[BACK]Return to uipc_socket.c CVS log [TXT][DIR] Up to [cvs.NetBSD.org] / src / sys / kern

Annotation of src/sys/kern/uipc_socket.c, Revision 1.235.2.3

1.235.2.3! skrll       1: /*     $NetBSD: uipc_socket.c,v 1.235.2.2 2015/06/06 14:40:22 skrll Exp $      */
1.64      thorpej     2:
                      3: /*-
1.188     ad          4:  * Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc.
1.64      thorpej     5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
1.188     ad          8:  * by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran.
1.64      thorpej     9:  *
                     10:  * Redistribution and use in source and binary forms, with or without
                     11:  * modification, are permitted provided that the following conditions
                     12:  * are met:
                     13:  * 1. Redistributions of source code must retain the above copyright
                     14:  *    notice, this list of conditions and the following disclaimer.
                     15:  * 2. Redistributions in binary form must reproduce the above copyright
                     16:  *    notice, this list of conditions and the following disclaimer in the
                     17:  *    documentation and/or other materials provided with the distribution.
                     18:  *
                     19:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     20:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     21:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     22:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     23:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     24:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     25:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     26:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     27:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     28:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     29:  * POSSIBILITY OF SUCH DAMAGE.
                     30:  */
1.16      cgd        31:
1.1       cgd        32: /*
1.159     ad         33:  * Copyright (c) 2004 The FreeBSD Foundation
                     34:  * Copyright (c) 2004 Robert Watson
1.15      mycroft    35:  * Copyright (c) 1982, 1986, 1988, 1990, 1993
                     36:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        37:  *
                     38:  * Redistribution and use in source and binary forms, with or without
                     39:  * modification, are permitted provided that the following conditions
                     40:  * are met:
                     41:  * 1. Redistributions of source code must retain the above copyright
                     42:  *    notice, this list of conditions and the following disclaimer.
                     43:  * 2. Redistributions in binary form must reproduce the above copyright
                     44:  *    notice, this list of conditions and the following disclaimer in the
                     45:  *    documentation and/or other materials provided with the distribution.
1.85      agc        46:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        47:  *    may be used to endorse or promote products derived from this software
                     48:  *    without specific prior written permission.
                     49:  *
                     50:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     51:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     52:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     53:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     54:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     55:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     56:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     57:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     58:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     59:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     60:  * SUCH DAMAGE.
                     61:  *
1.32      fvdl       62:  *     @(#)uipc_socket.c       8.6 (Berkeley) 5/2/95
1.1       cgd        63:  */
1.59      lukem      64:
1.222     rmind      65: /*
                     66:  * Socket operation routines.
                     67:  *
                     68:  * These routines are called by the routines in sys_socket.c or from a
                     69:  * system process, and implement the semantics of socket operations by
                     70:  * switching out to the protocol specific routines.
                     71:  */
                     72:
1.59      lukem      73: #include <sys/cdefs.h>
1.235.2.3! skrll      74: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.235.2.2 2015/06/06 14:40:22 skrll Exp $");
1.64      thorpej    75:
1.235.2.3! skrll      76: #ifdef _KERNEL_OPT
1.179     christos   77: #include "opt_compat_netbsd.h"
1.64      thorpej    78: #include "opt_sock_counters.h"
                     79: #include "opt_sosend_loan.h"
1.81      martin     80: #include "opt_mbuftrace.h"
1.84      ragge      81: #include "opt_somaxkva.h"
1.167     ad         82: #include "opt_multiprocessor.h"        /* XXX */
1.235.2.3! skrll      83: #endif
1.1       cgd        84:
1.9       mycroft    85: #include <sys/param.h>
                     86: #include <sys/systm.h>
                     87: #include <sys/proc.h>
                     88: #include <sys/file.h>
1.142     dyoung     89: #include <sys/filedesc.h>
1.173     plunky     90: #include <sys/kmem.h>
1.9       mycroft    91: #include <sys/mbuf.h>
                     92: #include <sys/domain.h>
                     93: #include <sys/kernel.h>
                     94: #include <sys/protosw.h>
                     95: #include <sys/socket.h>
                     96: #include <sys/socketvar.h>
1.21      christos   97: #include <sys/signalvar.h>
1.9       mycroft    98: #include <sys/resourcevar.h>
1.174     pooka      99: #include <sys/uidinfo.h>
1.72      jdolecek  100: #include <sys/event.h>
1.89      christos  101: #include <sys/poll.h>
1.118     elad      102: #include <sys/kauth.h>
1.136     ad        103: #include <sys/mutex.h>
                    104: #include <sys/condvar.h>
1.205     bouyer    105: #include <sys/kthread.h>
1.37      thorpej   106:
1.179     christos  107: #ifdef COMPAT_50
                    108: #include <compat/sys/time.h>
1.184     christos  109: #include <compat/sys/socket.h>
1.179     christos  110: #endif
                    111:
1.202     uebayasi  112: #include <uvm/uvm_extern.h>
                    113: #include <uvm/uvm_loan.h>
                    114: #include <uvm/uvm_page.h>
1.64      thorpej   115:
1.77      thorpej   116: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37      thorpej   117:
1.142     dyoung    118: extern const struct fileops socketops;
                    119:
1.54      lukem     120: extern int     somaxconn;                      /* patchable (XXX sysctl) */
                    121: int            somaxconn = SOMAXCONN;
1.160     ad        122: kmutex_t       *softnet_lock;
1.49      jonathan  123:
1.64      thorpej   124: #ifdef SOSEND_COUNTERS
                    125: #include <sys/device.h>
                    126:
1.113     thorpej   127: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   128:     NULL, "sosend", "loan big");
1.113     thorpej   129: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   130:     NULL, "sosend", "copy big");
1.113     thorpej   131: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   132:     NULL, "sosend", "copy small");
1.113     thorpej   133: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   134:     NULL, "sosend", "kva limit");
                    135:
                    136: #define        SOSEND_COUNTER_INCR(ev)         (ev)->ev_count++
                    137:
1.101     matt      138: EVCNT_ATTACH_STATIC(sosend_loan_big);
                    139: EVCNT_ATTACH_STATIC(sosend_copy_big);
                    140: EVCNT_ATTACH_STATIC(sosend_copy_small);
                    141: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64      thorpej   142: #else
                    143:
                    144: #define        SOSEND_COUNTER_INCR(ev)         /* nothing */
                    145:
                    146: #endif /* SOSEND_COUNTERS */
                    147:
1.167     ad        148: #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR)
1.121     yamt      149: int sock_loan_thresh = -1;
1.71      thorpej   150: #else
1.121     yamt      151: int sock_loan_thresh = 4096;
1.65      thorpej   152: #endif
1.64      thorpej   153:
1.136     ad        154: static kmutex_t so_pendfree_lock;
1.205     bouyer    155: static struct mbuf *so_pendfree = NULL;
1.64      thorpej   156:
1.84      ragge     157: #ifndef SOMAXKVA
                    158: #define        SOMAXKVA (16 * 1024 * 1024)
                    159: #endif
                    160: int somaxkva = SOMAXKVA;
1.113     thorpej   161: static int socurkva;
1.136     ad        162: static kcondvar_t socurkva_cv;
1.64      thorpej   163:
1.191     elad      164: static kauth_listener_t socket_listener;
                    165:
1.64      thorpej   166: #define        SOCK_LOAN_CHUNK         65536
                    167:
1.205     bouyer    168: static void sopendfree_thread(void *);
                    169: static kcondvar_t pendfree_thread_cv;
                    170: static lwp_t *sopendfree_lwp;
1.93      yamt      171:
1.212     pooka     172: static void sysctl_kern_socket_setup(void);
1.178     pooka     173: static struct sysctllog *socket_sysctllog;
                    174:
1.113     thorpej   175: static vsize_t
1.129     yamt      176: sokvareserve(struct socket *so, vsize_t len)
1.80      yamt      177: {
1.98      christos  178:        int error;
1.80      yamt      179:
1.136     ad        180:        mutex_enter(&so_pendfree_lock);
1.80      yamt      181:        while (socurkva + len > somaxkva) {
                    182:                SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136     ad        183:                error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98      christos  184:                if (error) {
                    185:                        len = 0;
                    186:                        break;
                    187:                }
1.80      yamt      188:        }
1.93      yamt      189:        socurkva += len;
1.136     ad        190:        mutex_exit(&so_pendfree_lock);
1.98      christos  191:        return len;
1.95      yamt      192: }
                    193:
1.113     thorpej   194: static void
1.95      yamt      195: sokvaunreserve(vsize_t len)
                    196: {
                    197:
1.136     ad        198:        mutex_enter(&so_pendfree_lock);
1.95      yamt      199:        socurkva -= len;
1.136     ad        200:        cv_broadcast(&socurkva_cv);
                    201:        mutex_exit(&so_pendfree_lock);
1.95      yamt      202: }
                    203:
                    204: /*
                    205:  * sokvaalloc: allocate kva for loan.
                    206:  */
                    207:
                    208: vaddr_t
1.209     matt      209: sokvaalloc(vaddr_t sva, vsize_t len, struct socket *so)
1.95      yamt      210: {
                    211:        vaddr_t lva;
                    212:
                    213:        /*
                    214:         * reserve kva.
                    215:         */
                    216:
1.98      christos  217:        if (sokvareserve(so, len) == 0)
                    218:                return 0;
1.93      yamt      219:
                    220:        /*
                    221:         * allocate kva.
                    222:         */
1.80      yamt      223:
1.209     matt      224:        lva = uvm_km_alloc(kernel_map, len, atop(sva) & uvmexp.colormask,
                    225:            UVM_KMF_COLORMATCH | UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95      yamt      226:        if (lva == 0) {
                    227:                sokvaunreserve(len);
1.80      yamt      228:                return (0);
1.95      yamt      229:        }
1.80      yamt      230:
                    231:        return lva;
                    232: }
                    233:
1.93      yamt      234: /*
                    235:  * sokvafree: free kva for loan.
                    236:  */
                    237:
1.80      yamt      238: void
                    239: sokvafree(vaddr_t sva, vsize_t len)
                    240: {
1.93      yamt      241:
                    242:        /*
                    243:         * free kva.
                    244:         */
1.80      yamt      245:
1.109     yamt      246:        uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93      yamt      247:
                    248:        /*
                    249:         * unreserve kva.
                    250:         */
                    251:
1.95      yamt      252:        sokvaunreserve(len);
1.80      yamt      253: }
                    254:
1.64      thorpej   255: static void
1.134     christos  256: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64      thorpej   257: {
1.156     yamt      258:        vaddr_t sva, eva;
1.64      thorpej   259:        vsize_t len;
1.156     yamt      260:        int npgs;
                    261:
                    262:        KASSERT(pgs != NULL);
1.64      thorpej   263:
                    264:        eva = round_page((vaddr_t) buf + size);
                    265:        sva = trunc_page((vaddr_t) buf);
                    266:        len = eva - sva;
                    267:        npgs = len >> PAGE_SHIFT;
                    268:
                    269:        pmap_kremove(sva, len);
                    270:        pmap_update(pmap_kernel());
                    271:        uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80      yamt      272:        sokvafree(sva, len);
1.64      thorpej   273: }
                    274:
1.93      yamt      275: /*
1.205     bouyer    276:  * sopendfree_thread: free mbufs on "pendfree" list.
1.136     ad        277:  * unlock and relock so_pendfree_lock when freeing mbufs.
1.93      yamt      278:  */
                    279:
1.205     bouyer    280: static void
                    281: sopendfree_thread(void *v)
1.93      yamt      282: {
1.137     ad        283:        struct mbuf *m, *next;
1.205     bouyer    284:        size_t rv;
1.93      yamt      285:
1.205     bouyer    286:        mutex_enter(&so_pendfree_lock);
1.64      thorpej   287:
1.205     bouyer    288:        for (;;) {
                    289:                rv = 0;
                    290:                while (so_pendfree != NULL) {
                    291:                        m = so_pendfree;
                    292:                        so_pendfree = NULL;
                    293:                        mutex_exit(&so_pendfree_lock);
                    294:
                    295:                        for (; m != NULL; m = next) {
                    296:                                next = m->m_next;
                    297:                                KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
                    298:                                KASSERT(m->m_ext.ext_refcnt == 0);
                    299:
                    300:                                rv += m->m_ext.ext_size;
                    301:                                sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
                    302:                                    m->m_ext.ext_size);
                    303:                                pool_cache_put(mb_cache, m);
                    304:                        }
1.93      yamt      305:
1.205     bouyer    306:                        mutex_enter(&so_pendfree_lock);
1.93      yamt      307:                }
1.205     bouyer    308:                if (rv)
                    309:                        cv_broadcast(&socurkva_cv);
                    310:                cv_wait(&pendfree_thread_cv, &so_pendfree_lock);
1.64      thorpej   311:        }
1.205     bouyer    312:        panic("sopendfree_thread");
                    313:        /* NOTREACHED */
1.64      thorpej   314: }
                    315:
1.80      yamt      316: void
1.134     christos  317: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64      thorpej   318: {
                    319:
1.156     yamt      320:        KASSERT(m != NULL);
1.64      thorpej   321:
1.93      yamt      322:        /*
                    323:         * postpone freeing mbuf.
                    324:         *
                    325:         * we can't do it in interrupt context
                    326:         * because we need to put kva back to kernel_map.
                    327:         */
                    328:
1.136     ad        329:        mutex_enter(&so_pendfree_lock);
1.92      yamt      330:        m->m_next = so_pendfree;
                    331:        so_pendfree = m;
1.205     bouyer    332:        cv_signal(&pendfree_thread_cv);
1.136     ad        333:        mutex_exit(&so_pendfree_lock);
1.64      thorpej   334: }
                    335:
                    336: static long
                    337: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
                    338: {
                    339:        struct iovec *iov = uio->uio_iov;
                    340:        vaddr_t sva, eva;
                    341:        vsize_t len;
1.156     yamt      342:        vaddr_t lva;
                    343:        int npgs, error;
                    344:        vaddr_t va;
                    345:        int i;
1.64      thorpej   346:
1.116     yamt      347:        if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64      thorpej   348:                return (0);
                    349:
                    350:        if (iov->iov_len < (size_t) space)
                    351:                space = iov->iov_len;
                    352:        if (space > SOCK_LOAN_CHUNK)
                    353:                space = SOCK_LOAN_CHUNK;
                    354:
                    355:        eva = round_page((vaddr_t) iov->iov_base + space);
                    356:        sva = trunc_page((vaddr_t) iov->iov_base);
                    357:        len = eva - sva;
                    358:        npgs = len >> PAGE_SHIFT;
                    359:
1.79      thorpej   360:        KASSERT(npgs <= M_EXT_MAXPAGES);
                    361:
1.209     matt      362:        lva = sokvaalloc(sva, len, so);
1.64      thorpej   363:        if (lva == 0)
1.80      yamt      364:                return 0;
1.64      thorpej   365:
1.116     yamt      366:        error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79      thorpej   367:            m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64      thorpej   368:        if (error) {
1.80      yamt      369:                sokvafree(lva, len);
1.64      thorpej   370:                return (0);
                    371:        }
                    372:
                    373:        for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79      thorpej   374:                pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
1.194     cegger    375:                    VM_PROT_READ, 0);
1.64      thorpej   376:        pmap_update(pmap_kernel());
                    377:
                    378:        lva += (vaddr_t) iov->iov_base & PAGE_MASK;
                    379:
1.134     christos  380:        MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79      thorpej   381:        m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64      thorpej   382:
                    383:        uio->uio_resid -= space;
                    384:        /* uio_offset not updated, not set/used for write(2) */
1.134     christos  385:        uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64      thorpej   386:        uio->uio_iov->iov_len -= space;
                    387:        if (uio->uio_iov->iov_len == 0) {
                    388:                uio->uio_iov++;
                    389:                uio->uio_iovcnt--;
                    390:        }
                    391:
                    392:        return (space);
                    393: }
                    394:
1.142     dyoung    395: struct mbuf *
1.147     dyoung    396: getsombuf(struct socket *so, int type)
1.142     dyoung    397: {
                    398:        struct mbuf *m;
                    399:
1.147     dyoung    400:        m = m_get(M_WAIT, type);
1.142     dyoung    401:        MCLAIM(m, so->so_mowner);
                    402:        return m;
                    403: }
                    404:
1.191     elad      405: static int
                    406: socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
                    407:     void *arg0, void *arg1, void *arg2, void *arg3)
                    408: {
                    409:        int result;
                    410:        enum kauth_network_req req;
                    411:
                    412:        result = KAUTH_RESULT_DEFER;
                    413:        req = (enum kauth_network_req)arg0;
                    414:
1.193     elad      415:        if ((action != KAUTH_NETWORK_SOCKET) &&
                    416:            (action != KAUTH_NETWORK_BIND))
1.191     elad      417:                return result;
                    418:
                    419:        switch (req) {
1.193     elad      420:        case KAUTH_REQ_NETWORK_BIND_PORT:
                    421:                result = KAUTH_RESULT_ALLOW;
                    422:                break;
                    423:
1.191     elad      424:        case KAUTH_REQ_NETWORK_SOCKET_DROP: {
                    425:                /* Normal users can only drop their own connections. */
                    426:                struct socket *so = (struct socket *)arg1;
                    427:
1.220     christos  428:                if (so->so_cred && proc_uidmatch(cred, so->so_cred) == 0)
1.191     elad      429:                        result = KAUTH_RESULT_ALLOW;
                    430:
                    431:                break;
                    432:                }
                    433:
                    434:        case KAUTH_REQ_NETWORK_SOCKET_OPEN:
                    435:                /* We allow "raw" routing/bluetooth sockets to anyone. */
1.203     matt      436:                if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_OROUTE
                    437:                    || (u_long)arg1 == PF_BLUETOOTH) {
1.191     elad      438:                        result = KAUTH_RESULT_ALLOW;
1.203     matt      439:                } else {
1.191     elad      440:                        /* Privileged, let secmodel handle this. */
                    441:                        if ((u_long)arg2 == SOCK_RAW)
                    442:                                break;
                    443:                }
                    444:
                    445:                result = KAUTH_RESULT_ALLOW;
                    446:
                    447:                break;
                    448:
1.192     elad      449:        case KAUTH_REQ_NETWORK_SOCKET_CANSEE:
                    450:                result = KAUTH_RESULT_ALLOW;
                    451:
                    452:                break;
                    453:
1.191     elad      454:        default:
                    455:                break;
                    456:        }
                    457:
                    458:        return result;
                    459: }
                    460:
1.119     yamt      461: void
                    462: soinit(void)
                    463: {
                    464:
1.212     pooka     465:        sysctl_kern_socket_setup();
1.178     pooka     466:
1.148     ad        467:        mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.160     ad        468:        softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.136     ad        469:        cv_init(&socurkva_cv, "sokva");
1.205     bouyer    470:        cv_init(&pendfree_thread_cv, "sopendfr");
1.166     ad        471:        soinit2();
1.136     ad        472:
1.119     yamt      473:        /* Set the initial adjusted socket buffer size. */
                    474:        if (sb_max_set(sb_max))
                    475:                panic("bad initial sb_max value: %lu", sb_max);
                    476:
1.191     elad      477:        socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
                    478:            socket_listener_cb, NULL);
1.119     yamt      479: }
                    480:
1.205     bouyer    481: void
                    482: soinit1(void)
                    483: {
                    484:        int error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL,
                    485:            sopendfree_thread, NULL, &sopendfree_lwp, "sopendfree");
                    486:        if (error)
                    487:                panic("soinit1 %d", error);
                    488: }
                    489:
1.1       cgd       490: /*
1.222     rmind     491:  * socreate: create a new socket of the specified type and the protocol.
                    492:  *
                    493:  * => Caller may specify another socket for lock sharing (must not be held).
                    494:  * => Returns the new socket without lock held.
1.224     rmind     495:  */
1.3       andrew    496: int
1.160     ad        497: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l,
                    498:         struct socket *lockso)
1.1       cgd       499: {
1.99      matt      500:        const struct protosw    *prp;
1.54      lukem     501:        struct socket   *so;
1.115     yamt      502:        uid_t           uid;
1.160     ad        503:        int             error;
                    504:        kmutex_t        *lock;
1.1       cgd       505:
1.132     elad      506:        error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
                    507:            KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
                    508:            KAUTH_ARG(proto));
1.140     dyoung    509:        if (error != 0)
                    510:                return error;
1.127     elad      511:
1.1       cgd       512:        if (proto)
                    513:                prp = pffindproto(dom, proto, type);
                    514:        else
                    515:                prp = pffindtype(dom, type);
1.140     dyoung    516:        if (prp == NULL) {
1.120     ginsbach  517:                /* no support for domain */
                    518:                if (pffinddomain(dom) == 0)
1.140     dyoung    519:                        return EAFNOSUPPORT;
1.120     ginsbach  520:                /* no support for socket type */
                    521:                if (proto == 0 && type != 0)
1.140     dyoung    522:                        return EPROTOTYPE;
                    523:                return EPROTONOSUPPORT;
1.120     ginsbach  524:        }
1.223     rmind     525:        if (prp->pr_usrreqs == NULL)
1.140     dyoung    526:                return EPROTONOSUPPORT;
1.1       cgd       527:        if (prp->pr_type != type)
1.140     dyoung    528:                return EPROTOTYPE;
1.160     ad        529:
                    530:        so = soget(true);
1.1       cgd       531:        so->so_type = type;
                    532:        so->so_proto = prp;
1.33      matt      533:        so->so_send = sosend;
                    534:        so->so_receive = soreceive;
1.78      matt      535: #ifdef MBUFTRACE
                    536:        so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
                    537:        so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
                    538:        so->so_mowner = &prp->pr_domain->dom_mowner;
                    539: #endif
1.138     rmind     540:        uid = kauth_cred_geteuid(l->l_cred);
1.115     yamt      541:        so->so_uidinfo = uid_find(uid);
1.168     yamt      542:        so->so_cpid = l->l_proc->p_pid;
1.224     rmind     543:
                    544:        /*
                    545:         * Lock assigned and taken during PCB attach, unless we share
                    546:         * the lock with another socket, e.g. socketpair(2) case.
                    547:         */
                    548:        if (lockso) {
1.160     ad        549:                lock = lockso->so_lock;
                    550:                so->so_lock = lock;
                    551:                mutex_obj_hold(lock);
                    552:                mutex_enter(lock);
                    553:        }
1.224     rmind     554:
                    555:        /* Attach the PCB (returns with the socket lock held). */
                    556:        error = (*prp->pr_usrreqs->pr_attach)(so, proto);
1.160     ad        557:        KASSERT(solocked(so));
1.224     rmind     558:
                    559:        if (error) {
1.222     rmind     560:                KASSERT(so->so_pcb == NULL);
1.1       cgd       561:                so->so_state |= SS_NOFDREF;
                    562:                sofree(so);
1.140     dyoung    563:                return error;
1.1       cgd       564:        }
1.198     elad      565:        so->so_cred = kauth_cred_dup(l->l_cred);
1.160     ad        566:        sounlock(so);
1.224     rmind     567:
1.1       cgd       568:        *aso = so;
1.140     dyoung    569:        return 0;
1.1       cgd       570: }
                    571:
1.222     rmind     572: /*
                    573:  * fsocreate: create a socket and a file descriptor associated with it.
                    574:  *
                    575:  * => On success, write file descriptor to fdout and return zero.
                    576:  * => On failure, return non-zero; *fdout will be undefined.
1.142     dyoung    577:  */
                    578: int
1.222     rmind     579: fsocreate(int domain, struct socket **sop, int type, int proto, int *fdout)
1.142     dyoung    580: {
1.222     rmind     581:        lwp_t *l = curlwp;
                    582:        int error, fd, flags;
                    583:        struct socket *so;
                    584:        struct file *fp;
1.142     dyoung    585:
1.222     rmind     586:        if ((error = fd_allocfile(&fp, &fd)) != 0) {
1.204     christos  587:                return error;
1.222     rmind     588:        }
                    589:        flags = type & SOCK_FLAGS_MASK;
1.204     christos  590:        fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0);
1.207     christos  591:        fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
                    592:            ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
1.142     dyoung    593:        fp->f_type = DTYPE_SOCKET;
                    594:        fp->f_ops = &socketops;
1.222     rmind     595:
                    596:        type &= ~SOCK_FLAGS_MASK;
                    597:        error = socreate(domain, &so, type, proto, l, NULL);
                    598:        if (error) {
1.155     ad        599:                fd_abort(curproc, fp, fd);
1.222     rmind     600:                return error;
                    601:        }
                    602:        if (flags & SOCK_NONBLOCK) {
                    603:                so->so_state |= SS_NBIO;
                    604:        }
1.235     matt      605:        fp->f_socket = so;
1.222     rmind     606:        fd_affix(curproc, fp, fd);
                    607:
                    608:        if (sop != NULL) {
                    609:                *sop = so;
1.142     dyoung    610:        }
1.222     rmind     611:        *fdout = fd;
1.142     dyoung    612:        return error;
                    613: }
                    614:
1.3       andrew    615: int
1.190     dyoung    616: sofamily(const struct socket *so)
                    617: {
                    618:        const struct protosw *pr;
                    619:        const struct domain *dom;
                    620:
                    621:        if ((pr = so->so_proto) == NULL)
                    622:                return AF_UNSPEC;
                    623:        if ((dom = pr->pr_domain) == NULL)
                    624:                return AF_UNSPEC;
                    625:        return dom->dom_family;
                    626: }
                    627:
                    628: int
1.235.2.1  skrll     629: sobind(struct socket *so, struct sockaddr *nam, struct lwp *l)
1.1       cgd       630: {
1.160     ad        631:        int     error;
1.1       cgd       632:
1.160     ad        633:        solock(so);
1.235.2.1  skrll     634:        if (nam->sa_family != so->so_proto->pr_domain->dom_family) {
                    635:                sounlock(so);
                    636:                return EAFNOSUPPORT;
                    637:        }
1.231     rtr       638:        error = (*so->so_proto->pr_usrreqs->pr_bind)(so, nam, l);
1.160     ad        639:        sounlock(so);
1.140     dyoung    640:        return error;
1.1       cgd       641: }
                    642:
1.3       andrew    643: int
1.150     elad      644: solisten(struct socket *so, int backlog, struct lwp *l)
1.1       cgd       645: {
1.160     ad        646:        int     error;
1.1       cgd       647:
1.160     ad        648:        solock(so);
1.235.2.2  skrll     649:        if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
1.163     ad        650:            SS_ISDISCONNECTING)) != 0) {
1.222     rmind     651:                sounlock(so);
                    652:                return EINVAL;
1.163     ad        653:        }
1.231     rtr       654:        error = (*so->so_proto->pr_usrreqs->pr_listen)(so, l);
1.140     dyoung    655:        if (error != 0) {
1.160     ad        656:                sounlock(so);
1.140     dyoung    657:                return error;
1.1       cgd       658:        }
1.63      matt      659:        if (TAILQ_EMPTY(&so->so_q))
1.1       cgd       660:                so->so_options |= SO_ACCEPTCONN;
                    661:        if (backlog < 0)
                    662:                backlog = 0;
1.49      jonathan  663:        so->so_qlimit = min(backlog, somaxconn);
1.160     ad        664:        sounlock(so);
1.140     dyoung    665:        return 0;
1.1       cgd       666: }
                    667:
1.21      christos  668: void
1.54      lukem     669: sofree(struct socket *so)
1.1       cgd       670: {
1.161     ad        671:        u_int refs;
1.1       cgd       672:
1.160     ad        673:        KASSERT(solocked(so));
                    674:
                    675:        if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) {
                    676:                sounlock(so);
1.1       cgd       677:                return;
1.160     ad        678:        }
1.43      mycroft   679:        if (so->so_head) {
                    680:                /*
                    681:                 * We must not decommission a socket that's on the accept(2)
                    682:                 * queue.  If we do, then accept(2) may hang after select(2)
                    683:                 * indicated that the listening socket was ready.
                    684:                 */
1.160     ad        685:                if (!soqremque(so, 0)) {
                    686:                        sounlock(so);
1.43      mycroft   687:                        return;
1.160     ad        688:                }
1.43      mycroft   689:        }
1.98      christos  690:        if (so->so_rcv.sb_hiwat)
1.110     christos  691:                (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98      christos  692:                    RLIM_INFINITY);
                    693:        if (so->so_snd.sb_hiwat)
1.110     christos  694:                (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98      christos  695:                    RLIM_INFINITY);
                    696:        sbrelease(&so->so_snd, so);
1.160     ad        697:        KASSERT(!cv_has_waiters(&so->so_cv));
                    698:        KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv));
                    699:        KASSERT(!cv_has_waiters(&so->so_snd.sb_cv));
1.1       cgd       700:        sorflush(so);
1.161     ad        701:        refs = so->so_aborting; /* XXX */
1.177     ad        702:        /* Remove acccept filter if one is present. */
1.170     tls       703:        if (so->so_accf != NULL)
1.177     ad        704:                (void)accept_filt_clear(so);
1.160     ad        705:        sounlock(so);
1.161     ad        706:        if (refs == 0)          /* XXX */
                    707:                soput(so);
1.1       cgd       708: }
                    709:
                    710: /*
1.222     rmind     711:  * soclose: close a socket on last file table reference removal.
                    712:  * Initiate disconnect if connected.  Free socket when disconnect complete.
1.1       cgd       713:  */
1.3       andrew    714: int
1.54      lukem     715: soclose(struct socket *so)
1.1       cgd       716: {
1.222     rmind     717:        struct socket *so2;
                    718:        int error = 0;
1.1       cgd       719:
1.160     ad        720:        solock(so);
1.1       cgd       721:        if (so->so_options & SO_ACCEPTCONN) {
1.172     ad        722:                for (;;) {
                    723:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.160     ad        724:                                KASSERT(solocked2(so, so2));
                    725:                                (void) soqremque(so2, 0);
                    726:                                /* soabort drops the lock. */
                    727:                                (void) soabort(so2);
                    728:                                solock(so);
1.172     ad        729:                                continue;
1.160     ad        730:                        }
1.172     ad        731:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.160     ad        732:                                KASSERT(solocked2(so, so2));
                    733:                                (void) soqremque(so2, 1);
                    734:                                /* soabort drops the lock. */
                    735:                                (void) soabort(so2);
                    736:                                solock(so);
1.172     ad        737:                                continue;
1.160     ad        738:                        }
1.172     ad        739:                        break;
                    740:                }
1.1       cgd       741:        }
1.222     rmind     742:        if (so->so_pcb == NULL)
1.1       cgd       743:                goto discard;
                    744:        if (so->so_state & SS_ISCONNECTED) {
                    745:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    746:                        error = sodisconnect(so);
                    747:                        if (error)
                    748:                                goto drop;
                    749:                }
                    750:                if (so->so_options & SO_LINGER) {
1.206     christos  751:                        if ((so->so_state & (SS_ISDISCONNECTING|SS_NBIO)) ==
                    752:                            (SS_ISDISCONNECTING|SS_NBIO))
1.1       cgd       753:                                goto drop;
1.21      christos  754:                        while (so->so_state & SS_ISCONNECTED) {
1.185     yamt      755:                                error = sowait(so, true, so->so_linger * hz);
1.21      christos  756:                                if (error)
1.1       cgd       757:                                        break;
1.21      christos  758:                        }
1.1       cgd       759:                }
                    760:        }
1.54      lukem     761:  drop:
1.1       cgd       762:        if (so->so_pcb) {
1.224     rmind     763:                KASSERT(solocked(so));
                    764:                (*so->so_proto->pr_usrreqs->pr_detach)(so);
1.1       cgd       765:        }
1.54      lukem     766:  discard:
1.222     rmind     767:        KASSERT((so->so_state & SS_NOFDREF) == 0);
1.198     elad      768:        kauth_cred_free(so->so_cred);
1.1       cgd       769:        so->so_state |= SS_NOFDREF;
                    770:        sofree(so);
1.222     rmind     771:        return error;
1.1       cgd       772: }
                    773:
                    774: /*
1.160     ad        775:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       776:  */
1.3       andrew    777: int
1.54      lukem     778: soabort(struct socket *so)
1.1       cgd       779: {
1.161     ad        780:        u_int refs;
1.139     yamt      781:        int error;
1.235.2.2  skrll     782:
1.160     ad        783:        KASSERT(solocked(so));
                    784:        KASSERT(so->so_head == NULL);
1.1       cgd       785:
1.161     ad        786:        so->so_aborting++;              /* XXX */
1.230     mrg       787:        error = (*so->so_proto->pr_usrreqs->pr_abort)(so);
1.161     ad        788:        refs = --so->so_aborting;       /* XXX */
1.164     drochner  789:        if (error || (refs == 0)) {
1.139     yamt      790:                sofree(so);
1.160     ad        791:        } else {
                    792:                sounlock(so);
1.139     yamt      793:        }
                    794:        return error;
1.1       cgd       795: }
                    796:
1.3       andrew    797: int
1.235.2.2  skrll     798: soaccept(struct socket *so, struct sockaddr *nam)
1.1       cgd       799: {
1.222     rmind     800:        int error;
1.160     ad        801:
                    802:        KASSERT(solocked(so));
1.222     rmind     803:        KASSERT((so->so_state & SS_NOFDREF) != 0);
1.1       cgd       804:
                    805:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   806:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    807:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.225     rtr       808:                error = (*so->so_proto->pr_usrreqs->pr_accept)(so, nam);
1.41      mycroft   809:        else
1.53      itojun    810:                error = ECONNABORTED;
1.52      itojun    811:
1.222     rmind     812:        return error;
1.1       cgd       813: }
                    814:
1.3       andrew    815: int
1.235.2.2  skrll     816: soconnect(struct socket *so, struct sockaddr *nam, struct lwp *l)
1.1       cgd       817: {
1.222     rmind     818:        int error;
1.160     ad        819:
                    820:        KASSERT(solocked(so));
1.1       cgd       821:
                    822:        if (so->so_options & SO_ACCEPTCONN)
1.222     rmind     823:                return EOPNOTSUPP;
1.1       cgd       824:        /*
                    825:         * If protocol is connection-based, can only connect once.
                    826:         * Otherwise, if connected, try to disconnect first.
                    827:         * This allows user to disconnect by connecting to, e.g.,
                    828:         * a null address.
                    829:         */
                    830:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    831:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
1.235.2.2  skrll     832:            (error = sodisconnect(so)))) {
1.1       cgd       833:                error = EISCONN;
1.235.2.2  skrll     834:        } else {
                    835:                if (nam->sa_family != so->so_proto->pr_domain->dom_family) {
                    836:                        return EAFNOSUPPORT;
                    837:                }
1.231     rtr       838:                error = (*so->so_proto->pr_usrreqs->pr_connect)(so, nam, l);
1.235.2.2  skrll     839:        }
1.222     rmind     840:
                    841:        return error;
1.1       cgd       842: }
                    843:
1.3       andrew    844: int
1.54      lukem     845: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       846: {
1.160     ad        847:        KASSERT(solocked2(so1, so2));
1.1       cgd       848:
1.234     rtr       849:        return (*so1->so_proto->pr_usrreqs->pr_connect2)(so1, so2);
1.1       cgd       850: }
                    851:
1.3       andrew    852: int
1.54      lukem     853: sodisconnect(struct socket *so)
1.1       cgd       854: {
1.160     ad        855:        int     error;
                    856:
                    857:        KASSERT(solocked(so));
1.1       cgd       858:
                    859:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    860:                error = ENOTCONN;
1.160     ad        861:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       862:                error = EALREADY;
1.160     ad        863:        } else {
1.229     rtr       864:                error = (*so->so_proto->pr_usrreqs->pr_disconnect)(so);
1.1       cgd       865:        }
                    866:        return (error);
                    867: }
                    868:
1.15      mycroft   869: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       870: /*
                    871:  * Send on a socket.
                    872:  * If send must go all at once and message is larger than
                    873:  * send buffering, then hard error.
                    874:  * Lock against other senders.
                    875:  * If must go all at once and not enough room now, then
                    876:  * inform user that this would block and do nothing.
                    877:  * Otherwise, if nonblocking, send as much as possible.
                    878:  * The data to be sent is described by "uio" if nonzero,
                    879:  * otherwise by the mbuf chain "top" (which must be null
                    880:  * if uio is not).  Data provided in mbuf chain must be small
                    881:  * enough to send all at once.
                    882:  *
                    883:  * Returns nonzero on error, timeout or signal; callers
                    884:  * must check for short counts if EINTR/ERESTART are returned.
                    885:  * Data and control buffers are freed on return.
                    886:  */
1.3       andrew    887: int
1.235.2.2  skrll     888: sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
                    889:        struct mbuf *top, struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       890: {
1.54      lukem     891:        struct mbuf     **mp, *m;
1.58      jdolecek  892:        long            space, len, resid, clen, mlen;
                    893:        int             error, s, dontroute, atomic;
1.196     dsl       894:        short           wakeup_state = 0;
1.54      lukem     895:
1.160     ad        896:        clen = 0;
1.64      thorpej   897:
1.160     ad        898:        /*
                    899:         * solock() provides atomicity of access.  splsoftnet() prevents
                    900:         * protocol processing soft interrupts from interrupting us and
                    901:         * blocking (expensive).
                    902:         */
                    903:        s = splsoftnet();
                    904:        solock(so);
1.54      lukem     905:        atomic = sosendallatonce(so) || top;
1.1       cgd       906:        if (uio)
                    907:                resid = uio->uio_resid;
                    908:        else
                    909:                resid = top->m_pkthdr.len;
1.7       cgd       910:        /*
                    911:         * In theory resid should be unsigned.
                    912:         * However, space must be signed, as it might be less than 0
                    913:         * if we over-committed, and we must use a signed comparison
                    914:         * of space and resid.  On the other hand, a negative resid
                    915:         * causes us to loop sending 0-length segments to the protocol.
                    916:         */
1.29      mycroft   917:        if (resid < 0) {
                    918:                error = EINVAL;
                    919:                goto out;
                    920:        }
1.1       cgd       921:        dontroute =
                    922:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    923:            (so->so_proto->pr_flags & PR_ATOMIC);
1.165     christos  924:        l->l_ru.ru_msgsnd++;
1.1       cgd       925:        if (control)
                    926:                clen = control->m_len;
1.54      lukem     927:  restart:
1.21      christos  928:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       929:                goto out;
                    930:        do {
1.160     ad        931:                if (so->so_state & SS_CANTSENDMORE) {
                    932:                        error = EPIPE;
                    933:                        goto release;
                    934:                }
1.48      thorpej   935:                if (so->so_error) {
                    936:                        error = so->so_error;
                    937:                        so->so_error = 0;
                    938:                        goto release;
                    939:                }
1.1       cgd       940:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    941:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
1.217     rmind     942:                                if (resid || clen == 0) {
1.160     ad        943:                                        error = ENOTCONN;
                    944:                                        goto release;
                    945:                                }
1.235.2.2  skrll     946:                        } else if (addr == NULL) {
1.160     ad        947:                                error = EDESTADDRREQ;
                    948:                                goto release;
                    949:                        }
1.1       cgd       950:                }
                    951:                space = sbspace(&so->so_snd);
                    952:                if (flags & MSG_OOB)
                    953:                        space += 1024;
1.21      christos  954:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        955:                    clen > so->so_snd.sb_hiwat) {
                    956:                        error = EMSGSIZE;
                    957:                        goto release;
                    958:                }
1.96      mycroft   959:                if (space < resid + clen &&
1.1       cgd       960:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.206     christos  961:                        if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) {
1.160     ad        962:                                error = EWOULDBLOCK;
                    963:                                goto release;
                    964:                        }
1.1       cgd       965:                        sbunlock(&so->so_snd);
1.196     dsl       966:                        if (wakeup_state & SS_RESTARTSYS) {
                    967:                                error = ERESTART;
                    968:                                goto out;
                    969:                        }
1.1       cgd       970:                        error = sbwait(&so->so_snd);
                    971:                        if (error)
                    972:                                goto out;
1.196     dsl       973:                        wakeup_state = so->so_state;
1.1       cgd       974:                        goto restart;
                    975:                }
1.196     dsl       976:                wakeup_state = 0;
1.1       cgd       977:                mp = &top;
                    978:                space -= clen;
                    979:                do {
1.45      tv        980:                        if (uio == NULL) {
                    981:                                /*
                    982:                                 * Data is prepackaged in "top".
                    983:                                 */
                    984:                                resid = 0;
                    985:                                if (flags & MSG_EOR)
                    986:                                        top->m_flags |= M_EOR;
                    987:                        } else do {
1.160     ad        988:                                sounlock(so);
                    989:                                splx(s);
1.144     dyoung    990:                                if (top == NULL) {
1.78      matt      991:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv        992:                                        mlen = MHLEN;
                    993:                                        m->m_pkthdr.len = 0;
1.140     dyoung    994:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv        995:                                } else {
1.78      matt      996:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv        997:                                        mlen = MLEN;
                    998:                                }
1.78      matt      999:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt     1000:                                if (sock_loan_thresh >= 0 &&
                   1001:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                   1002:                                    space >= sock_loan_thresh &&
1.64      thorpej  1003:                                    (len = sosend_loan(so, uio, m,
                   1004:                                                       space)) != 0) {
                   1005:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                   1006:                                        space -= len;
                   1007:                                        goto have_data;
                   1008:                                }
1.45      tv       1009:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej  1010:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.201     oki      1011:                                        m_clget(m, M_DONTWAIT);
1.45      tv       1012:                                        if ((m->m_flags & M_EXT) == 0)
                   1013:                                                goto nopages;
                   1014:                                        mlen = MCLBYTES;
                   1015:                                        if (atomic && top == 0) {
1.58      jdolecek 1016:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem    1017:                                                    resid);
1.45      tv       1018:                                                m->m_data += max_hdr;
                   1019:                                        } else
1.58      jdolecek 1020:                                                len = lmin(MCLBYTES, resid);
1.45      tv       1021:                                        space -= len;
                   1022:                                } else {
1.64      thorpej  1023:  nopages:
                   1024:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek 1025:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv       1026:                                        space -= len;
                   1027:                                        /*
                   1028:                                         * For datagram protocols, leave room
                   1029:                                         * for protocol headers in first mbuf.
                   1030:                                         */
                   1031:                                        if (atomic && top == 0 && len < mlen)
                   1032:                                                MH_ALIGN(m, len);
                   1033:                                }
1.144     dyoung   1034:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej  1035:  have_data:
1.45      tv       1036:                                resid = uio->uio_resid;
                   1037:                                m->m_len = len;
                   1038:                                *mp = m;
                   1039:                                top->m_pkthdr.len += len;
1.160     ad       1040:                                s = splsoftnet();
                   1041:                                solock(so);
1.144     dyoung   1042:                                if (error != 0)
1.45      tv       1043:                                        goto release;
                   1044:                                mp = &m->m_next;
                   1045:                                if (resid <= 0) {
                   1046:                                        if (flags & MSG_EOR)
                   1047:                                                top->m_flags |= M_EOR;
                   1048:                                        break;
                   1049:                                }
                   1050:                        } while (space > 0 && atomic);
1.108     perry    1051:
1.160     ad       1052:                        if (so->so_state & SS_CANTSENDMORE) {
                   1053:                                error = EPIPE;
                   1054:                                goto release;
                   1055:                        }
1.45      tv       1056:                        if (dontroute)
                   1057:                                so->so_options |= SO_DONTROUTE;
                   1058:                        if (resid > 0)
                   1059:                                so->so_state |= SS_MORETOCOME;
1.235.2.2  skrll    1060:                        if (flags & MSG_OOB) {
1.226     rtr      1061:                                error = (*so->so_proto->pr_usrreqs->pr_sendoob)(so,
                   1062:                                    top, control);
1.235.2.2  skrll    1063:                        } else {
1.232     rtr      1064:                                error = (*so->so_proto->pr_usrreqs->pr_send)(so,
                   1065:                                    top, addr, control, l);
1.235.2.2  skrll    1066:                        }
1.45      tv       1067:                        if (dontroute)
                   1068:                                so->so_options &= ~SO_DONTROUTE;
                   1069:                        if (resid > 0)
                   1070:                                so->so_state &= ~SS_MORETOCOME;
                   1071:                        clen = 0;
1.144     dyoung   1072:                        control = NULL;
                   1073:                        top = NULL;
1.45      tv       1074:                        mp = &top;
1.144     dyoung   1075:                        if (error != 0)
1.1       cgd      1076:                                goto release;
                   1077:                } while (resid && space > 0);
                   1078:        } while (resid);
                   1079:
1.54      lukem    1080:  release:
1.1       cgd      1081:        sbunlock(&so->so_snd);
1.54      lukem    1082:  out:
1.160     ad       1083:        sounlock(so);
                   1084:        splx(s);
1.1       cgd      1085:        if (top)
                   1086:                m_freem(top);
                   1087:        if (control)
                   1088:                m_freem(control);
                   1089:        return (error);
                   1090: }
                   1091:
                   1092: /*
1.159     ad       1093:  * Following replacement or removal of the first mbuf on the first
                   1094:  * mbuf chain of a socket buffer, push necessary state changes back
                   1095:  * into the socket buffer so that other consumers see the values
                   1096:  * consistently.  'nextrecord' is the callers locally stored value of
                   1097:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1098:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1099:  */
                   1100: static void
                   1101: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1102: {
                   1103:
1.160     ad       1104:        KASSERT(solocked(sb->sb_so));
                   1105:
1.159     ad       1106:        /*
                   1107:         * First, update for the new value of nextrecord.  If necessary,
                   1108:         * make it the first record.
                   1109:         */
                   1110:        if (sb->sb_mb != NULL)
                   1111:                sb->sb_mb->m_nextpkt = nextrecord;
                   1112:        else
                   1113:                sb->sb_mb = nextrecord;
                   1114:
                   1115:         /*
                   1116:          * Now update any dependent socket buffer fields to reflect
                   1117:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1118:          * the addition of a second clause that takes care of the
                   1119:          * case where sb_mb has been updated, but remains the last
                   1120:          * record.
                   1121:          */
                   1122:         if (sb->sb_mb == NULL) {
                   1123:                 sb->sb_mbtail = NULL;
                   1124:                 sb->sb_lastrecord = NULL;
                   1125:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1126:                 sb->sb_lastrecord = sb->sb_mb;
                   1127: }
                   1128:
                   1129: /*
1.1       cgd      1130:  * Implement receive operations on a socket.
                   1131:  * We depend on the way that records are added to the sockbuf
                   1132:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1133:  * must begin with an address if the protocol so specifies,
                   1134:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1135:  * and then zero or more mbufs of data.
                   1136:  * In order to avoid blocking network interrupts for the entire time here,
                   1137:  * we splx() while doing the actual copy to user space.
                   1138:  * Although the sockbuf is locked, new data may still be appended,
                   1139:  * and thus we must maintain consistency of the sockbuf during that time.
                   1140:  *
                   1141:  * The caller may receive the data as a single mbuf chain by supplying
                   1142:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1143:  * only for the count in uio_resid.
                   1144:  */
1.3       andrew   1145: int
1.54      lukem    1146: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1147:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1148: {
1.116     yamt     1149:        struct lwp *l = curlwp;
1.160     ad       1150:        struct mbuf     *m, **mp, *mt;
1.211     chs      1151:        size_t len, offset, moff, orig_resid;
                   1152:        int atomic, flags, error, s, type;
1.99      matt     1153:        const struct protosw    *pr;
1.54      lukem    1154:        struct mbuf     *nextrecord;
1.67      he       1155:        int             mbuf_removed = 0;
1.146     dyoung   1156:        const struct domain *dom;
1.196     dsl      1157:        short           wakeup_state = 0;
1.64      thorpej  1158:
1.54      lukem    1159:        pr = so->so_proto;
1.146     dyoung   1160:        atomic = pr->pr_flags & PR_ATOMIC;
                   1161:        dom = pr->pr_domain;
1.1       cgd      1162:        mp = mp0;
1.54      lukem    1163:        type = 0;
                   1164:        orig_resid = uio->uio_resid;
1.102     jonathan 1165:
1.144     dyoung   1166:        if (paddr != NULL)
                   1167:                *paddr = NULL;
                   1168:        if (controlp != NULL)
                   1169:                *controlp = NULL;
                   1170:        if (flagsp != NULL)
1.1       cgd      1171:                flags = *flagsp &~ MSG_EOR;
                   1172:        else
                   1173:                flags = 0;
1.66      enami    1174:
1.1       cgd      1175:        if (flags & MSG_OOB) {
                   1176:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1177:                solock(so);
1.226     rtr      1178:                error = (*pr->pr_usrreqs->pr_recvoob)(so, m, flags & MSG_PEEK);
1.160     ad       1179:                sounlock(so);
1.1       cgd      1180:                if (error)
                   1181:                        goto bad;
                   1182:                do {
1.134     christos 1183:                        error = uiomove(mtod(m, void *),
1.211     chs      1184:                            MIN(uio->uio_resid, m->m_len), uio);
1.1       cgd      1185:                        m = m_free(m);
1.144     dyoung   1186:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1187:  bad:
1.144     dyoung   1188:                if (m != NULL)
1.1       cgd      1189:                        m_freem(m);
1.144     dyoung   1190:                return error;
1.1       cgd      1191:        }
1.144     dyoung   1192:        if (mp != NULL)
1.140     dyoung   1193:                *mp = NULL;
1.160     ad       1194:
                   1195:        /*
                   1196:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1197:         * protocol processing soft interrupts from interrupting us and
                   1198:         * blocking (expensive).
                   1199:         */
                   1200:        s = splsoftnet();
                   1201:        solock(so);
1.54      lukem    1202:  restart:
1.160     ad       1203:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1204:                sounlock(so);
                   1205:                splx(s);
1.144     dyoung   1206:                return error;
1.160     ad       1207:        }
1.1       cgd      1208:
                   1209:        m = so->so_rcv.sb_mb;
                   1210:        /*
                   1211:         * If we have less data than requested, block awaiting more
                   1212:         * (subject to any timeout) if:
1.15      mycroft  1213:         *   1. the current count is less than the low water mark,
1.1       cgd      1214:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1215:         *      receive operation at once if we block (resid <= hiwat), or
                   1216:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1217:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1218:         * we have to do the receive in sections, and thus risk returning
                   1219:         * a short count if a timeout or signal occurs after we start.
                   1220:         */
1.144     dyoung   1221:        if (m == NULL ||
                   1222:            ((flags & MSG_DONTWAIT) == 0 &&
                   1223:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1224:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1225:              ((flags & MSG_WAITALL) &&
                   1226:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1227:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1228: #ifdef DIAGNOSTIC
1.144     dyoung   1229:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1230:                        panic("receive 1");
                   1231: #endif
                   1232:                if (so->so_error) {
1.144     dyoung   1233:                        if (m != NULL)
1.15      mycroft  1234:                                goto dontblock;
1.1       cgd      1235:                        error = so->so_error;
                   1236:                        if ((flags & MSG_PEEK) == 0)
                   1237:                                so->so_error = 0;
                   1238:                        goto release;
                   1239:                }
                   1240:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1241:                        if (m != NULL)
1.15      mycroft  1242:                                goto dontblock;
1.1       cgd      1243:                        else
                   1244:                                goto release;
                   1245:                }
1.144     dyoung   1246:                for (; m != NULL; m = m->m_next)
1.1       cgd      1247:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1248:                                m = so->so_rcv.sb_mb;
                   1249:                                goto dontblock;
                   1250:                        }
                   1251:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1252:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1253:                        error = ENOTCONN;
                   1254:                        goto release;
                   1255:                }
                   1256:                if (uio->uio_resid == 0)
                   1257:                        goto release;
1.206     christos 1258:                if ((so->so_state & SS_NBIO) ||
                   1259:                    (flags & (MSG_DONTWAIT|MSG_NBIO))) {
1.1       cgd      1260:                        error = EWOULDBLOCK;
                   1261:                        goto release;
                   1262:                }
1.69      thorpej  1263:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1264:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1265:                sbunlock(&so->so_rcv);
1.196     dsl      1266:                if (wakeup_state & SS_RESTARTSYS)
                   1267:                        error = ERESTART;
                   1268:                else
                   1269:                        error = sbwait(&so->so_rcv);
1.160     ad       1270:                if (error != 0) {
                   1271:                        sounlock(so);
                   1272:                        splx(s);
1.144     dyoung   1273:                        return error;
1.160     ad       1274:                }
1.196     dsl      1275:                wakeup_state = so->so_state;
1.1       cgd      1276:                goto restart;
                   1277:        }
1.54      lukem    1278:  dontblock:
1.69      thorpej  1279:        /*
                   1280:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1281:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1282:         * pointer to the next record in the socket buffer.  We must keep the
                   1283:         * various socket buffer pointers and local stack versions of the
                   1284:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1285:         * socket lock, and re-reading them when picking it up.
1.159     ad       1286:         *
                   1287:         * Otherwise, we will race with the network stack appending new data
                   1288:         * or records onto the socket buffer by using inconsistent/stale
                   1289:         * versions of the field, possibly resulting in socket buffer
                   1290:         * corruption.
                   1291:         *
                   1292:         * By holding the high-level sblock(), we prevent simultaneous
                   1293:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1294:         */
1.144     dyoung   1295:        if (l != NULL)
1.157     ad       1296:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1297:        KASSERT(m == so->so_rcv.sb_mb);
                   1298:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1299:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1300:        nextrecord = m->m_nextpkt;
                   1301:        if (pr->pr_flags & PR_ADDR) {
                   1302: #ifdef DIAGNOSTIC
                   1303:                if (m->m_type != MT_SONAME)
                   1304:                        panic("receive 1a");
                   1305: #endif
1.3       andrew   1306:                orig_resid = 0;
1.1       cgd      1307:                if (flags & MSG_PEEK) {
                   1308:                        if (paddr)
                   1309:                                *paddr = m_copy(m, 0, m->m_len);
                   1310:                        m = m->m_next;
                   1311:                } else {
                   1312:                        sbfree(&so->so_rcv, m);
1.67      he       1313:                        mbuf_removed = 1;
1.144     dyoung   1314:                        if (paddr != NULL) {
1.1       cgd      1315:                                *paddr = m;
                   1316:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1317:                                m->m_next = NULL;
1.1       cgd      1318:                                m = so->so_rcv.sb_mb;
                   1319:                        } else {
                   1320:                                MFREE(m, so->so_rcv.sb_mb);
                   1321:                                m = so->so_rcv.sb_mb;
                   1322:                        }
1.159     ad       1323:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1324:                }
                   1325:        }
1.159     ad       1326:
                   1327:        /*
                   1328:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1329:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1330:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1331:         * perform externalization (or freeing if controlp == NULL).
                   1332:         */
                   1333:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1334:                struct mbuf *cm = NULL, *cmn;
                   1335:                struct mbuf **cme = &cm;
                   1336:
                   1337:                do {
                   1338:                        if (flags & MSG_PEEK) {
                   1339:                                if (controlp != NULL) {
                   1340:                                        *controlp = m_copy(m, 0, m->m_len);
                   1341:                                        controlp = &(*controlp)->m_next;
                   1342:                                }
                   1343:                                m = m->m_next;
                   1344:                        } else {
                   1345:                                sbfree(&so->so_rcv, m);
1.1       cgd      1346:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1347:                                m->m_next = NULL;
1.159     ad       1348:                                *cme = m;
                   1349:                                cme = &(*cme)->m_next;
1.1       cgd      1350:                                m = so->so_rcv.sb_mb;
1.159     ad       1351:                        }
                   1352:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1353:                if ((flags & MSG_PEEK) == 0)
                   1354:                        sbsync(&so->so_rcv, nextrecord);
                   1355:                for (; cm != NULL; cm = cmn) {
                   1356:                        cmn = cm->m_next;
                   1357:                        cm->m_next = NULL;
                   1358:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1359:                        if (controlp != NULL) {
                   1360:                                if (dom->dom_externalize != NULL &&
                   1361:                                    type == SCM_RIGHTS) {
1.160     ad       1362:                                        sounlock(so);
1.159     ad       1363:                                        splx(s);
1.204     christos 1364:                                        error = (*dom->dom_externalize)(cm, l,
                   1365:                                            (flags & MSG_CMSG_CLOEXEC) ?
                   1366:                                            O_CLOEXEC : 0);
1.159     ad       1367:                                        s = splsoftnet();
1.160     ad       1368:                                        solock(so);
1.159     ad       1369:                                }
                   1370:                                *controlp = cm;
                   1371:                                while (*controlp != NULL)
                   1372:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1373:                        } else {
1.106     itojun   1374:                                /*
                   1375:                                 * Dispose of any SCM_RIGHTS message that went
                   1376:                                 * through the read path rather than recv.
                   1377:                                 */
1.159     ad       1378:                                if (dom->dom_dispose != NULL &&
                   1379:                                    type == SCM_RIGHTS) {
1.160     ad       1380:                                        sounlock(so);
1.159     ad       1381:                                        (*dom->dom_dispose)(cm);
1.160     ad       1382:                                        solock(so);
1.159     ad       1383:                                }
                   1384:                                m_freem(cm);
1.1       cgd      1385:                        }
                   1386:                }
1.159     ad       1387:                if (m != NULL)
                   1388:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1389:                else
                   1390:                        nextrecord = so->so_rcv.sb_mb;
                   1391:                orig_resid = 0;
1.1       cgd      1392:        }
1.69      thorpej  1393:
1.159     ad       1394:        /* If m is non-NULL, we have some data to read. */
                   1395:        if (__predict_true(m != NULL)) {
1.1       cgd      1396:                type = m->m_type;
                   1397:                if (type == MT_OOBDATA)
                   1398:                        flags |= MSG_OOB;
                   1399:        }
1.69      thorpej  1400:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1401:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1402:
1.1       cgd      1403:        moff = 0;
                   1404:        offset = 0;
1.144     dyoung   1405:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1406:                if (m->m_type == MT_OOBDATA) {
                   1407:                        if (type != MT_OOBDATA)
                   1408:                                break;
                   1409:                } else if (type == MT_OOBDATA)
                   1410:                        break;
                   1411: #ifdef DIAGNOSTIC
                   1412:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1413:                        panic("receive 3");
                   1414: #endif
                   1415:                so->so_state &= ~SS_RCVATMARK;
1.196     dsl      1416:                wakeup_state = 0;
1.1       cgd      1417:                len = uio->uio_resid;
                   1418:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1419:                        len = so->so_oobmark - offset;
                   1420:                if (len > m->m_len - moff)
                   1421:                        len = m->m_len - moff;
                   1422:                /*
                   1423:                 * If mp is set, just pass back the mbufs.
                   1424:                 * Otherwise copy them out via the uio, then free.
                   1425:                 * Sockbuf must be consistent here (points to current mbuf,
                   1426:                 * it points to next record) when we drop priority;
                   1427:                 * we must note any additions to the sockbuf when we
                   1428:                 * block interrupts again.
                   1429:                 */
1.144     dyoung   1430:                if (mp == NULL) {
1.69      thorpej  1431:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1432:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1433:                        sounlock(so);
1.1       cgd      1434:                        splx(s);
1.211     chs      1435:                        error = uiomove(mtod(m, char *) + moff, len, uio);
1.20      mycroft  1436:                        s = splsoftnet();
1.160     ad       1437:                        solock(so);
1.144     dyoung   1438:                        if (error != 0) {
1.67      he       1439:                                /*
                   1440:                                 * If any part of the record has been removed
                   1441:                                 * (such as the MT_SONAME mbuf, which will
                   1442:                                 * happen when PR_ADDR, and thus also
                   1443:                                 * PR_ATOMIC, is set), then drop the entire
                   1444:                                 * record to maintain the atomicity of the
                   1445:                                 * receive operation.
                   1446:                                 *
                   1447:                                 * This avoids a later panic("receive 1a")
                   1448:                                 * when compiled with DIAGNOSTIC.
                   1449:                                 */
1.146     dyoung   1450:                                if (m && mbuf_removed && atomic)
1.67      he       1451:                                        (void) sbdroprecord(&so->so_rcv);
                   1452:
1.57      jdolecek 1453:                                goto release;
1.67      he       1454:                        }
1.1       cgd      1455:                } else
                   1456:                        uio->uio_resid -= len;
                   1457:                if (len == m->m_len - moff) {
                   1458:                        if (m->m_flags & M_EOR)
                   1459:                                flags |= MSG_EOR;
                   1460:                        if (flags & MSG_PEEK) {
                   1461:                                m = m->m_next;
                   1462:                                moff = 0;
                   1463:                        } else {
                   1464:                                nextrecord = m->m_nextpkt;
                   1465:                                sbfree(&so->so_rcv, m);
                   1466:                                if (mp) {
                   1467:                                        *mp = m;
                   1468:                                        mp = &m->m_next;
                   1469:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1470:                                        *mp = NULL;
1.1       cgd      1471:                                } else {
                   1472:                                        MFREE(m, so->so_rcv.sb_mb);
                   1473:                                        m = so->so_rcv.sb_mb;
                   1474:                                }
1.69      thorpej  1475:                                /*
                   1476:                                 * If m != NULL, we also know that
                   1477:                                 * so->so_rcv.sb_mb != NULL.
                   1478:                                 */
                   1479:                                KASSERT(so->so_rcv.sb_mb == m);
                   1480:                                if (m) {
1.1       cgd      1481:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1482:                                        if (nextrecord == NULL)
                   1483:                                                so->so_rcv.sb_lastrecord = m;
                   1484:                                } else {
                   1485:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1486:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1487:                                }
                   1488:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1489:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1490:                        }
1.144     dyoung   1491:                } else if (flags & MSG_PEEK)
                   1492:                        moff += len;
                   1493:                else {
1.160     ad       1494:                        if (mp != NULL) {
                   1495:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1496:                                if (__predict_false(mt == NULL)) {
                   1497:                                        sounlock(so);
                   1498:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1499:                                        solock(so);
                   1500:                                }
                   1501:                                *mp = mt;
                   1502:                        }
1.144     dyoung   1503:                        m->m_data += len;
                   1504:                        m->m_len -= len;
                   1505:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1506:                }
                   1507:                if (so->so_oobmark) {
                   1508:                        if ((flags & MSG_PEEK) == 0) {
                   1509:                                so->so_oobmark -= len;
                   1510:                                if (so->so_oobmark == 0) {
                   1511:                                        so->so_state |= SS_RCVATMARK;
                   1512:                                        break;
                   1513:                                }
1.7       cgd      1514:                        } else {
1.1       cgd      1515:                                offset += len;
1.7       cgd      1516:                                if (offset == so->so_oobmark)
                   1517:                                        break;
                   1518:                        }
1.1       cgd      1519:                }
                   1520:                if (flags & MSG_EOR)
                   1521:                        break;
                   1522:                /*
                   1523:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1524:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1525:                 * termination.  If a signal/timeout occurs, return
                   1526:                 * with a short count but without error.
                   1527:                 * Keep sockbuf locked against other readers.
                   1528:                 */
1.144     dyoung   1529:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1530:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1531:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1532:                                break;
1.68      matt     1533:                        /*
                   1534:                         * If we are peeking and the socket receive buffer is
                   1535:                         * full, stop since we can't get more data to peek at.
                   1536:                         */
                   1537:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1538:                                break;
                   1539:                        /*
                   1540:                         * If we've drained the socket buffer, tell the
                   1541:                         * protocol in case it needs to do something to
                   1542:                         * get it filled again.
                   1543:                         */
                   1544:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1.233     rtr      1545:                                (*pr->pr_usrreqs->pr_rcvd)(so, flags, l);
1.69      thorpej  1546:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1547:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.196     dsl      1548:                        if (wakeup_state & SS_RESTARTSYS)
                   1549:                                error = ERESTART;
                   1550:                        else
                   1551:                                error = sbwait(&so->so_rcv);
1.144     dyoung   1552:                        if (error != 0) {
1.1       cgd      1553:                                sbunlock(&so->so_rcv);
1.160     ad       1554:                                sounlock(so);
1.1       cgd      1555:                                splx(s);
1.144     dyoung   1556:                                return 0;
1.1       cgd      1557:                        }
1.21      christos 1558:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1559:                                nextrecord = m->m_nextpkt;
1.196     dsl      1560:                        wakeup_state = so->so_state;
1.1       cgd      1561:                }
                   1562:        }
1.3       andrew   1563:
1.146     dyoung   1564:        if (m && atomic) {
1.3       andrew   1565:                flags |= MSG_TRUNC;
                   1566:                if ((flags & MSG_PEEK) == 0)
                   1567:                        (void) sbdroprecord(&so->so_rcv);
                   1568:        }
1.1       cgd      1569:        if ((flags & MSG_PEEK) == 0) {
1.144     dyoung   1570:                if (m == NULL) {
1.69      thorpej  1571:                        /*
1.70      thorpej  1572:                         * First part is an inline SB_EMPTY_FIXUP().  Second
1.69      thorpej  1573:                         * part makes sure sb_lastrecord is up-to-date if
                   1574:                         * there is still data in the socket buffer.
                   1575:                         */
1.1       cgd      1576:                        so->so_rcv.sb_mb = nextrecord;
1.69      thorpej  1577:                        if (so->so_rcv.sb_mb == NULL) {
                   1578:                                so->so_rcv.sb_mbtail = NULL;
                   1579:                                so->so_rcv.sb_lastrecord = NULL;
                   1580:                        } else if (nextrecord->m_nextpkt == NULL)
                   1581:                                so->so_rcv.sb_lastrecord = nextrecord;
                   1582:                }
                   1583:                SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
                   1584:                SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1       cgd      1585:                if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.233     rtr      1586:                        (*pr->pr_usrreqs->pr_rcvd)(so, flags, l);
1.1       cgd      1587:        }
1.3       andrew   1588:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1589:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1590:                sbunlock(&so->so_rcv);
                   1591:                goto restart;
                   1592:        }
1.108     perry    1593:
1.144     dyoung   1594:        if (flagsp != NULL)
1.1       cgd      1595:                *flagsp |= flags;
1.54      lukem    1596:  release:
1.1       cgd      1597:        sbunlock(&so->so_rcv);
1.160     ad       1598:        sounlock(so);
1.1       cgd      1599:        splx(s);
1.144     dyoung   1600:        return error;
1.1       cgd      1601: }
                   1602:
1.14      mycroft  1603: int
1.54      lukem    1604: soshutdown(struct socket *so, int how)
1.1       cgd      1605: {
1.99      matt     1606:        const struct protosw    *pr;
1.160     ad       1607:        int     error;
                   1608:
                   1609:        KASSERT(solocked(so));
1.34      kleink   1610:
1.54      lukem    1611:        pr = so->so_proto;
1.34      kleink   1612:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1613:                return (EINVAL);
1.1       cgd      1614:
1.160     ad       1615:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1616:                sorflush(so);
1.160     ad       1617:                error = 0;
                   1618:        }
1.34      kleink   1619:        if (how == SHUT_WR || how == SHUT_RDWR)
1.229     rtr      1620:                error = (*pr->pr_usrreqs->pr_shutdown)(so);
1.160     ad       1621:
                   1622:        return error;
1.1       cgd      1623: }
                   1624:
1.195     dsl      1625: void
1.196     dsl      1626: sorestart(struct socket *so)
1.188     ad       1627: {
1.196     dsl      1628:        /*
                   1629:         * An application has called close() on an fd on which another
                   1630:         * of its threads has called a socket system call.
                   1631:         * Mark this and wake everyone up, and code that would block again
                   1632:         * instead returns ERESTART.
                   1633:         * On system call re-entry the fd is validated and EBADF returned.
                   1634:         * Any other fd will block again on the 2nd syscall.
                   1635:         */
1.188     ad       1636:        solock(so);
1.196     dsl      1637:        so->so_state |= SS_RESTARTSYS;
1.188     ad       1638:        cv_broadcast(&so->so_cv);
1.196     dsl      1639:        cv_broadcast(&so->so_snd.sb_cv);
                   1640:        cv_broadcast(&so->so_rcv.sb_cv);
1.188     ad       1641:        sounlock(so);
                   1642: }
                   1643:
1.14      mycroft  1644: void
1.54      lukem    1645: sorflush(struct socket *so)
1.1       cgd      1646: {
1.54      lukem    1647:        struct sockbuf  *sb, asb;
1.99      matt     1648:        const struct protosw    *pr;
1.160     ad       1649:
                   1650:        KASSERT(solocked(so));
1.1       cgd      1651:
1.54      lukem    1652:        sb = &so->so_rcv;
                   1653:        pr = so->so_proto;
1.160     ad       1654:        socantrcvmore(so);
1.1       cgd      1655:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1656:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1657:        sbunlock(sb);
                   1658:        asb = *sb;
1.86      wrstuden 1659:        /*
                   1660:         * Clear most of the sockbuf structure, but leave some of the
                   1661:         * fields valid.
                   1662:         */
                   1663:        memset(&sb->sb_startzero, 0,
                   1664:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1665:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1666:                sounlock(so);
1.1       cgd      1667:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1668:                solock(so);
                   1669:        }
1.98      christos 1670:        sbrelease(&asb, so);
1.1       cgd      1671: }
                   1672:
1.171     plunky   1673: /*
                   1674:  * internal set SOL_SOCKET options
                   1675:  */
1.142     dyoung   1676: static int
1.171     plunky   1677: sosetopt1(struct socket *so, const struct sockopt *sopt)
1.1       cgd      1678: {
1.219     christos 1679:        int error = EINVAL, opt;
                   1680:        int optval = 0; /* XXX: gcc */
1.171     plunky   1681:        struct linger l;
                   1682:        struct timeval tv;
1.142     dyoung   1683:
1.179     christos 1684:        switch ((opt = sopt->sopt_name)) {
1.142     dyoung   1685:
1.170     tls      1686:        case SO_ACCEPTFILTER:
1.177     ad       1687:                error = accept_filt_setopt(so, sopt);
                   1688:                KASSERT(solocked(so));
1.170     tls      1689:                break;
                   1690:
1.171     plunky   1691:        case SO_LINGER:
                   1692:                error = sockopt_get(sopt, &l, sizeof(l));
1.177     ad       1693:                solock(so);
1.171     plunky   1694:                if (error)
1.177     ad       1695:                        break;
1.171     plunky   1696:                if (l.l_linger < 0 || l.l_linger > USHRT_MAX ||
1.177     ad       1697:                    l.l_linger > (INT_MAX / hz)) {
                   1698:                        error = EDOM;
                   1699:                        break;
                   1700:                }
1.171     plunky   1701:                so->so_linger = l.l_linger;
                   1702:                if (l.l_onoff)
                   1703:                        so->so_options |= SO_LINGER;
                   1704:                else
                   1705:                        so->so_options &= ~SO_LINGER;
1.177     ad       1706:                break;
1.1       cgd      1707:
1.142     dyoung   1708:        case SO_DEBUG:
                   1709:        case SO_KEEPALIVE:
                   1710:        case SO_DONTROUTE:
                   1711:        case SO_USELOOPBACK:
                   1712:        case SO_BROADCAST:
                   1713:        case SO_REUSEADDR:
                   1714:        case SO_REUSEPORT:
                   1715:        case SO_OOBINLINE:
                   1716:        case SO_TIMESTAMP:
1.207     christos 1717:        case SO_NOSIGPIPE:
1.184     christos 1718: #ifdef SO_OTIMESTAMP
                   1719:        case SO_OTIMESTAMP:
                   1720: #endif
1.171     plunky   1721:                error = sockopt_getint(sopt, &optval);
1.177     ad       1722:                solock(so);
1.171     plunky   1723:                if (error)
1.177     ad       1724:                        break;
1.171     plunky   1725:                if (optval)
1.179     christos 1726:                        so->so_options |= opt;
1.142     dyoung   1727:                else
1.179     christos 1728:                        so->so_options &= ~opt;
1.142     dyoung   1729:                break;
                   1730:
                   1731:        case SO_SNDBUF:
                   1732:        case SO_RCVBUF:
                   1733:        case SO_SNDLOWAT:
                   1734:        case SO_RCVLOWAT:
1.171     plunky   1735:                error = sockopt_getint(sopt, &optval);
1.177     ad       1736:                solock(so);
1.171     plunky   1737:                if (error)
1.177     ad       1738:                        break;
1.1       cgd      1739:
1.142     dyoung   1740:                /*
                   1741:                 * Values < 1 make no sense for any of these
                   1742:                 * options, so disallow them.
                   1743:                 */
1.177     ad       1744:                if (optval < 1) {
                   1745:                        error = EINVAL;
                   1746:                        break;
                   1747:                }
1.1       cgd      1748:
1.179     christos 1749:                switch (opt) {
1.171     plunky   1750:                case SO_SNDBUF:
1.177     ad       1751:                        if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) {
                   1752:                                error = ENOBUFS;
                   1753:                                break;
                   1754:                        }
1.171     plunky   1755:                        so->so_snd.sb_flags &= ~SB_AUTOSIZE;
                   1756:                        break;
1.1       cgd      1757:
                   1758:                case SO_RCVBUF:
1.177     ad       1759:                        if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) {
                   1760:                                error = ENOBUFS;
                   1761:                                break;
                   1762:                        }
1.171     plunky   1763:                        so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1.142     dyoung   1764:                        break;
                   1765:
                   1766:                /*
                   1767:                 * Make sure the low-water is never greater than
                   1768:                 * the high-water.
                   1769:                 */
1.1       cgd      1770:                case SO_SNDLOWAT:
1.171     plunky   1771:                        if (optval > so->so_snd.sb_hiwat)
                   1772:                                optval = so->so_snd.sb_hiwat;
                   1773:
                   1774:                        so->so_snd.sb_lowat = optval;
1.142     dyoung   1775:                        break;
1.171     plunky   1776:
1.1       cgd      1777:                case SO_RCVLOWAT:
1.171     plunky   1778:                        if (optval > so->so_rcv.sb_hiwat)
                   1779:                                optval = so->so_rcv.sb_hiwat;
                   1780:
                   1781:                        so->so_rcv.sb_lowat = optval;
1.142     dyoung   1782:                        break;
                   1783:                }
                   1784:                break;
1.28      thorpej  1785:
1.179     christos 1786: #ifdef COMPAT_50
                   1787:        case SO_OSNDTIMEO:
                   1788:        case SO_ORCVTIMEO: {
                   1789:                struct timeval50 otv;
                   1790:                error = sockopt_get(sopt, &otv, sizeof(otv));
1.186     pooka    1791:                if (error) {
                   1792:                        solock(so);
1.183     christos 1793:                        break;
1.186     pooka    1794:                }
1.179     christos 1795:                timeval50_to_timeval(&otv, &tv);
                   1796:                opt = opt == SO_OSNDTIMEO ? SO_SNDTIMEO : SO_RCVTIMEO;
1.182     christos 1797:                error = 0;
1.179     christos 1798:                /*FALLTHROUGH*/
                   1799:        }
                   1800: #endif /* COMPAT_50 */
                   1801:
1.142     dyoung   1802:        case SO_SNDTIMEO:
                   1803:        case SO_RCVTIMEO:
1.182     christos 1804:                if (error)
1.179     christos 1805:                        error = sockopt_get(sopt, &tv, sizeof(tv));
1.177     ad       1806:                solock(so);
1.171     plunky   1807:                if (error)
1.177     ad       1808:                        break;
1.171     plunky   1809:
1.177     ad       1810:                if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) {
                   1811:                        error = EDOM;
                   1812:                        break;
                   1813:                }
1.28      thorpej  1814:
1.171     plunky   1815:                optval = tv.tv_sec * hz + tv.tv_usec / tick;
                   1816:                if (optval == 0 && tv.tv_usec != 0)
                   1817:                        optval = 1;
1.28      thorpej  1818:
1.179     christos 1819:                switch (opt) {
1.142     dyoung   1820:                case SO_SNDTIMEO:
1.171     plunky   1821:                        so->so_snd.sb_timeo = optval;
1.1       cgd      1822:                        break;
                   1823:                case SO_RCVTIMEO:
1.171     plunky   1824:                        so->so_rcv.sb_timeo = optval;
1.142     dyoung   1825:                        break;
                   1826:                }
                   1827:                break;
1.1       cgd      1828:
1.142     dyoung   1829:        default:
1.177     ad       1830:                solock(so);
                   1831:                error = ENOPROTOOPT;
                   1832:                break;
1.142     dyoung   1833:        }
1.177     ad       1834:        KASSERT(solocked(so));
                   1835:        return error;
1.142     dyoung   1836: }
1.1       cgd      1837:
1.142     dyoung   1838: int
1.171     plunky   1839: sosetopt(struct socket *so, struct sockopt *sopt)
1.142     dyoung   1840: {
                   1841:        int error, prerr;
1.1       cgd      1842:
1.177     ad       1843:        if (sopt->sopt_level == SOL_SOCKET) {
1.171     plunky   1844:                error = sosetopt1(so, sopt);
1.177     ad       1845:                KASSERT(solocked(so));
                   1846:        } else {
1.142     dyoung   1847:                error = ENOPROTOOPT;
1.177     ad       1848:                solock(so);
                   1849:        }
1.1       cgd      1850:
1.142     dyoung   1851:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1852:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1853:                /* give the protocol stack a shot */
1.171     plunky   1854:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt);
1.142     dyoung   1855:                if (prerr == 0)
                   1856:                        error = 0;
                   1857:                else if (prerr != ENOPROTOOPT)
                   1858:                        error = prerr;
1.171     plunky   1859:        }
1.160     ad       1860:        sounlock(so);
1.142     dyoung   1861:        return error;
1.1       cgd      1862: }
                   1863:
1.171     plunky   1864: /*
                   1865:  * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt()
                   1866:  */
                   1867: int
                   1868: so_setsockopt(struct lwp *l, struct socket *so, int level, int name,
                   1869:     const void *val, size_t valsize)
                   1870: {
                   1871:        struct sockopt sopt;
                   1872:        int error;
                   1873:
                   1874:        KASSERT(valsize == 0 || val != NULL);
                   1875:
                   1876:        sockopt_init(&sopt, level, name, valsize);
                   1877:        sockopt_set(&sopt, val, valsize);
                   1878:
                   1879:        error = sosetopt(so, &sopt);
                   1880:
                   1881:        sockopt_destroy(&sopt);
                   1882:
                   1883:        return error;
                   1884: }
1.235.2.2  skrll    1885:
1.171     plunky   1886: /*
                   1887:  * internal get SOL_SOCKET options
                   1888:  */
                   1889: static int
                   1890: sogetopt1(struct socket *so, struct sockopt *sopt)
                   1891: {
1.179     christos 1892:        int error, optval, opt;
1.171     plunky   1893:        struct linger l;
                   1894:        struct timeval tv;
                   1895:
1.179     christos 1896:        switch ((opt = sopt->sopt_name)) {
1.171     plunky   1897:
                   1898:        case SO_ACCEPTFILTER:
1.177     ad       1899:                error = accept_filt_getopt(so, sopt);
1.171     plunky   1900:                break;
                   1901:
                   1902:        case SO_LINGER:
                   1903:                l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0;
                   1904:                l.l_linger = so->so_linger;
                   1905:
                   1906:                error = sockopt_set(sopt, &l, sizeof(l));
                   1907:                break;
                   1908:
                   1909:        case SO_USELOOPBACK:
                   1910:        case SO_DONTROUTE:
                   1911:        case SO_DEBUG:
                   1912:        case SO_KEEPALIVE:
                   1913:        case SO_REUSEADDR:
                   1914:        case SO_REUSEPORT:
                   1915:        case SO_BROADCAST:
                   1916:        case SO_OOBINLINE:
                   1917:        case SO_TIMESTAMP:
1.207     christos 1918:        case SO_NOSIGPIPE:
1.184     christos 1919: #ifdef SO_OTIMESTAMP
                   1920:        case SO_OTIMESTAMP:
                   1921: #endif
1.218     seanb    1922:        case SO_ACCEPTCONN:
1.179     christos 1923:                error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0);
1.171     plunky   1924:                break;
                   1925:
                   1926:        case SO_TYPE:
                   1927:                error = sockopt_setint(sopt, so->so_type);
                   1928:                break;
                   1929:
                   1930:        case SO_ERROR:
                   1931:                error = sockopt_setint(sopt, so->so_error);
                   1932:                so->so_error = 0;
                   1933:                break;
                   1934:
                   1935:        case SO_SNDBUF:
                   1936:                error = sockopt_setint(sopt, so->so_snd.sb_hiwat);
                   1937:                break;
                   1938:
                   1939:        case SO_RCVBUF:
                   1940:                error = sockopt_setint(sopt, so->so_rcv.sb_hiwat);
                   1941:                break;
                   1942:
                   1943:        case SO_SNDLOWAT:
                   1944:                error = sockopt_setint(sopt, so->so_snd.sb_lowat);
                   1945:                break;
                   1946:
                   1947:        case SO_RCVLOWAT:
                   1948:                error = sockopt_setint(sopt, so->so_rcv.sb_lowat);
                   1949:                break;
                   1950:
1.179     christos 1951: #ifdef COMPAT_50
                   1952:        case SO_OSNDTIMEO:
                   1953:        case SO_ORCVTIMEO: {
                   1954:                struct timeval50 otv;
                   1955:
                   1956:                optval = (opt == SO_OSNDTIMEO ?
                   1957:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1958:
                   1959:                otv.tv_sec = optval / hz;
                   1960:                otv.tv_usec = (optval % hz) * tick;
                   1961:
                   1962:                error = sockopt_set(sopt, &otv, sizeof(otv));
                   1963:                break;
                   1964:        }
                   1965: #endif /* COMPAT_50 */
                   1966:
1.171     plunky   1967:        case SO_SNDTIMEO:
                   1968:        case SO_RCVTIMEO:
1.179     christos 1969:                optval = (opt == SO_SNDTIMEO ?
1.171     plunky   1970:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1971:
                   1972:                tv.tv_sec = optval / hz;
                   1973:                tv.tv_usec = (optval % hz) * tick;
                   1974:
                   1975:                error = sockopt_set(sopt, &tv, sizeof(tv));
                   1976:                break;
                   1977:
                   1978:        case SO_OVERFLOWED:
                   1979:                error = sockopt_setint(sopt, so->so_rcv.sb_overflowed);
                   1980:                break;
                   1981:
                   1982:        default:
                   1983:                error = ENOPROTOOPT;
                   1984:                break;
                   1985:        }
                   1986:
                   1987:        return (error);
                   1988: }
                   1989:
1.14      mycroft  1990: int
1.171     plunky   1991: sogetopt(struct socket *so, struct sockopt *sopt)
1.1       cgd      1992: {
1.160     ad       1993:        int             error;
1.1       cgd      1994:
1.160     ad       1995:        solock(so);
1.171     plunky   1996:        if (sopt->sopt_level != SOL_SOCKET) {
1.1       cgd      1997:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       1998:                        error = ((*so->so_proto->pr_ctloutput)
1.171     plunky   1999:                            (PRCO_GETOPT, so, sopt));
1.1       cgd      2000:                } else
1.160     ad       2001:                        error = (ENOPROTOOPT);
1.1       cgd      2002:        } else {
1.171     plunky   2003:                error = sogetopt1(so, sopt);
                   2004:        }
                   2005:        sounlock(so);
                   2006:        return (error);
                   2007: }
                   2008:
                   2009: /*
                   2010:  * alloc sockopt data buffer buffer
                   2011:  *     - will be released at destroy
                   2012:  */
1.176     plunky   2013: static int
                   2014: sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag)
1.171     plunky   2015: {
                   2016:
                   2017:        KASSERT(sopt->sopt_size == 0);
                   2018:
1.176     plunky   2019:        if (len > sizeof(sopt->sopt_buf)) {
                   2020:                sopt->sopt_data = kmem_zalloc(len, kmflag);
                   2021:                if (sopt->sopt_data == NULL)
                   2022:                        return ENOMEM;
                   2023:        } else
1.171     plunky   2024:                sopt->sopt_data = sopt->sopt_buf;
                   2025:
                   2026:        sopt->sopt_size = len;
1.176     plunky   2027:        return 0;
1.171     plunky   2028: }
                   2029:
                   2030: /*
                   2031:  * initialise sockopt storage
1.176     plunky   2032:  *     - MAY sleep during allocation
1.171     plunky   2033:  */
                   2034: void
                   2035: sockopt_init(struct sockopt *sopt, int level, int name, size_t size)
                   2036: {
1.1       cgd      2037:
1.171     plunky   2038:        memset(sopt, 0, sizeof(*sopt));
1.1       cgd      2039:
1.171     plunky   2040:        sopt->sopt_level = level;
                   2041:        sopt->sopt_name = name;
1.176     plunky   2042:        (void)sockopt_alloc(sopt, size, KM_SLEEP);
1.171     plunky   2043: }
                   2044:
                   2045: /*
                   2046:  * destroy sockopt storage
                   2047:  *     - will release any held memory references
                   2048:  */
                   2049: void
                   2050: sockopt_destroy(struct sockopt *sopt)
                   2051: {
                   2052:
                   2053:        if (sopt->sopt_data != sopt->sopt_buf)
1.173     plunky   2054:                kmem_free(sopt->sopt_data, sopt->sopt_size);
1.171     plunky   2055:
                   2056:        memset(sopt, 0, sizeof(*sopt));
                   2057: }
                   2058:
                   2059: /*
                   2060:  * set sockopt value
                   2061:  *     - value is copied into sockopt
1.176     plunky   2062:  *     - memory is allocated when necessary, will not sleep
1.171     plunky   2063:  */
                   2064: int
                   2065: sockopt_set(struct sockopt *sopt, const void *buf, size_t len)
                   2066: {
1.176     plunky   2067:        int error;
1.171     plunky   2068:
1.176     plunky   2069:        if (sopt->sopt_size == 0) {
                   2070:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2071:                if (error)
                   2072:                        return error;
                   2073:        }
1.171     plunky   2074:
                   2075:        KASSERT(sopt->sopt_size == len);
                   2076:        memcpy(sopt->sopt_data, buf, len);
                   2077:        return 0;
                   2078: }
                   2079:
                   2080: /*
                   2081:  * common case of set sockopt integer value
                   2082:  */
                   2083: int
                   2084: sockopt_setint(struct sockopt *sopt, int val)
                   2085: {
                   2086:
                   2087:        return sockopt_set(sopt, &val, sizeof(int));
                   2088: }
                   2089:
                   2090: /*
                   2091:  * get sockopt value
                   2092:  *     - correct size must be given
                   2093:  */
                   2094: int
                   2095: sockopt_get(const struct sockopt *sopt, void *buf, size_t len)
                   2096: {
1.170     tls      2097:
1.171     plunky   2098:        if (sopt->sopt_size != len)
                   2099:                return EINVAL;
1.1       cgd      2100:
1.171     plunky   2101:        memcpy(buf, sopt->sopt_data, len);
                   2102:        return 0;
                   2103: }
1.1       cgd      2104:
1.171     plunky   2105: /*
                   2106:  * common case of get sockopt integer value
                   2107:  */
                   2108: int
                   2109: sockopt_getint(const struct sockopt *sopt, int *valp)
                   2110: {
1.1       cgd      2111:
1.171     plunky   2112:        return sockopt_get(sopt, valp, sizeof(int));
                   2113: }
1.1       cgd      2114:
1.171     plunky   2115: /*
                   2116:  * set sockopt value from mbuf
                   2117:  *     - ONLY for legacy code
                   2118:  *     - mbuf is released by sockopt
1.176     plunky   2119:  *     - will not sleep
1.171     plunky   2120:  */
                   2121: int
                   2122: sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m)
                   2123: {
                   2124:        size_t len;
1.176     plunky   2125:        int error;
1.1       cgd      2126:
1.171     plunky   2127:        len = m_length(m);
1.1       cgd      2128:
1.176     plunky   2129:        if (sopt->sopt_size == 0) {
                   2130:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2131:                if (error)
                   2132:                        return error;
                   2133:        }
1.1       cgd      2134:
1.171     plunky   2135:        KASSERT(sopt->sopt_size == len);
                   2136:        m_copydata(m, 0, len, sopt->sopt_data);
                   2137:        m_freem(m);
1.1       cgd      2138:
1.171     plunky   2139:        return 0;
                   2140: }
1.1       cgd      2141:
1.171     plunky   2142: /*
                   2143:  * get sockopt value into mbuf
                   2144:  *     - ONLY for legacy code
                   2145:  *     - mbuf to be released by the caller
1.176     plunky   2146:  *     - will not sleep
1.171     plunky   2147:  */
                   2148: struct mbuf *
                   2149: sockopt_getmbuf(const struct sockopt *sopt)
                   2150: {
                   2151:        struct mbuf *m;
1.107     darrenr  2152:
1.176     plunky   2153:        if (sopt->sopt_size > MCLBYTES)
                   2154:                return NULL;
                   2155:
                   2156:        m = m_get(M_DONTWAIT, MT_SOOPTS);
1.171     plunky   2157:        if (m == NULL)
                   2158:                return NULL;
                   2159:
1.176     plunky   2160:        if (sopt->sopt_size > MLEN) {
                   2161:                MCLGET(m, M_DONTWAIT);
                   2162:                if ((m->m_flags & M_EXT) == 0) {
                   2163:                        m_free(m);
                   2164:                        return NULL;
                   2165:                }
1.1       cgd      2166:        }
1.176     plunky   2167:
                   2168:        memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size);
                   2169:        m->m_len = sopt->sopt_size;
1.160     ad       2170:
1.171     plunky   2171:        return m;
1.1       cgd      2172: }
                   2173:
1.14      mycroft  2174: void
1.54      lukem    2175: sohasoutofband(struct socket *so)
1.1       cgd      2176: {
1.153     rmind    2177:
1.90      christos 2178:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.189     ad       2179:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT);
1.1       cgd      2180: }
1.72      jdolecek 2181:
                   2182: static void
                   2183: filt_sordetach(struct knote *kn)
                   2184: {
                   2185:        struct socket   *so;
                   2186:
1.235     matt     2187:        so = ((file_t *)kn->kn_obj)->f_socket;
1.160     ad       2188:        solock(so);
1.73      christos 2189:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2190:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 2191:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       2192:        sounlock(so);
1.72      jdolecek 2193: }
                   2194:
                   2195: /*ARGSUSED*/
                   2196: static int
1.129     yamt     2197: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 2198: {
                   2199:        struct socket   *so;
1.160     ad       2200:        int rv;
1.72      jdolecek 2201:
1.235     matt     2202:        so = ((file_t *)kn->kn_obj)->f_socket;
1.160     ad       2203:        if (hint != NOTE_SUBMIT)
                   2204:                solock(so);
1.72      jdolecek 2205:        kn->kn_data = so->so_rcv.sb_cc;
                   2206:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    2207:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2208:                kn->kn_fflags = so->so_error;
1.160     ad       2209:                rv = 1;
                   2210:        } else if (so->so_error)        /* temporary udp error */
                   2211:                rv = 1;
                   2212:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2213:                rv = (kn->kn_data >= kn->kn_sdata);
1.235.2.2  skrll    2214:        else
1.160     ad       2215:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   2216:        if (hint != NOTE_SUBMIT)
                   2217:                sounlock(so);
                   2218:        return rv;
1.72      jdolecek 2219: }
                   2220:
                   2221: static void
                   2222: filt_sowdetach(struct knote *kn)
                   2223: {
                   2224:        struct socket   *so;
                   2225:
1.235     matt     2226:        so = ((file_t *)kn->kn_obj)->f_socket;
1.160     ad       2227:        solock(so);
1.73      christos 2228:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2229:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 2230:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       2231:        sounlock(so);
1.72      jdolecek 2232: }
                   2233:
                   2234: /*ARGSUSED*/
                   2235: static int
1.129     yamt     2236: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 2237: {
                   2238:        struct socket   *so;
1.160     ad       2239:        int rv;
1.72      jdolecek 2240:
1.235     matt     2241:        so = ((file_t *)kn->kn_obj)->f_socket;
1.160     ad       2242:        if (hint != NOTE_SUBMIT)
                   2243:                solock(so);
1.72      jdolecek 2244:        kn->kn_data = sbspace(&so->so_snd);
                   2245:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    2246:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2247:                kn->kn_fflags = so->so_error;
1.160     ad       2248:                rv = 1;
                   2249:        } else if (so->so_error)        /* temporary udp error */
                   2250:                rv = 1;
                   2251:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 2252:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       2253:                rv = 0;
                   2254:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2255:                rv = (kn->kn_data >= kn->kn_sdata);
                   2256:        else
                   2257:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   2258:        if (hint != NOTE_SUBMIT)
                   2259:                sounlock(so);
                   2260:        return rv;
1.72      jdolecek 2261: }
                   2262:
                   2263: /*ARGSUSED*/
                   2264: static int
1.129     yamt     2265: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 2266: {
                   2267:        struct socket   *so;
1.160     ad       2268:        int rv;
1.72      jdolecek 2269:
1.235     matt     2270:        so = ((file_t *)kn->kn_obj)->f_socket;
1.72      jdolecek 2271:
                   2272:        /*
                   2273:         * Set kn_data to number of incoming connections, not
                   2274:         * counting partial (incomplete) connections.
1.108     perry    2275:         */
1.160     ad       2276:        if (hint != NOTE_SUBMIT)
                   2277:                solock(so);
1.72      jdolecek 2278:        kn->kn_data = so->so_qlen;
1.160     ad       2279:        rv = (kn->kn_data > 0);
                   2280:        if (hint != NOTE_SUBMIT)
                   2281:                sounlock(so);
                   2282:        return rv;
1.72      jdolecek 2283: }
                   2284:
                   2285: static const struct filterops solisten_filtops =
                   2286:        { 1, NULL, filt_sordetach, filt_solisten };
                   2287: static const struct filterops soread_filtops =
                   2288:        { 1, NULL, filt_sordetach, filt_soread };
                   2289: static const struct filterops sowrite_filtops =
                   2290:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   2291:
                   2292: int
1.129     yamt     2293: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 2294: {
                   2295:        struct socket   *so;
                   2296:        struct sockbuf  *sb;
                   2297:
1.235     matt     2298:        so = ((file_t *)kn->kn_obj)->f_socket;
1.160     ad       2299:        solock(so);
1.72      jdolecek 2300:        switch (kn->kn_filter) {
                   2301:        case EVFILT_READ:
                   2302:                if (so->so_options & SO_ACCEPTCONN)
                   2303:                        kn->kn_fop = &solisten_filtops;
                   2304:                else
                   2305:                        kn->kn_fop = &soread_filtops;
                   2306:                sb = &so->so_rcv;
                   2307:                break;
                   2308:        case EVFILT_WRITE:
                   2309:                kn->kn_fop = &sowrite_filtops;
                   2310:                sb = &so->so_snd;
                   2311:                break;
                   2312:        default:
1.160     ad       2313:                sounlock(so);
1.149     pooka    2314:                return (EINVAL);
1.72      jdolecek 2315:        }
1.73      christos 2316:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 2317:        sb->sb_flags |= SB_KNOTE;
1.160     ad       2318:        sounlock(so);
1.72      jdolecek 2319:        return (0);
                   2320: }
                   2321:
1.154     ad       2322: static int
                   2323: sodopoll(struct socket *so, int events)
                   2324: {
                   2325:        int revents;
                   2326:
                   2327:        revents = 0;
                   2328:
                   2329:        if (events & (POLLIN | POLLRDNORM))
                   2330:                if (soreadable(so))
                   2331:                        revents |= events & (POLLIN | POLLRDNORM);
                   2332:
                   2333:        if (events & (POLLOUT | POLLWRNORM))
                   2334:                if (sowritable(so))
                   2335:                        revents |= events & (POLLOUT | POLLWRNORM);
                   2336:
                   2337:        if (events & (POLLPRI | POLLRDBAND))
                   2338:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   2339:                        revents |= events & (POLLPRI | POLLRDBAND);
                   2340:
                   2341:        return revents;
                   2342: }
                   2343:
                   2344: int
                   2345: sopoll(struct socket *so, int events)
                   2346: {
                   2347:        int revents = 0;
                   2348:
1.160     ad       2349: #ifndef DIAGNOSTIC
                   2350:        /*
                   2351:         * Do a quick, unlocked check in expectation that the socket
                   2352:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   2353:         * as the solocked() assertions will fail.
                   2354:         */
1.154     ad       2355:        if ((revents = sodopoll(so, events)) != 0)
                   2356:                return revents;
1.160     ad       2357: #endif
1.154     ad       2358:
1.160     ad       2359:        solock(so);
1.154     ad       2360:        if ((revents = sodopoll(so, events)) == 0) {
                   2361:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   2362:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       2363:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       2364:                }
                   2365:
                   2366:                if (events & (POLLOUT | POLLWRNORM)) {
                   2367:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2368:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2369:                }
                   2370:        }
1.160     ad       2371:        sounlock(so);
1.154     ad       2372:
                   2373:        return revents;
                   2374: }
                   2375:
                   2376:
1.94      yamt     2377: #include <sys/sysctl.h>
                   2378:
                   2379: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
1.212     pooka    2380: static int sysctl_kern_sbmax(SYSCTLFN_PROTO);
1.94      yamt     2381:
                   2382: /*
                   2383:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2384:  * value is not too small.
                   2385:  * (XXX should we maybe make sure it's not too large as well?)
                   2386:  */
                   2387: static int
                   2388: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2389: {
                   2390:        int error, new_somaxkva;
                   2391:        struct sysctlnode node;
                   2392:
                   2393:        new_somaxkva = somaxkva;
                   2394:        node = *rnode;
                   2395:        node.sysctl_data = &new_somaxkva;
                   2396:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2397:        if (error || newp == NULL)
                   2398:                return (error);
                   2399:
                   2400:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2401:                return (EINVAL);
                   2402:
1.136     ad       2403:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2404:        somaxkva = new_somaxkva;
1.136     ad       2405:        cv_broadcast(&socurkva_cv);
                   2406:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2407:
                   2408:        return (error);
                   2409: }
                   2410:
1.212     pooka    2411: /*
                   2412:  * sysctl helper routine for kern.sbmax. Basically just ensures that
                   2413:  * any new value is not too small.
                   2414:  */
                   2415: static int
                   2416: sysctl_kern_sbmax(SYSCTLFN_ARGS)
                   2417: {
                   2418:        int error, new_sbmax;
                   2419:        struct sysctlnode node;
                   2420:
                   2421:        new_sbmax = sb_max;
                   2422:        node = *rnode;
                   2423:        node.sysctl_data = &new_sbmax;
                   2424:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2425:        if (error || newp == NULL)
                   2426:                return (error);
                   2427:
                   2428:        KERNEL_LOCK(1, NULL);
                   2429:        error = sb_max_set(new_sbmax);
                   2430:        KERNEL_UNLOCK_ONE(NULL);
                   2431:
                   2432:        return (error);
                   2433: }
                   2434:
1.178     pooka    2435: static void
1.212     pooka    2436: sysctl_kern_socket_setup(void)
1.94      yamt     2437: {
                   2438:
1.178     pooka    2439:        KASSERT(socket_sysctllog == NULL);
1.97      atatat   2440:
1.178     pooka    2441:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2442:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2443:                       CTLTYPE_INT, "somaxkva",
                   2444:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2445:                                    "used for socket buffers"),
1.94      yamt     2446:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2447:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
1.212     pooka    2448:
                   2449:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
                   2450:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
                   2451:                       CTLTYPE_INT, "sbmax",
                   2452:                       SYSCTL_DESCR("Maximum socket buffer size"),
                   2453:                       sysctl_kern_sbmax, 0, NULL, 0,
                   2454:                       CTL_KERN, KERN_SBMAX, CTL_EOL);
1.94      yamt     2455: }

CVSweb <webmaster@jp.NetBSD.org>