[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 1.161

1.161   ! ad          1: /*     $NetBSD: uipc_socket.c,v 1.160 2008/04/24 11:38:36 ad Exp $     */
1.64      thorpej     2:
                      3: /*-
1.154     ad          4:  * Copyright (c) 2002, 2007, 2008 The NetBSD Foundation, Inc.
1.64      thorpej     5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
                      8:  * by Jason R. Thorpe of Wasabi Systems, Inc.
                      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:  * 3. All advertising materials mentioning features or use of this software
                     19:  *    must display the following acknowledgement:
                     20:  *     This product includes software developed by the NetBSD
                     21:  *     Foundation, Inc. and its contributors.
                     22:  * 4. Neither the name of The NetBSD Foundation nor the names of its
                     23:  *    contributors may be used to endorse or promote products derived
                     24:  *    from this software without specific prior written permission.
                     25:  *
                     26:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     27:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     28:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     29:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     30:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     31:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     32:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     33:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     34:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     35:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     36:  * POSSIBILITY OF SUCH DAMAGE.
                     37:  */
1.16      cgd        38:
1.1       cgd        39: /*
1.159     ad         40:  * Copyright (c) 2004 The FreeBSD Foundation
                     41:  * Copyright (c) 2004 Robert Watson
1.15      mycroft    42:  * Copyright (c) 1982, 1986, 1988, 1990, 1993
                     43:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        44:  *
                     45:  * Redistribution and use in source and binary forms, with or without
                     46:  * modification, are permitted provided that the following conditions
                     47:  * are met:
                     48:  * 1. Redistributions of source code must retain the above copyright
                     49:  *    notice, this list of conditions and the following disclaimer.
                     50:  * 2. Redistributions in binary form must reproduce the above copyright
                     51:  *    notice, this list of conditions and the following disclaimer in the
                     52:  *    documentation and/or other materials provided with the distribution.
1.85      agc        53:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        54:  *    may be used to endorse or promote products derived from this software
                     55:  *    without specific prior written permission.
                     56:  *
                     57:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     58:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     59:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     60:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     61:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     62:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     63:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     64:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     65:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     66:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     67:  * SUCH DAMAGE.
                     68:  *
1.32      fvdl       69:  *     @(#)uipc_socket.c       8.6 (Berkeley) 5/2/95
1.1       cgd        70:  */
1.59      lukem      71:
                     72: #include <sys/cdefs.h>
1.161   ! ad         73: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.160 2008/04/24 11:38:36 ad Exp $");
1.64      thorpej    74:
                     75: #include "opt_sock_counters.h"
                     76: #include "opt_sosend_loan.h"
1.81      martin     77: #include "opt_mbuftrace.h"
1.84      ragge      78: #include "opt_somaxkva.h"
1.1       cgd        79:
1.9       mycroft    80: #include <sys/param.h>
                     81: #include <sys/systm.h>
                     82: #include <sys/proc.h>
                     83: #include <sys/file.h>
1.142     dyoung     84: #include <sys/filedesc.h>
1.9       mycroft    85: #include <sys/malloc.h>
                     86: #include <sys/mbuf.h>
                     87: #include <sys/domain.h>
                     88: #include <sys/kernel.h>
                     89: #include <sys/protosw.h>
                     90: #include <sys/socket.h>
                     91: #include <sys/socketvar.h>
1.21      christos   92: #include <sys/signalvar.h>
1.9       mycroft    93: #include <sys/resourcevar.h>
1.72      jdolecek   94: #include <sys/event.h>
1.89      christos   95: #include <sys/poll.h>
1.118     elad       96: #include <sys/kauth.h>
1.136     ad         97: #include <sys/mutex.h>
                     98: #include <sys/condvar.h>
1.37      thorpej    99:
1.64      thorpej   100: #include <uvm/uvm.h>
                    101:
1.77      thorpej   102: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
                    103: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37      thorpej   104:
1.142     dyoung    105: extern const struct fileops socketops;
                    106:
1.54      lukem     107: extern int     somaxconn;                      /* patchable (XXX sysctl) */
                    108: int            somaxconn = SOMAXCONN;
1.160     ad        109: kmutex_t       *softnet_lock;
1.49      jonathan  110:
1.64      thorpej   111: #ifdef SOSEND_COUNTERS
                    112: #include <sys/device.h>
                    113:
1.113     thorpej   114: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   115:     NULL, "sosend", "loan big");
1.113     thorpej   116: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   117:     NULL, "sosend", "copy big");
1.113     thorpej   118: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   119:     NULL, "sosend", "copy small");
1.113     thorpej   120: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   121:     NULL, "sosend", "kva limit");
                    122:
                    123: #define        SOSEND_COUNTER_INCR(ev)         (ev)->ev_count++
                    124:
1.101     matt      125: EVCNT_ATTACH_STATIC(sosend_loan_big);
                    126: EVCNT_ATTACH_STATIC(sosend_copy_big);
                    127: EVCNT_ATTACH_STATIC(sosend_copy_small);
                    128: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64      thorpej   129: #else
                    130:
                    131: #define        SOSEND_COUNTER_INCR(ev)         /* nothing */
                    132:
                    133: #endif /* SOSEND_COUNTERS */
                    134:
1.119     yamt      135: static struct callback_entry sokva_reclaimerentry;
1.1       cgd       136:
1.71      thorpej   137: #ifdef SOSEND_NO_LOAN
1.121     yamt      138: int sock_loan_thresh = -1;
1.71      thorpej   139: #else
1.121     yamt      140: int sock_loan_thresh = 4096;
1.65      thorpej   141: #endif
1.64      thorpej   142:
1.136     ad        143: static kmutex_t so_pendfree_lock;
1.113     thorpej   144: static struct mbuf *so_pendfree;
1.64      thorpej   145:
1.84      ragge     146: #ifndef SOMAXKVA
                    147: #define        SOMAXKVA (16 * 1024 * 1024)
                    148: #endif
                    149: int somaxkva = SOMAXKVA;
1.113     thorpej   150: static int socurkva;
1.136     ad        151: static kcondvar_t socurkva_cv;
1.64      thorpej   152:
                    153: #define        SOCK_LOAN_CHUNK         65536
                    154:
1.117     yamt      155: static size_t sodopendfree(void);
                    156: static size_t sodopendfreel(void);
1.93      yamt      157:
1.113     thorpej   158: static vsize_t
1.129     yamt      159: sokvareserve(struct socket *so, vsize_t len)
1.80      yamt      160: {
1.98      christos  161:        int error;
1.80      yamt      162:
1.136     ad        163:        mutex_enter(&so_pendfree_lock);
1.80      yamt      164:        while (socurkva + len > somaxkva) {
1.93      yamt      165:                size_t freed;
                    166:
                    167:                /*
                    168:                 * try to do pendfree.
                    169:                 */
                    170:
1.117     yamt      171:                freed = sodopendfreel();
1.93      yamt      172:
                    173:                /*
                    174:                 * if some kva was freed, try again.
                    175:                 */
                    176:
                    177:                if (freed)
1.80      yamt      178:                        continue;
1.93      yamt      179:
1.80      yamt      180:                SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136     ad        181:                error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98      christos  182:                if (error) {
                    183:                        len = 0;
                    184:                        break;
                    185:                }
1.80      yamt      186:        }
1.93      yamt      187:        socurkva += len;
1.136     ad        188:        mutex_exit(&so_pendfree_lock);
1.98      christos  189:        return len;
1.95      yamt      190: }
                    191:
1.113     thorpej   192: static void
1.95      yamt      193: sokvaunreserve(vsize_t len)
                    194: {
                    195:
1.136     ad        196:        mutex_enter(&so_pendfree_lock);
1.95      yamt      197:        socurkva -= len;
1.136     ad        198:        cv_broadcast(&socurkva_cv);
                    199:        mutex_exit(&so_pendfree_lock);
1.95      yamt      200: }
                    201:
                    202: /*
                    203:  * sokvaalloc: allocate kva for loan.
                    204:  */
                    205:
                    206: vaddr_t
                    207: sokvaalloc(vsize_t len, struct socket *so)
                    208: {
                    209:        vaddr_t lva;
                    210:
                    211:        /*
                    212:         * reserve kva.
                    213:         */
                    214:
1.98      christos  215:        if (sokvareserve(so, len) == 0)
                    216:                return 0;
1.93      yamt      217:
                    218:        /*
                    219:         * allocate kva.
                    220:         */
1.80      yamt      221:
1.109     yamt      222:        lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95      yamt      223:        if (lva == 0) {
                    224:                sokvaunreserve(len);
1.80      yamt      225:                return (0);
1.95      yamt      226:        }
1.80      yamt      227:
                    228:        return lva;
                    229: }
                    230:
1.93      yamt      231: /*
                    232:  * sokvafree: free kva for loan.
                    233:  */
                    234:
1.80      yamt      235: void
                    236: sokvafree(vaddr_t sva, vsize_t len)
                    237: {
1.93      yamt      238:
                    239:        /*
                    240:         * free kva.
                    241:         */
1.80      yamt      242:
1.109     yamt      243:        uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93      yamt      244:
                    245:        /*
                    246:         * unreserve kva.
                    247:         */
                    248:
1.95      yamt      249:        sokvaunreserve(len);
1.80      yamt      250: }
                    251:
1.64      thorpej   252: static void
1.134     christos  253: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64      thorpej   254: {
1.156     yamt      255:        vaddr_t sva, eva;
1.64      thorpej   256:        vsize_t len;
1.156     yamt      257:        int npgs;
                    258:
                    259:        KASSERT(pgs != NULL);
1.64      thorpej   260:
                    261:        eva = round_page((vaddr_t) buf + size);
                    262:        sva = trunc_page((vaddr_t) buf);
                    263:        len = eva - sva;
                    264:        npgs = len >> PAGE_SHIFT;
                    265:
                    266:        pmap_kremove(sva, len);
                    267:        pmap_update(pmap_kernel());
                    268:        uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80      yamt      269:        sokvafree(sva, len);
1.64      thorpej   270: }
                    271:
                    272: static size_t
1.152     matt      273: sodopendfree(void)
1.64      thorpej   274: {
1.93      yamt      275:        size_t rv;
1.64      thorpej   276:
1.160     ad        277:        if (__predict_true(so_pendfree == NULL))
                    278:                return 0;
                    279:
1.136     ad        280:        mutex_enter(&so_pendfree_lock);
1.117     yamt      281:        rv = sodopendfreel();
1.136     ad        282:        mutex_exit(&so_pendfree_lock);
1.93      yamt      283:
                    284:        return rv;
                    285: }
                    286:
                    287: /*
                    288:  * sodopendfreel: free mbufs on "pendfree" list.
1.136     ad        289:  * unlock and relock so_pendfree_lock when freeing mbufs.
1.93      yamt      290:  *
1.136     ad        291:  * => called with so_pendfree_lock held.
1.93      yamt      292:  */
                    293:
                    294: static size_t
1.152     matt      295: sodopendfreel(void)
1.93      yamt      296: {
1.137     ad        297:        struct mbuf *m, *next;
1.93      yamt      298:        size_t rv = 0;
                    299:
1.136     ad        300:        KASSERT(mutex_owned(&so_pendfree_lock));
1.64      thorpej   301:
1.137     ad        302:        while (so_pendfree != NULL) {
1.64      thorpej   303:                m = so_pendfree;
1.93      yamt      304:                so_pendfree = NULL;
1.136     ad        305:                mutex_exit(&so_pendfree_lock);
1.93      yamt      306:
                    307:                for (; m != NULL; m = next) {
                    308:                        next = m->m_next;
1.156     yamt      309:                        KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
                    310:                        KASSERT(m->m_ext.ext_refcnt == 0);
1.93      yamt      311:
                    312:                        rv += m->m_ext.ext_size;
1.156     yamt      313:                        sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
1.93      yamt      314:                            m->m_ext.ext_size);
1.145     ad        315:                        pool_cache_put(mb_cache, m);
1.93      yamt      316:                }
1.64      thorpej   317:
1.136     ad        318:                mutex_enter(&so_pendfree_lock);
1.64      thorpej   319:        }
                    320:
                    321:        return (rv);
                    322: }
                    323:
1.80      yamt      324: void
1.134     christos  325: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64      thorpej   326: {
                    327:
1.156     yamt      328:        KASSERT(m != NULL);
1.64      thorpej   329:
1.93      yamt      330:        /*
                    331:         * postpone freeing mbuf.
                    332:         *
                    333:         * we can't do it in interrupt context
                    334:         * because we need to put kva back to kernel_map.
                    335:         */
                    336:
1.136     ad        337:        mutex_enter(&so_pendfree_lock);
1.92      yamt      338:        m->m_next = so_pendfree;
                    339:        so_pendfree = m;
1.136     ad        340:        cv_broadcast(&socurkva_cv);
                    341:        mutex_exit(&so_pendfree_lock);
1.64      thorpej   342: }
                    343:
                    344: static long
                    345: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
                    346: {
                    347:        struct iovec *iov = uio->uio_iov;
                    348:        vaddr_t sva, eva;
                    349:        vsize_t len;
1.156     yamt      350:        vaddr_t lva;
                    351:        int npgs, error;
                    352:        vaddr_t va;
                    353:        int i;
1.64      thorpej   354:
1.116     yamt      355:        if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64      thorpej   356:                return (0);
                    357:
                    358:        if (iov->iov_len < (size_t) space)
                    359:                space = iov->iov_len;
                    360:        if (space > SOCK_LOAN_CHUNK)
                    361:                space = SOCK_LOAN_CHUNK;
                    362:
                    363:        eva = round_page((vaddr_t) iov->iov_base + space);
                    364:        sva = trunc_page((vaddr_t) iov->iov_base);
                    365:        len = eva - sva;
                    366:        npgs = len >> PAGE_SHIFT;
                    367:
1.79      thorpej   368:        KASSERT(npgs <= M_EXT_MAXPAGES);
                    369:
1.80      yamt      370:        lva = sokvaalloc(len, so);
1.64      thorpej   371:        if (lva == 0)
1.80      yamt      372:                return 0;
1.64      thorpej   373:
1.116     yamt      374:        error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79      thorpej   375:            m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64      thorpej   376:        if (error) {
1.80      yamt      377:                sokvafree(lva, len);
1.64      thorpej   378:                return (0);
                    379:        }
                    380:
                    381:        for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79      thorpej   382:                pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
                    383:                    VM_PROT_READ);
1.64      thorpej   384:        pmap_update(pmap_kernel());
                    385:
                    386:        lva += (vaddr_t) iov->iov_base & PAGE_MASK;
                    387:
1.134     christos  388:        MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79      thorpej   389:        m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64      thorpej   390:
                    391:        uio->uio_resid -= space;
                    392:        /* uio_offset not updated, not set/used for write(2) */
1.134     christos  393:        uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64      thorpej   394:        uio->uio_iov->iov_len -= space;
                    395:        if (uio->uio_iov->iov_len == 0) {
                    396:                uio->uio_iov++;
                    397:                uio->uio_iovcnt--;
                    398:        }
                    399:
                    400:        return (space);
                    401: }
                    402:
1.119     yamt      403: static int
1.129     yamt      404: sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg)
1.119     yamt      405: {
                    406:
                    407:        KASSERT(ce == &sokva_reclaimerentry);
                    408:        KASSERT(obj == NULL);
                    409:
                    410:        sodopendfree();
                    411:        if (!vm_map_starved_p(kernel_map)) {
                    412:                return CALLBACK_CHAIN_ABORT;
                    413:        }
                    414:        return CALLBACK_CHAIN_CONTINUE;
                    415: }
                    416:
1.142     dyoung    417: struct mbuf *
1.147     dyoung    418: getsombuf(struct socket *so, int type)
1.142     dyoung    419: {
                    420:        struct mbuf *m;
                    421:
1.147     dyoung    422:        m = m_get(M_WAIT, type);
1.142     dyoung    423:        MCLAIM(m, so->so_mowner);
                    424:        return m;
                    425: }
                    426:
                    427: struct mbuf *
                    428: m_intopt(struct socket *so, int val)
                    429: {
                    430:        struct mbuf *m;
                    431:
1.147     dyoung    432:        m = getsombuf(so, MT_SOOPTS);
1.142     dyoung    433:        m->m_len = sizeof(int);
                    434:        *mtod(m, int *) = val;
                    435:        return m;
                    436: }
                    437:
1.119     yamt      438: void
                    439: soinit(void)
                    440: {
                    441:
1.148     ad        442:        mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.160     ad        443:        softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.136     ad        444:        cv_init(&socurkva_cv, "sokva");
                    445:
1.119     yamt      446:        /* Set the initial adjusted socket buffer size. */
                    447:        if (sb_max_set(sb_max))
                    448:                panic("bad initial sb_max value: %lu", sb_max);
                    449:
                    450:        callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback,
                    451:            &sokva_reclaimerentry, NULL, sokva_reclaim_callback);
                    452: }
                    453:
1.1       cgd       454: /*
                    455:  * Socket operation routines.
                    456:  * These routines are called by the routines in
                    457:  * sys_socket.c or from a system process, and
                    458:  * implement the semantics of socket operations by
                    459:  * switching out to the protocol specific routines.
                    460:  */
                    461: /*ARGSUSED*/
1.3       andrew    462: int
1.160     ad        463: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l,
                    464:         struct socket *lockso)
1.1       cgd       465: {
1.99      matt      466:        const struct protosw    *prp;
1.54      lukem     467:        struct socket   *so;
1.115     yamt      468:        uid_t           uid;
1.160     ad        469:        int             error;
                    470:        kmutex_t        *lock;
1.1       cgd       471:
1.132     elad      472:        error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
                    473:            KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
                    474:            KAUTH_ARG(proto));
1.140     dyoung    475:        if (error != 0)
                    476:                return error;
1.127     elad      477:
1.1       cgd       478:        if (proto)
                    479:                prp = pffindproto(dom, proto, type);
                    480:        else
                    481:                prp = pffindtype(dom, type);
1.140     dyoung    482:        if (prp == NULL) {
1.120     ginsbach  483:                /* no support for domain */
                    484:                if (pffinddomain(dom) == 0)
1.140     dyoung    485:                        return EAFNOSUPPORT;
1.120     ginsbach  486:                /* no support for socket type */
                    487:                if (proto == 0 && type != 0)
1.140     dyoung    488:                        return EPROTOTYPE;
                    489:                return EPROTONOSUPPORT;
1.120     ginsbach  490:        }
1.140     dyoung    491:        if (prp->pr_usrreq == NULL)
                    492:                return EPROTONOSUPPORT;
1.1       cgd       493:        if (prp->pr_type != type)
1.140     dyoung    494:                return EPROTOTYPE;
1.160     ad        495:
                    496:        so = soget(true);
1.1       cgd       497:        so->so_type = type;
                    498:        so->so_proto = prp;
1.33      matt      499:        so->so_send = sosend;
                    500:        so->so_receive = soreceive;
1.78      matt      501: #ifdef MBUFTRACE
                    502:        so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
                    503:        so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
                    504:        so->so_mowner = &prp->pr_domain->dom_mowner;
                    505: #endif
1.138     rmind     506:        uid = kauth_cred_geteuid(l->l_cred);
1.115     yamt      507:        so->so_uidinfo = uid_find(uid);
1.160     ad        508:        if (lockso != NULL) {
                    509:                /* Caller wants us to share a lock. */
                    510:                lock = lockso->so_lock;
                    511:                so->so_lock = lock;
                    512:                mutex_obj_hold(lock);
                    513:                mutex_enter(lock);
                    514:        } else {
                    515:                /* Lock assigned and taken during PRU_ATTACH. */
                    516:        }
1.140     dyoung    517:        error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
                    518:            (struct mbuf *)(long)proto, NULL, l);
1.160     ad        519:        KASSERT(solocked(so));
1.140     dyoung    520:        if (error != 0) {
1.1       cgd       521:                so->so_state |= SS_NOFDREF;
                    522:                sofree(so);
1.140     dyoung    523:                return error;
1.1       cgd       524:        }
1.160     ad        525:        sounlock(so);
1.1       cgd       526:        *aso = so;
1.140     dyoung    527:        return 0;
1.1       cgd       528: }
                    529:
1.142     dyoung    530: /* On success, write file descriptor to fdout and return zero.  On
                    531:  * failure, return non-zero; *fdout will be undefined.
                    532:  */
                    533: int
                    534: fsocreate(int domain, struct socket **sop, int type, int protocol,
                    535:     struct lwp *l, int *fdout)
                    536: {
                    537:        struct socket   *so;
                    538:        struct file     *fp;
                    539:        int             fd, error;
                    540:
1.155     ad        541:        if ((error = fd_allocfile(&fp, &fd)) != 0)
1.142     dyoung    542:                return (error);
                    543:        fp->f_flag = FREAD|FWRITE;
                    544:        fp->f_type = DTYPE_SOCKET;
                    545:        fp->f_ops = &socketops;
1.160     ad        546:        error = socreate(domain, &so, type, protocol, l, NULL);
1.142     dyoung    547:        if (error != 0) {
1.155     ad        548:                fd_abort(curproc, fp, fd);
1.142     dyoung    549:        } else {
                    550:                if (sop != NULL)
                    551:                        *sop = so;
                    552:                fp->f_data = so;
1.155     ad        553:                fd_affix(curproc, fp, fd);
1.142     dyoung    554:                *fdout = fd;
                    555:        }
                    556:        return error;
                    557: }
                    558:
1.3       andrew    559: int
1.114     christos  560: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       561: {
1.160     ad        562:        int     error;
1.1       cgd       563:
1.160     ad        564:        solock(so);
1.140     dyoung    565:        error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.160     ad        566:        sounlock(so);
1.140     dyoung    567:        return error;
1.1       cgd       568: }
                    569:
1.3       andrew    570: int
1.150     elad      571: solisten(struct socket *so, int backlog, struct lwp *l)
1.1       cgd       572: {
1.160     ad        573:        int     error;
1.1       cgd       574:
1.160     ad        575:        solock(so);
1.158     ad        576:        if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
                    577:            SS_ISDISCONNECTING)) != 0)
                    578:                return (EOPNOTSUPP);
1.140     dyoung    579:        error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
1.150     elad      580:            NULL, NULL, l);
1.140     dyoung    581:        if (error != 0) {
1.160     ad        582:                sounlock(so);
1.140     dyoung    583:                return error;
1.1       cgd       584:        }
1.63      matt      585:        if (TAILQ_EMPTY(&so->so_q))
1.1       cgd       586:                so->so_options |= SO_ACCEPTCONN;
                    587:        if (backlog < 0)
                    588:                backlog = 0;
1.49      jonathan  589:        so->so_qlimit = min(backlog, somaxconn);
1.160     ad        590:        sounlock(so);
1.140     dyoung    591:        return 0;
1.1       cgd       592: }
                    593:
1.21      christos  594: void
1.54      lukem     595: sofree(struct socket *so)
1.1       cgd       596: {
1.161   ! ad        597:        u_int refs;
1.1       cgd       598:
1.160     ad        599:        KASSERT(solocked(so));
                    600:
                    601:        if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) {
                    602:                sounlock(so);
1.1       cgd       603:                return;
1.160     ad        604:        }
1.43      mycroft   605:        if (so->so_head) {
                    606:                /*
                    607:                 * We must not decommission a socket that's on the accept(2)
                    608:                 * queue.  If we do, then accept(2) may hang after select(2)
                    609:                 * indicated that the listening socket was ready.
                    610:                 */
1.160     ad        611:                if (!soqremque(so, 0)) {
                    612:                        sounlock(so);
1.43      mycroft   613:                        return;
1.160     ad        614:                }
1.43      mycroft   615:        }
1.98      christos  616:        if (so->so_rcv.sb_hiwat)
1.110     christos  617:                (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98      christos  618:                    RLIM_INFINITY);
                    619:        if (so->so_snd.sb_hiwat)
1.110     christos  620:                (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98      christos  621:                    RLIM_INFINITY);
                    622:        sbrelease(&so->so_snd, so);
1.160     ad        623:        KASSERT(!cv_has_waiters(&so->so_cv));
                    624:        KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv));
                    625:        KASSERT(!cv_has_waiters(&so->so_snd.sb_cv));
1.1       cgd       626:        sorflush(so);
1.161   ! ad        627:        refs = so->so_aborting; /* XXX */
1.160     ad        628:        sounlock(so);
1.161   ! ad        629:        if (refs == 0)          /* XXX */
        !           630:                soput(so);
1.1       cgd       631: }
                    632:
                    633: /*
                    634:  * Close a socket on last file table reference removal.
                    635:  * Initiate disconnect if connected.
                    636:  * Free socket when disconnect complete.
                    637:  */
1.3       andrew    638: int
1.54      lukem     639: soclose(struct socket *so)
1.1       cgd       640: {
1.54      lukem     641:        struct socket   *so2;
1.160     ad        642:        int             error;
                    643:        int             error2;
1.1       cgd       644:
1.54      lukem     645:        error = 0;
1.160     ad        646:        solock(so);
1.1       cgd       647:        if (so->so_options & SO_ACCEPTCONN) {
1.160     ad        648:                do {
                    649:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
                    650:                                KASSERT(solocked2(so, so2));
                    651:                                (void) soqremque(so2, 0);
                    652:                                /* soabort drops the lock. */
                    653:                                (void) soabort(so2);
                    654:                                solock(so);
                    655:                                continue;
                    656:                        }
                    657:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
                    658:                                KASSERT(solocked2(so, so2));
                    659:                                (void) soqremque(so2, 1);
                    660:                                /* soabort drops the lock. */
                    661:                                (void) soabort(so2);
                    662:                                solock(so);
                    663:                                continue;
                    664:                        }
                    665:                } while (0);
1.1       cgd       666:        }
                    667:        if (so->so_pcb == 0)
                    668:                goto discard;
                    669:        if (so->so_state & SS_ISCONNECTED) {
                    670:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    671:                        error = sodisconnect(so);
                    672:                        if (error)
                    673:                                goto drop;
                    674:                }
                    675:                if (so->so_options & SO_LINGER) {
1.151     ad        676:                        if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio)
1.1       cgd       677:                                goto drop;
1.21      christos  678:                        while (so->so_state & SS_ISCONNECTED) {
1.160     ad        679:                                error = sowait(so, so->so_linger * hz);
1.21      christos  680:                                if (error)
1.1       cgd       681:                                        break;
1.21      christos  682:                        }
1.1       cgd       683:                }
                    684:        }
1.54      lukem     685:  drop:
1.1       cgd       686:        if (so->so_pcb) {
1.160     ad        687:                error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140     dyoung    688:                    NULL, NULL, NULL, NULL);
1.1       cgd       689:                if (error == 0)
                    690:                        error = error2;
                    691:        }
1.54      lukem     692:  discard:
1.1       cgd       693:        if (so->so_state & SS_NOFDREF)
                    694:                panic("soclose: NOFDREF");
                    695:        so->so_state |= SS_NOFDREF;
                    696:        sofree(so);
                    697:        return (error);
                    698: }
                    699:
                    700: /*
1.160     ad        701:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       702:  */
1.3       andrew    703: int
1.54      lukem     704: soabort(struct socket *so)
1.1       cgd       705: {
1.161   ! ad        706:        u_int refs;
1.139     yamt      707:        int error;
1.160     ad        708:
                    709:        KASSERT(solocked(so));
                    710:        KASSERT(so->so_head == NULL);
1.1       cgd       711:
1.161   ! ad        712:        so->so_aborting++;              /* XXX */
1.140     dyoung    713:        error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
                    714:            NULL, NULL, NULL);
1.161   ! ad        715:        refs = --so->so_aborting;       /* XXX */
1.139     yamt      716:        if (error) {
                    717:                sofree(so);
1.160     ad        718:        } else {
                    719:                sounlock(so);
1.161   ! ad        720:                if (refs == 0)
        !           721:                        sofree(so);
1.139     yamt      722:        }
                    723:        return error;
1.1       cgd       724: }
                    725:
1.3       andrew    726: int
1.54      lukem     727: soaccept(struct socket *so, struct mbuf *nam)
1.1       cgd       728: {
1.160     ad        729:        int     error;
                    730:
                    731:        KASSERT(solocked(so));
1.1       cgd       732:
1.54      lukem     733:        error = 0;
1.1       cgd       734:        if ((so->so_state & SS_NOFDREF) == 0)
                    735:                panic("soaccept: !NOFDREF");
                    736:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   737:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    738:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41      mycroft   739:                error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140     dyoung    740:                    NULL, nam, NULL, NULL);
1.41      mycroft   741:        else
1.53      itojun    742:                error = ECONNABORTED;
1.52      itojun    743:
1.1       cgd       744:        return (error);
                    745: }
                    746:
1.3       andrew    747: int
1.114     christos  748: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       749: {
1.160     ad        750:        int             error;
                    751:
                    752:        KASSERT(solocked(so));
1.1       cgd       753:
                    754:        if (so->so_options & SO_ACCEPTCONN)
                    755:                return (EOPNOTSUPP);
                    756:        /*
                    757:         * If protocol is connection-based, can only connect once.
                    758:         * Otherwise, if connected, try to disconnect first.
                    759:         * This allows user to disconnect by connecting to, e.g.,
                    760:         * a null address.
                    761:         */
                    762:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    763:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
                    764:            (error = sodisconnect(so))))
                    765:                error = EISCONN;
                    766:        else
                    767:                error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140     dyoung    768:                    NULL, nam, NULL, l);
1.1       cgd       769:        return (error);
                    770: }
                    771:
1.3       andrew    772: int
1.54      lukem     773: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       774: {
1.160     ad        775:        int     error;
                    776:
                    777:        KASSERT(solocked2(so1, so2));
1.1       cgd       778:
1.22      mycroft   779:        error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140     dyoung    780:            NULL, (struct mbuf *)so2, NULL, NULL);
1.1       cgd       781:        return (error);
                    782: }
                    783:
1.3       andrew    784: int
1.54      lukem     785: sodisconnect(struct socket *so)
1.1       cgd       786: {
1.160     ad        787:        int     error;
                    788:
                    789:        KASSERT(solocked(so));
1.1       cgd       790:
                    791:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    792:                error = ENOTCONN;
1.160     ad        793:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       794:                error = EALREADY;
1.160     ad        795:        } else {
                    796:                error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
                    797:                    NULL, NULL, NULL, NULL);
1.1       cgd       798:        }
1.117     yamt      799:        sodopendfree();
1.1       cgd       800:        return (error);
                    801: }
                    802:
1.15      mycroft   803: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       804: /*
                    805:  * Send on a socket.
                    806:  * If send must go all at once and message is larger than
                    807:  * send buffering, then hard error.
                    808:  * Lock against other senders.
                    809:  * If must go all at once and not enough room now, then
                    810:  * inform user that this would block and do nothing.
                    811:  * Otherwise, if nonblocking, send as much as possible.
                    812:  * The data to be sent is described by "uio" if nonzero,
                    813:  * otherwise by the mbuf chain "top" (which must be null
                    814:  * if uio is not).  Data provided in mbuf chain must be small
                    815:  * enough to send all at once.
                    816:  *
                    817:  * Returns nonzero on error, timeout or signal; callers
                    818:  * must check for short counts if EINTR/ERESTART are returned.
                    819:  * Data and control buffers are freed on return.
                    820:  */
1.3       andrew    821: int
1.54      lukem     822: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114     christos  823:        struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       824: {
1.54      lukem     825:        struct mbuf     **mp, *m;
1.114     christos  826:        struct proc     *p;
1.58      jdolecek  827:        long            space, len, resid, clen, mlen;
                    828:        int             error, s, dontroute, atomic;
1.54      lukem     829:
1.114     christos  830:        p = l->l_proc;
1.117     yamt      831:        sodopendfree();
1.160     ad        832:        clen = 0;
1.64      thorpej   833:
1.160     ad        834:        /*
                    835:         * solock() provides atomicity of access.  splsoftnet() prevents
                    836:         * protocol processing soft interrupts from interrupting us and
                    837:         * blocking (expensive).
                    838:         */
                    839:        s = splsoftnet();
                    840:        solock(so);
1.54      lukem     841:        atomic = sosendallatonce(so) || top;
1.1       cgd       842:        if (uio)
                    843:                resid = uio->uio_resid;
                    844:        else
                    845:                resid = top->m_pkthdr.len;
1.7       cgd       846:        /*
                    847:         * In theory resid should be unsigned.
                    848:         * However, space must be signed, as it might be less than 0
                    849:         * if we over-committed, and we must use a signed comparison
                    850:         * of space and resid.  On the other hand, a negative resid
                    851:         * causes us to loop sending 0-length segments to the protocol.
                    852:         */
1.29      mycroft   853:        if (resid < 0) {
                    854:                error = EINVAL;
                    855:                goto out;
                    856:        }
1.1       cgd       857:        dontroute =
                    858:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    859:            (so->so_proto->pr_flags & PR_ATOMIC);
1.157     ad        860:        if (l)
                    861:                l->l_ru.ru_msgsnd++;
1.1       cgd       862:        if (control)
                    863:                clen = control->m_len;
1.54      lukem     864:  restart:
1.21      christos  865:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       866:                goto out;
                    867:        do {
1.160     ad        868:                if (so->so_state & SS_CANTSENDMORE) {
                    869:                        error = EPIPE;
                    870:                        goto release;
                    871:                }
1.48      thorpej   872:                if (so->so_error) {
                    873:                        error = so->so_error;
                    874:                        so->so_error = 0;
                    875:                        goto release;
                    876:                }
1.1       cgd       877:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    878:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
                    879:                                if ((so->so_state & SS_ISCONFIRMING) == 0 &&
1.160     ad        880:                                    !(resid == 0 && clen != 0)) {
                    881:                                        error = ENOTCONN;
                    882:                                        goto release;
                    883:                                }
                    884:                        } else if (addr == 0) {
                    885:                                error = EDESTADDRREQ;
                    886:                                goto release;
                    887:                        }
1.1       cgd       888:                }
                    889:                space = sbspace(&so->so_snd);
                    890:                if (flags & MSG_OOB)
                    891:                        space += 1024;
1.21      christos  892:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        893:                    clen > so->so_snd.sb_hiwat) {
                    894:                        error = EMSGSIZE;
                    895:                        goto release;
                    896:                }
1.96      mycroft   897:                if (space < resid + clen &&
1.1       cgd       898:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.160     ad        899:                        if (so->so_nbio) {
                    900:                                error = EWOULDBLOCK;
                    901:                                goto release;
                    902:                        }
1.1       cgd       903:                        sbunlock(&so->so_snd);
                    904:                        error = sbwait(&so->so_snd);
                    905:                        if (error)
                    906:                                goto out;
                    907:                        goto restart;
                    908:                }
                    909:                mp = &top;
                    910:                space -= clen;
                    911:                do {
1.45      tv        912:                        if (uio == NULL) {
                    913:                                /*
                    914:                                 * Data is prepackaged in "top".
                    915:                                 */
                    916:                                resid = 0;
                    917:                                if (flags & MSG_EOR)
                    918:                                        top->m_flags |= M_EOR;
                    919:                        } else do {
1.160     ad        920:                                sounlock(so);
                    921:                                splx(s);
1.144     dyoung    922:                                if (top == NULL) {
1.78      matt      923:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv        924:                                        mlen = MHLEN;
                    925:                                        m->m_pkthdr.len = 0;
1.140     dyoung    926:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv        927:                                } else {
1.78      matt      928:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv        929:                                        mlen = MLEN;
                    930:                                }
1.78      matt      931:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt      932:                                if (sock_loan_thresh >= 0 &&
                    933:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                    934:                                    space >= sock_loan_thresh &&
1.64      thorpej   935:                                    (len = sosend_loan(so, uio, m,
                    936:                                                       space)) != 0) {
                    937:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                    938:                                        space -= len;
                    939:                                        goto have_data;
                    940:                                }
1.45      tv        941:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej   942:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78      matt      943:                                        m_clget(m, M_WAIT);
1.45      tv        944:                                        if ((m->m_flags & M_EXT) == 0)
                    945:                                                goto nopages;
                    946:                                        mlen = MCLBYTES;
                    947:                                        if (atomic && top == 0) {
1.58      jdolecek  948:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem     949:                                                    resid);
1.45      tv        950:                                                m->m_data += max_hdr;
                    951:                                        } else
1.58      jdolecek  952:                                                len = lmin(MCLBYTES, resid);
1.45      tv        953:                                        space -= len;
                    954:                                } else {
1.64      thorpej   955:  nopages:
                    956:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek  957:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv        958:                                        space -= len;
                    959:                                        /*
                    960:                                         * For datagram protocols, leave room
                    961:                                         * for protocol headers in first mbuf.
                    962:                                         */
                    963:                                        if (atomic && top == 0 && len < mlen)
                    964:                                                MH_ALIGN(m, len);
                    965:                                }
1.144     dyoung    966:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej   967:  have_data:
1.45      tv        968:                                resid = uio->uio_resid;
                    969:                                m->m_len = len;
                    970:                                *mp = m;
                    971:                                top->m_pkthdr.len += len;
1.160     ad        972:                                s = splsoftnet();
                    973:                                solock(so);
1.144     dyoung    974:                                if (error != 0)
1.45      tv        975:                                        goto release;
                    976:                                mp = &m->m_next;
                    977:                                if (resid <= 0) {
                    978:                                        if (flags & MSG_EOR)
                    979:                                                top->m_flags |= M_EOR;
                    980:                                        break;
                    981:                                }
                    982:                        } while (space > 0 && atomic);
1.108     perry     983:
1.160     ad        984:                        if (so->so_state & SS_CANTSENDMORE) {
                    985:                                error = EPIPE;
                    986:                                goto release;
                    987:                        }
1.45      tv        988:                        if (dontroute)
                    989:                                so->so_options |= SO_DONTROUTE;
                    990:                        if (resid > 0)
                    991:                                so->so_state |= SS_MORETOCOME;
1.46      sommerfe  992:                        error = (*so->so_proto->pr_usrreq)(so,
                    993:                            (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.160     ad        994:                            top, addr, control, curlwp);
1.45      tv        995:                        if (dontroute)
                    996:                                so->so_options &= ~SO_DONTROUTE;
                    997:                        if (resid > 0)
                    998:                                so->so_state &= ~SS_MORETOCOME;
                    999:                        clen = 0;
1.144     dyoung   1000:                        control = NULL;
                   1001:                        top = NULL;
1.45      tv       1002:                        mp = &top;
1.144     dyoung   1003:                        if (error != 0)
1.1       cgd      1004:                                goto release;
                   1005:                } while (resid && space > 0);
                   1006:        } while (resid);
                   1007:
1.54      lukem    1008:  release:
1.1       cgd      1009:        sbunlock(&so->so_snd);
1.54      lukem    1010:  out:
1.160     ad       1011:        sounlock(so);
                   1012:        splx(s);
1.1       cgd      1013:        if (top)
                   1014:                m_freem(top);
                   1015:        if (control)
                   1016:                m_freem(control);
                   1017:        return (error);
                   1018: }
                   1019:
                   1020: /*
1.159     ad       1021:  * Following replacement or removal of the first mbuf on the first
                   1022:  * mbuf chain of a socket buffer, push necessary state changes back
                   1023:  * into the socket buffer so that other consumers see the values
                   1024:  * consistently.  'nextrecord' is the callers locally stored value of
                   1025:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1026:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1027:  */
                   1028: static void
                   1029: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1030: {
                   1031:
1.160     ad       1032:        KASSERT(solocked(sb->sb_so));
                   1033:
1.159     ad       1034:        /*
                   1035:         * First, update for the new value of nextrecord.  If necessary,
                   1036:         * make it the first record.
                   1037:         */
                   1038:        if (sb->sb_mb != NULL)
                   1039:                sb->sb_mb->m_nextpkt = nextrecord;
                   1040:        else
                   1041:                sb->sb_mb = nextrecord;
                   1042:
                   1043:         /*
                   1044:          * Now update any dependent socket buffer fields to reflect
                   1045:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1046:          * the addition of a second clause that takes care of the
                   1047:          * case where sb_mb has been updated, but remains the last
                   1048:          * record.
                   1049:          */
                   1050:         if (sb->sb_mb == NULL) {
                   1051:                 sb->sb_mbtail = NULL;
                   1052:                 sb->sb_lastrecord = NULL;
                   1053:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1054:                 sb->sb_lastrecord = sb->sb_mb;
                   1055: }
                   1056:
                   1057: /*
1.1       cgd      1058:  * Implement receive operations on a socket.
                   1059:  * We depend on the way that records are added to the sockbuf
                   1060:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1061:  * must begin with an address if the protocol so specifies,
                   1062:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1063:  * and then zero or more mbufs of data.
                   1064:  * In order to avoid blocking network interrupts for the entire time here,
                   1065:  * we splx() while doing the actual copy to user space.
                   1066:  * Although the sockbuf is locked, new data may still be appended,
                   1067:  * and thus we must maintain consistency of the sockbuf during that time.
                   1068:  *
                   1069:  * The caller may receive the data as a single mbuf chain by supplying
                   1070:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1071:  * only for the count in uio_resid.
                   1072:  */
1.3       andrew   1073: int
1.54      lukem    1074: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1075:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1076: {
1.116     yamt     1077:        struct lwp *l = curlwp;
1.160     ad       1078:        struct mbuf     *m, **mp, *mt;
1.146     dyoung   1079:        int atomic, flags, len, error, s, offset, moff, type, orig_resid;
1.99      matt     1080:        const struct protosw    *pr;
1.54      lukem    1081:        struct mbuf     *nextrecord;
1.67      he       1082:        int             mbuf_removed = 0;
1.146     dyoung   1083:        const struct domain *dom;
1.64      thorpej  1084:
1.54      lukem    1085:        pr = so->so_proto;
1.146     dyoung   1086:        atomic = pr->pr_flags & PR_ATOMIC;
                   1087:        dom = pr->pr_domain;
1.1       cgd      1088:        mp = mp0;
1.54      lukem    1089:        type = 0;
                   1090:        orig_resid = uio->uio_resid;
1.102     jonathan 1091:
1.144     dyoung   1092:        if (paddr != NULL)
                   1093:                *paddr = NULL;
                   1094:        if (controlp != NULL)
                   1095:                *controlp = NULL;
                   1096:        if (flagsp != NULL)
1.1       cgd      1097:                flags = *flagsp &~ MSG_EOR;
                   1098:        else
                   1099:                flags = 0;
1.66      enami    1100:
                   1101:        if ((flags & MSG_DONTWAIT) == 0)
1.117     yamt     1102:                sodopendfree();
1.66      enami    1103:
1.1       cgd      1104:        if (flags & MSG_OOB) {
                   1105:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1106:                solock(so);
1.17      cgd      1107:                error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140     dyoung   1108:                    (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.160     ad       1109:                sounlock(so);
1.1       cgd      1110:                if (error)
                   1111:                        goto bad;
                   1112:                do {
1.134     christos 1113:                        error = uiomove(mtod(m, void *),
1.1       cgd      1114:                            (int) min(uio->uio_resid, m->m_len), uio);
                   1115:                        m = m_free(m);
1.144     dyoung   1116:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1117:  bad:
1.144     dyoung   1118:                if (m != NULL)
1.1       cgd      1119:                        m_freem(m);
1.144     dyoung   1120:                return error;
1.1       cgd      1121:        }
1.144     dyoung   1122:        if (mp != NULL)
1.140     dyoung   1123:                *mp = NULL;
1.160     ad       1124:
                   1125:        /*
                   1126:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1127:         * protocol processing soft interrupts from interrupting us and
                   1128:         * blocking (expensive).
                   1129:         */
                   1130:        s = splsoftnet();
                   1131:        solock(so);
1.1       cgd      1132:        if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140     dyoung   1133:                (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1       cgd      1134:
1.54      lukem    1135:  restart:
1.160     ad       1136:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1137:                sounlock(so);
                   1138:                splx(s);
1.144     dyoung   1139:                return error;
1.160     ad       1140:        }
1.1       cgd      1141:
                   1142:        m = so->so_rcv.sb_mb;
                   1143:        /*
                   1144:         * If we have less data than requested, block awaiting more
                   1145:         * (subject to any timeout) if:
1.15      mycroft  1146:         *   1. the current count is less than the low water mark,
1.1       cgd      1147:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1148:         *      receive operation at once if we block (resid <= hiwat), or
                   1149:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1150:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1151:         * we have to do the receive in sections, and thus risk returning
                   1152:         * a short count if a timeout or signal occurs after we start.
                   1153:         */
1.144     dyoung   1154:        if (m == NULL ||
                   1155:            ((flags & MSG_DONTWAIT) == 0 &&
                   1156:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1157:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1158:              ((flags & MSG_WAITALL) &&
                   1159:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1160:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1161: #ifdef DIAGNOSTIC
1.144     dyoung   1162:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1163:                        panic("receive 1");
                   1164: #endif
                   1165:                if (so->so_error) {
1.144     dyoung   1166:                        if (m != NULL)
1.15      mycroft  1167:                                goto dontblock;
1.1       cgd      1168:                        error = so->so_error;
                   1169:                        if ((flags & MSG_PEEK) == 0)
                   1170:                                so->so_error = 0;
                   1171:                        goto release;
                   1172:                }
                   1173:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1174:                        if (m != NULL)
1.15      mycroft  1175:                                goto dontblock;
1.1       cgd      1176:                        else
                   1177:                                goto release;
                   1178:                }
1.144     dyoung   1179:                for (; m != NULL; m = m->m_next)
1.1       cgd      1180:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1181:                                m = so->so_rcv.sb_mb;
                   1182:                                goto dontblock;
                   1183:                        }
                   1184:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1185:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1186:                        error = ENOTCONN;
                   1187:                        goto release;
                   1188:                }
                   1189:                if (uio->uio_resid == 0)
                   1190:                        goto release;
1.151     ad       1191:                if (so->so_nbio || (flags & MSG_DONTWAIT)) {
1.1       cgd      1192:                        error = EWOULDBLOCK;
                   1193:                        goto release;
                   1194:                }
1.69      thorpej  1195:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1196:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1197:                sbunlock(&so->so_rcv);
                   1198:                error = sbwait(&so->so_rcv);
1.160     ad       1199:                if (error != 0) {
                   1200:                        sounlock(so);
                   1201:                        splx(s);
1.144     dyoung   1202:                        return error;
1.160     ad       1203:                }
1.1       cgd      1204:                goto restart;
                   1205:        }
1.54      lukem    1206:  dontblock:
1.69      thorpej  1207:        /*
                   1208:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1209:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1210:         * pointer to the next record in the socket buffer.  We must keep the
                   1211:         * various socket buffer pointers and local stack versions of the
                   1212:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1213:         * socket lock, and re-reading them when picking it up.
1.159     ad       1214:         *
                   1215:         * Otherwise, we will race with the network stack appending new data
                   1216:         * or records onto the socket buffer by using inconsistent/stale
                   1217:         * versions of the field, possibly resulting in socket buffer
                   1218:         * corruption.
                   1219:         *
                   1220:         * By holding the high-level sblock(), we prevent simultaneous
                   1221:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1222:         */
1.144     dyoung   1223:        if (l != NULL)
1.157     ad       1224:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1225:        KASSERT(m == so->so_rcv.sb_mb);
                   1226:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1227:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1228:        nextrecord = m->m_nextpkt;
                   1229:        if (pr->pr_flags & PR_ADDR) {
                   1230: #ifdef DIAGNOSTIC
                   1231:                if (m->m_type != MT_SONAME)
                   1232:                        panic("receive 1a");
                   1233: #endif
1.3       andrew   1234:                orig_resid = 0;
1.1       cgd      1235:                if (flags & MSG_PEEK) {
                   1236:                        if (paddr)
                   1237:                                *paddr = m_copy(m, 0, m->m_len);
                   1238:                        m = m->m_next;
                   1239:                } else {
                   1240:                        sbfree(&so->so_rcv, m);
1.67      he       1241:                        mbuf_removed = 1;
1.144     dyoung   1242:                        if (paddr != NULL) {
1.1       cgd      1243:                                *paddr = m;
                   1244:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1245:                                m->m_next = NULL;
1.1       cgd      1246:                                m = so->so_rcv.sb_mb;
                   1247:                        } else {
                   1248:                                MFREE(m, so->so_rcv.sb_mb);
                   1249:                                m = so->so_rcv.sb_mb;
                   1250:                        }
1.159     ad       1251:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1252:                }
                   1253:        }
1.159     ad       1254:
                   1255:        /*
                   1256:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1257:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1258:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1259:         * perform externalization (or freeing if controlp == NULL).
                   1260:         */
                   1261:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1262:                struct mbuf *cm = NULL, *cmn;
                   1263:                struct mbuf **cme = &cm;
                   1264:
                   1265:                do {
                   1266:                        if (flags & MSG_PEEK) {
                   1267:                                if (controlp != NULL) {
                   1268:                                        *controlp = m_copy(m, 0, m->m_len);
                   1269:                                        controlp = &(*controlp)->m_next;
                   1270:                                }
                   1271:                                m = m->m_next;
                   1272:                        } else {
                   1273:                                sbfree(&so->so_rcv, m);
1.1       cgd      1274:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1275:                                m->m_next = NULL;
1.159     ad       1276:                                *cme = m;
                   1277:                                cme = &(*cme)->m_next;
1.1       cgd      1278:                                m = so->so_rcv.sb_mb;
1.159     ad       1279:                        }
                   1280:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1281:                if ((flags & MSG_PEEK) == 0)
                   1282:                        sbsync(&so->so_rcv, nextrecord);
                   1283:                for (; cm != NULL; cm = cmn) {
                   1284:                        cmn = cm->m_next;
                   1285:                        cm->m_next = NULL;
                   1286:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1287:                        if (controlp != NULL) {
                   1288:                                if (dom->dom_externalize != NULL &&
                   1289:                                    type == SCM_RIGHTS) {
1.160     ad       1290:                                        sounlock(so);
1.159     ad       1291:                                        splx(s);
                   1292:                                        error = (*dom->dom_externalize)(cm, l);
                   1293:                                        s = splsoftnet();
1.160     ad       1294:                                        solock(so);
1.159     ad       1295:                                }
                   1296:                                *controlp = cm;
                   1297:                                while (*controlp != NULL)
                   1298:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1299:                        } else {
1.106     itojun   1300:                                /*
                   1301:                                 * Dispose of any SCM_RIGHTS message that went
                   1302:                                 * through the read path rather than recv.
                   1303:                                 */
1.159     ad       1304:                                if (dom->dom_dispose != NULL &&
                   1305:                                    type == SCM_RIGHTS) {
1.160     ad       1306:                                        sounlock(so);
1.159     ad       1307:                                        (*dom->dom_dispose)(cm);
1.160     ad       1308:                                        solock(so);
1.159     ad       1309:                                }
                   1310:                                m_freem(cm);
1.1       cgd      1311:                        }
                   1312:                }
1.159     ad       1313:                if (m != NULL)
                   1314:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1315:                else
                   1316:                        nextrecord = so->so_rcv.sb_mb;
                   1317:                orig_resid = 0;
1.1       cgd      1318:        }
1.69      thorpej  1319:
1.159     ad       1320:        /* If m is non-NULL, we have some data to read. */
                   1321:        if (__predict_true(m != NULL)) {
1.1       cgd      1322:                type = m->m_type;
                   1323:                if (type == MT_OOBDATA)
                   1324:                        flags |= MSG_OOB;
                   1325:        }
1.69      thorpej  1326:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1327:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1328:
1.1       cgd      1329:        moff = 0;
                   1330:        offset = 0;
1.144     dyoung   1331:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1332:                if (m->m_type == MT_OOBDATA) {
                   1333:                        if (type != MT_OOBDATA)
                   1334:                                break;
                   1335:                } else if (type == MT_OOBDATA)
                   1336:                        break;
                   1337: #ifdef DIAGNOSTIC
                   1338:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1339:                        panic("receive 3");
                   1340: #endif
                   1341:                so->so_state &= ~SS_RCVATMARK;
                   1342:                len = uio->uio_resid;
                   1343:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1344:                        len = so->so_oobmark - offset;
                   1345:                if (len > m->m_len - moff)
                   1346:                        len = m->m_len - moff;
                   1347:                /*
                   1348:                 * If mp is set, just pass back the mbufs.
                   1349:                 * Otherwise copy them out via the uio, then free.
                   1350:                 * Sockbuf must be consistent here (points to current mbuf,
                   1351:                 * it points to next record) when we drop priority;
                   1352:                 * we must note any additions to the sockbuf when we
                   1353:                 * block interrupts again.
                   1354:                 */
1.144     dyoung   1355:                if (mp == NULL) {
1.69      thorpej  1356:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1357:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1358:                        sounlock(so);
1.1       cgd      1359:                        splx(s);
1.134     christos 1360:                        error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20      mycroft  1361:                        s = splsoftnet();
1.160     ad       1362:                        solock(so);
1.144     dyoung   1363:                        if (error != 0) {
1.67      he       1364:                                /*
                   1365:                                 * If any part of the record has been removed
                   1366:                                 * (such as the MT_SONAME mbuf, which will
                   1367:                                 * happen when PR_ADDR, and thus also
                   1368:                                 * PR_ATOMIC, is set), then drop the entire
                   1369:                                 * record to maintain the atomicity of the
                   1370:                                 * receive operation.
                   1371:                                 *
                   1372:                                 * This avoids a later panic("receive 1a")
                   1373:                                 * when compiled with DIAGNOSTIC.
                   1374:                                 */
1.146     dyoung   1375:                                if (m && mbuf_removed && atomic)
1.67      he       1376:                                        (void) sbdroprecord(&so->so_rcv);
                   1377:
1.57      jdolecek 1378:                                goto release;
1.67      he       1379:                        }
1.1       cgd      1380:                } else
                   1381:                        uio->uio_resid -= len;
                   1382:                if (len == m->m_len - moff) {
                   1383:                        if (m->m_flags & M_EOR)
                   1384:                                flags |= MSG_EOR;
                   1385:                        if (flags & MSG_PEEK) {
                   1386:                                m = m->m_next;
                   1387:                                moff = 0;
                   1388:                        } else {
                   1389:                                nextrecord = m->m_nextpkt;
                   1390:                                sbfree(&so->so_rcv, m);
                   1391:                                if (mp) {
                   1392:                                        *mp = m;
                   1393:                                        mp = &m->m_next;
                   1394:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1395:                                        *mp = NULL;
1.1       cgd      1396:                                } else {
                   1397:                                        MFREE(m, so->so_rcv.sb_mb);
                   1398:                                        m = so->so_rcv.sb_mb;
                   1399:                                }
1.69      thorpej  1400:                                /*
                   1401:                                 * If m != NULL, we also know that
                   1402:                                 * so->so_rcv.sb_mb != NULL.
                   1403:                                 */
                   1404:                                KASSERT(so->so_rcv.sb_mb == m);
                   1405:                                if (m) {
1.1       cgd      1406:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1407:                                        if (nextrecord == NULL)
                   1408:                                                so->so_rcv.sb_lastrecord = m;
                   1409:                                } else {
                   1410:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1411:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1412:                                }
                   1413:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1414:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1415:                        }
1.144     dyoung   1416:                } else if (flags & MSG_PEEK)
                   1417:                        moff += len;
                   1418:                else {
1.160     ad       1419:                        if (mp != NULL) {
                   1420:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1421:                                if (__predict_false(mt == NULL)) {
                   1422:                                        sounlock(so);
                   1423:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1424:                                        solock(so);
                   1425:                                }
                   1426:                                *mp = mt;
                   1427:                        }
1.144     dyoung   1428:                        m->m_data += len;
                   1429:                        m->m_len -= len;
                   1430:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1431:                }
                   1432:                if (so->so_oobmark) {
                   1433:                        if ((flags & MSG_PEEK) == 0) {
                   1434:                                so->so_oobmark -= len;
                   1435:                                if (so->so_oobmark == 0) {
                   1436:                                        so->so_state |= SS_RCVATMARK;
                   1437:                                        break;
                   1438:                                }
1.7       cgd      1439:                        } else {
1.1       cgd      1440:                                offset += len;
1.7       cgd      1441:                                if (offset == so->so_oobmark)
                   1442:                                        break;
                   1443:                        }
1.1       cgd      1444:                }
                   1445:                if (flags & MSG_EOR)
                   1446:                        break;
                   1447:                /*
                   1448:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1449:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1450:                 * termination.  If a signal/timeout occurs, return
                   1451:                 * with a short count but without error.
                   1452:                 * Keep sockbuf locked against other readers.
                   1453:                 */
1.144     dyoung   1454:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1455:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1456:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1457:                                break;
1.68      matt     1458:                        /*
                   1459:                         * If we are peeking and the socket receive buffer is
                   1460:                         * full, stop since we can't get more data to peek at.
                   1461:                         */
                   1462:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1463:                                break;
                   1464:                        /*
                   1465:                         * If we've drained the socket buffer, tell the
                   1466:                         * protocol in case it needs to do something to
                   1467:                         * get it filled again.
                   1468:                         */
                   1469:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
                   1470:                                (*pr->pr_usrreq)(so, PRU_RCVD,
1.140     dyoung   1471:                                    NULL, (struct mbuf *)(long)flags, NULL, l);
1.69      thorpej  1472:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1473:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1       cgd      1474:                        error = sbwait(&so->so_rcv);
1.144     dyoung   1475:                        if (error != 0) {
1.1       cgd      1476:                                sbunlock(&so->so_rcv);
1.160     ad       1477:                                sounlock(so);
1.1       cgd      1478:                                splx(s);
1.144     dyoung   1479:                                return 0;
1.1       cgd      1480:                        }
1.21      christos 1481:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1482:                                nextrecord = m->m_nextpkt;
                   1483:                }
                   1484:        }
1.3       andrew   1485:
1.146     dyoung   1486:        if (m && atomic) {
1.3       andrew   1487:                flags |= MSG_TRUNC;
                   1488:                if ((flags & MSG_PEEK) == 0)
                   1489:                        (void) sbdroprecord(&so->so_rcv);
                   1490:        }
1.1       cgd      1491:        if ((flags & MSG_PEEK) == 0) {
1.144     dyoung   1492:                if (m == NULL) {
1.69      thorpej  1493:                        /*
1.70      thorpej  1494:                         * First part is an inline SB_EMPTY_FIXUP().  Second
1.69      thorpej  1495:                         * part makes sure sb_lastrecord is up-to-date if
                   1496:                         * there is still data in the socket buffer.
                   1497:                         */
1.1       cgd      1498:                        so->so_rcv.sb_mb = nextrecord;
1.69      thorpej  1499:                        if (so->so_rcv.sb_mb == NULL) {
                   1500:                                so->so_rcv.sb_mbtail = NULL;
                   1501:                                so->so_rcv.sb_lastrecord = NULL;
                   1502:                        } else if (nextrecord->m_nextpkt == NULL)
                   1503:                                so->so_rcv.sb_lastrecord = nextrecord;
                   1504:                }
                   1505:                SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
                   1506:                SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1       cgd      1507:                if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140     dyoung   1508:                        (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
                   1509:                            (struct mbuf *)(long)flags, NULL, l);
1.1       cgd      1510:        }
1.3       andrew   1511:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1512:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1513:                sbunlock(&so->so_rcv);
                   1514:                goto restart;
                   1515:        }
1.108     perry    1516:
1.144     dyoung   1517:        if (flagsp != NULL)
1.1       cgd      1518:                *flagsp |= flags;
1.54      lukem    1519:  release:
1.1       cgd      1520:        sbunlock(&so->so_rcv);
1.160     ad       1521:        sounlock(so);
1.1       cgd      1522:        splx(s);
1.144     dyoung   1523:        return error;
1.1       cgd      1524: }
                   1525:
1.14      mycroft  1526: int
1.54      lukem    1527: soshutdown(struct socket *so, int how)
1.1       cgd      1528: {
1.99      matt     1529:        const struct protosw    *pr;
1.160     ad       1530:        int     error;
                   1531:
                   1532:        KASSERT(solocked(so));
1.34      kleink   1533:
1.54      lukem    1534:        pr = so->so_proto;
1.34      kleink   1535:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1536:                return (EINVAL);
1.1       cgd      1537:
1.160     ad       1538:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1539:                sorflush(so);
1.160     ad       1540:                error = 0;
                   1541:        }
1.34      kleink   1542:        if (how == SHUT_WR || how == SHUT_RDWR)
1.160     ad       1543:                error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1.140     dyoung   1544:                    NULL, NULL, NULL);
1.160     ad       1545:
                   1546:        return error;
1.1       cgd      1547: }
                   1548:
1.14      mycroft  1549: void
1.54      lukem    1550: sorflush(struct socket *so)
1.1       cgd      1551: {
1.54      lukem    1552:        struct sockbuf  *sb, asb;
1.99      matt     1553:        const struct protosw    *pr;
1.160     ad       1554:
                   1555:        KASSERT(solocked(so));
1.1       cgd      1556:
1.54      lukem    1557:        sb = &so->so_rcv;
                   1558:        pr = so->so_proto;
1.160     ad       1559:        socantrcvmore(so);
1.1       cgd      1560:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1561:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1562:        sbunlock(sb);
                   1563:        asb = *sb;
1.86      wrstuden 1564:        /*
                   1565:         * Clear most of the sockbuf structure, but leave some of the
                   1566:         * fields valid.
                   1567:         */
                   1568:        memset(&sb->sb_startzero, 0,
                   1569:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1570:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1571:                sounlock(so);
1.1       cgd      1572:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1573:                solock(so);
                   1574:        }
1.98      christos 1575:        sbrelease(&asb, so);
1.1       cgd      1576: }
                   1577:
1.142     dyoung   1578: static int
                   1579: sosetopt1(struct socket *so, int level, int optname, struct mbuf *m)
1.1       cgd      1580: {
1.142     dyoung   1581:        int optval, val;
1.130     christos 1582:        struct linger   *l;
1.141     yamt     1583:        struct sockbuf  *sb;
1.142     dyoung   1584:        struct timeval *tv;
                   1585:
                   1586:        switch (optname) {
                   1587:
                   1588:        case SO_LINGER:
                   1589:                if (m == NULL || m->m_len != sizeof(struct linger))
                   1590:                        return EINVAL;
                   1591:                l = mtod(m, struct linger *);
                   1592:                if (l->l_linger < 0 || l->l_linger > USHRT_MAX ||
                   1593:                    l->l_linger > (INT_MAX / hz))
                   1594:                        return EDOM;
                   1595:                so->so_linger = l->l_linger;
                   1596:                if (l->l_onoff)
                   1597:                        so->so_options |= SO_LINGER;
                   1598:                else
                   1599:                        so->so_options &= ~SO_LINGER;
                   1600:                break;
1.1       cgd      1601:
1.142     dyoung   1602:        case SO_DEBUG:
                   1603:        case SO_KEEPALIVE:
                   1604:        case SO_DONTROUTE:
                   1605:        case SO_USELOOPBACK:
                   1606:        case SO_BROADCAST:
                   1607:        case SO_REUSEADDR:
                   1608:        case SO_REUSEPORT:
                   1609:        case SO_OOBINLINE:
                   1610:        case SO_TIMESTAMP:
                   1611:                if (m == NULL || m->m_len < sizeof(int))
                   1612:                        return EINVAL;
                   1613:                if (*mtod(m, int *))
                   1614:                        so->so_options |= optname;
                   1615:                else
                   1616:                        so->so_options &= ~optname;
                   1617:                break;
                   1618:
                   1619:        case SO_SNDBUF:
                   1620:        case SO_RCVBUF:
                   1621:        case SO_SNDLOWAT:
                   1622:        case SO_RCVLOWAT:
                   1623:                if (m == NULL || m->m_len < sizeof(int))
                   1624:                        return EINVAL;
1.1       cgd      1625:
1.142     dyoung   1626:                /*
                   1627:                 * Values < 1 make no sense for any of these
                   1628:                 * options, so disallow them.
                   1629:                 */
                   1630:                optval = *mtod(m, int *);
                   1631:                if (optval < 1)
                   1632:                        return EINVAL;
1.1       cgd      1633:
1.142     dyoung   1634:                switch (optname) {
1.1       cgd      1635:
                   1636:                case SO_SNDBUF:
                   1637:                case SO_RCVBUF:
1.142     dyoung   1638:                        sb = (optname == SO_SNDBUF) ?
                   1639:                            &so->so_snd : &so->so_rcv;
                   1640:                        if (sbreserve(sb, (u_long)optval, so) == 0)
                   1641:                                return ENOBUFS;
                   1642:                        sb->sb_flags &= ~SB_AUTOSIZE;
                   1643:                        break;
                   1644:
                   1645:                /*
                   1646:                 * Make sure the low-water is never greater than
                   1647:                 * the high-water.
                   1648:                 */
1.1       cgd      1649:                case SO_SNDLOWAT:
1.142     dyoung   1650:                        so->so_snd.sb_lowat =
                   1651:                            (optval > so->so_snd.sb_hiwat) ?
                   1652:                            so->so_snd.sb_hiwat : optval;
                   1653:                        break;
1.1       cgd      1654:                case SO_RCVLOWAT:
1.142     dyoung   1655:                        so->so_rcv.sb_lowat =
                   1656:                            (optval > so->so_rcv.sb_hiwat) ?
                   1657:                            so->so_rcv.sb_hiwat : optval;
                   1658:                        break;
                   1659:                }
                   1660:                break;
1.28      thorpej  1661:
1.142     dyoung   1662:        case SO_SNDTIMEO:
                   1663:        case SO_RCVTIMEO:
                   1664:                if (m == NULL || m->m_len < sizeof(*tv))
                   1665:                        return EINVAL;
                   1666:                tv = mtod(m, struct timeval *);
                   1667:                if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz)
                   1668:                        return EDOM;
                   1669:                val = tv->tv_sec * hz + tv->tv_usec / tick;
                   1670:                if (val == 0 && tv->tv_usec != 0)
                   1671:                        val = 1;
1.28      thorpej  1672:
1.142     dyoung   1673:                switch (optname) {
1.28      thorpej  1674:
1.142     dyoung   1675:                case SO_SNDTIMEO:
                   1676:                        so->so_snd.sb_timeo = val;
1.1       cgd      1677:                        break;
                   1678:                case SO_RCVTIMEO:
1.142     dyoung   1679:                        so->so_rcv.sb_timeo = val;
                   1680:                        break;
                   1681:                }
                   1682:                break;
1.1       cgd      1683:
1.142     dyoung   1684:        default:
                   1685:                return ENOPROTOOPT;
                   1686:        }
                   1687:        return 0;
                   1688: }
1.1       cgd      1689:
1.142     dyoung   1690: int
                   1691: sosetopt(struct socket *so, int level, int optname, struct mbuf *m)
                   1692: {
                   1693:        int error, prerr;
1.1       cgd      1694:
1.160     ad       1695:        solock(so);
1.142     dyoung   1696:        if (level == SOL_SOCKET)
                   1697:                error = sosetopt1(so, level, optname, m);
                   1698:        else
                   1699:                error = ENOPROTOOPT;
1.1       cgd      1700:
1.142     dyoung   1701:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1702:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1703:                /* give the protocol stack a shot */
                   1704:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, level,
                   1705:                    optname, &m);
                   1706:                if (prerr == 0)
                   1707:                        error = 0;
                   1708:                else if (prerr != ENOPROTOOPT)
                   1709:                        error = prerr;
                   1710:        } else if (m != NULL)
                   1711:                (void)m_free(m);
1.160     ad       1712:        sounlock(so);
1.142     dyoung   1713:        return error;
1.1       cgd      1714: }
                   1715:
1.14      mycroft  1716: int
1.54      lukem    1717: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1       cgd      1718: {
1.54      lukem    1719:        struct mbuf     *m;
1.160     ad       1720:        int             error;
1.1       cgd      1721:
1.160     ad       1722:        solock(so);
1.1       cgd      1723:        if (level != SOL_SOCKET) {
                   1724:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       1725:                        error = ((*so->so_proto->pr_ctloutput)
1.1       cgd      1726:                                  (PRCO_GETOPT, so, level, optname, mp));
                   1727:                } else
1.160     ad       1728:                        error = (ENOPROTOOPT);
1.1       cgd      1729:        } else {
                   1730:                m = m_get(M_WAIT, MT_SOOPTS);
1.36      perry    1731:                m->m_len = sizeof(int);
1.1       cgd      1732:
                   1733:                switch (optname) {
                   1734:
                   1735:                case SO_LINGER:
1.36      perry    1736:                        m->m_len = sizeof(struct linger);
1.1       cgd      1737:                        mtod(m, struct linger *)->l_onoff =
1.131     christos 1738:                            (so->so_options & SO_LINGER) ? 1 : 0;
1.1       cgd      1739:                        mtod(m, struct linger *)->l_linger = so->so_linger;
                   1740:                        break;
                   1741:
                   1742:                case SO_USELOOPBACK:
                   1743:                case SO_DONTROUTE:
                   1744:                case SO_DEBUG:
                   1745:                case SO_KEEPALIVE:
                   1746:                case SO_REUSEADDR:
1.15      mycroft  1747:                case SO_REUSEPORT:
1.1       cgd      1748:                case SO_BROADCAST:
                   1749:                case SO_OOBINLINE:
1.26      thorpej  1750:                case SO_TIMESTAMP:
1.131     christos 1751:                        *mtod(m, int *) = (so->so_options & optname) ? 1 : 0;
1.1       cgd      1752:                        break;
                   1753:
                   1754:                case SO_TYPE:
                   1755:                        *mtod(m, int *) = so->so_type;
                   1756:                        break;
                   1757:
                   1758:                case SO_ERROR:
                   1759:                        *mtod(m, int *) = so->so_error;
                   1760:                        so->so_error = 0;
                   1761:                        break;
                   1762:
                   1763:                case SO_SNDBUF:
                   1764:                        *mtod(m, int *) = so->so_snd.sb_hiwat;
                   1765:                        break;
                   1766:
                   1767:                case SO_RCVBUF:
                   1768:                        *mtod(m, int *) = so->so_rcv.sb_hiwat;
                   1769:                        break;
                   1770:
                   1771:                case SO_SNDLOWAT:
                   1772:                        *mtod(m, int *) = so->so_snd.sb_lowat;
                   1773:                        break;
                   1774:
                   1775:                case SO_RCVLOWAT:
                   1776:                        *mtod(m, int *) = so->so_rcv.sb_lowat;
                   1777:                        break;
                   1778:
                   1779:                case SO_SNDTIMEO:
                   1780:                case SO_RCVTIMEO:
                   1781:                    {
                   1782:                        int val = (optname == SO_SNDTIMEO ?
                   1783:                             so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1784:
                   1785:                        m->m_len = sizeof(struct timeval);
                   1786:                        mtod(m, struct timeval *)->tv_sec = val / hz;
                   1787:                        mtod(m, struct timeval *)->tv_usec =
1.27      kleink   1788:                            (val % hz) * tick;
1.1       cgd      1789:                        break;
                   1790:                    }
                   1791:
1.107     darrenr  1792:                case SO_OVERFLOWED:
                   1793:                        *mtod(m, int *) = so->so_rcv.sb_overflowed;
                   1794:                        break;
                   1795:
1.1       cgd      1796:                default:
1.160     ad       1797:                        sounlock(so);
1.1       cgd      1798:                        (void)m_free(m);
                   1799:                        return (ENOPROTOOPT);
                   1800:                }
                   1801:                *mp = m;
1.160     ad       1802:                error = 0;
1.1       cgd      1803:        }
1.160     ad       1804:
                   1805:        sounlock(so);
                   1806:        return (error);
1.1       cgd      1807: }
                   1808:
1.14      mycroft  1809: void
1.54      lukem    1810: sohasoutofband(struct socket *so)
1.1       cgd      1811: {
1.153     rmind    1812:
1.90      christos 1813:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.153     rmind    1814:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, 0);
1.1       cgd      1815: }
1.72      jdolecek 1816:
                   1817: static void
                   1818: filt_sordetach(struct knote *kn)
                   1819: {
                   1820:        struct socket   *so;
                   1821:
1.155     ad       1822:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       1823:        solock(so);
1.73      christos 1824:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   1825:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 1826:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       1827:        sounlock(so);
1.72      jdolecek 1828: }
                   1829:
                   1830: /*ARGSUSED*/
                   1831: static int
1.129     yamt     1832: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 1833: {
                   1834:        struct socket   *so;
1.160     ad       1835:        int rv;
1.72      jdolecek 1836:
1.155     ad       1837:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       1838:        if (hint != NOTE_SUBMIT)
                   1839:                solock(so);
1.72      jdolecek 1840:        kn->kn_data = so->so_rcv.sb_cc;
                   1841:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    1842:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 1843:                kn->kn_fflags = so->so_error;
1.160     ad       1844:                rv = 1;
                   1845:        } else if (so->so_error)        /* temporary udp error */
                   1846:                rv = 1;
                   1847:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   1848:                rv = (kn->kn_data >= kn->kn_sdata);
                   1849:        else
                   1850:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   1851:        if (hint != NOTE_SUBMIT)
                   1852:                sounlock(so);
                   1853:        return rv;
1.72      jdolecek 1854: }
                   1855:
                   1856: static void
                   1857: filt_sowdetach(struct knote *kn)
                   1858: {
                   1859:        struct socket   *so;
                   1860:
1.155     ad       1861:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       1862:        solock(so);
1.73      christos 1863:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   1864:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 1865:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       1866:        sounlock(so);
1.72      jdolecek 1867: }
                   1868:
                   1869: /*ARGSUSED*/
                   1870: static int
1.129     yamt     1871: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 1872: {
                   1873:        struct socket   *so;
1.160     ad       1874:        int rv;
1.72      jdolecek 1875:
1.155     ad       1876:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       1877:        if (hint != NOTE_SUBMIT)
                   1878:                solock(so);
1.72      jdolecek 1879:        kn->kn_data = sbspace(&so->so_snd);
                   1880:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    1881:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 1882:                kn->kn_fflags = so->so_error;
1.160     ad       1883:                rv = 1;
                   1884:        } else if (so->so_error)        /* temporary udp error */
                   1885:                rv = 1;
                   1886:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 1887:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       1888:                rv = 0;
                   1889:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   1890:                rv = (kn->kn_data >= kn->kn_sdata);
                   1891:        else
                   1892:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   1893:        if (hint != NOTE_SUBMIT)
                   1894:                sounlock(so);
                   1895:        return rv;
1.72      jdolecek 1896: }
                   1897:
                   1898: /*ARGSUSED*/
                   1899: static int
1.129     yamt     1900: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 1901: {
                   1902:        struct socket   *so;
1.160     ad       1903:        int rv;
1.72      jdolecek 1904:
1.155     ad       1905:        so = ((file_t *)kn->kn_obj)->f_data;
1.72      jdolecek 1906:
                   1907:        /*
                   1908:         * Set kn_data to number of incoming connections, not
                   1909:         * counting partial (incomplete) connections.
1.108     perry    1910:         */
1.160     ad       1911:        if (hint != NOTE_SUBMIT)
                   1912:                solock(so);
1.72      jdolecek 1913:        kn->kn_data = so->so_qlen;
1.160     ad       1914:        rv = (kn->kn_data > 0);
                   1915:        if (hint != NOTE_SUBMIT)
                   1916:                sounlock(so);
                   1917:        return rv;
1.72      jdolecek 1918: }
                   1919:
                   1920: static const struct filterops solisten_filtops =
                   1921:        { 1, NULL, filt_sordetach, filt_solisten };
                   1922: static const struct filterops soread_filtops =
                   1923:        { 1, NULL, filt_sordetach, filt_soread };
                   1924: static const struct filterops sowrite_filtops =
                   1925:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   1926:
                   1927: int
1.129     yamt     1928: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 1929: {
                   1930:        struct socket   *so;
                   1931:        struct sockbuf  *sb;
                   1932:
1.155     ad       1933:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       1934:        solock(so);
1.72      jdolecek 1935:        switch (kn->kn_filter) {
                   1936:        case EVFILT_READ:
                   1937:                if (so->so_options & SO_ACCEPTCONN)
                   1938:                        kn->kn_fop = &solisten_filtops;
                   1939:                else
                   1940:                        kn->kn_fop = &soread_filtops;
                   1941:                sb = &so->so_rcv;
                   1942:                break;
                   1943:        case EVFILT_WRITE:
                   1944:                kn->kn_fop = &sowrite_filtops;
                   1945:                sb = &so->so_snd;
                   1946:                break;
                   1947:        default:
1.160     ad       1948:                sounlock(so);
1.149     pooka    1949:                return (EINVAL);
1.72      jdolecek 1950:        }
1.73      christos 1951:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 1952:        sb->sb_flags |= SB_KNOTE;
1.160     ad       1953:        sounlock(so);
1.72      jdolecek 1954:        return (0);
                   1955: }
                   1956:
1.154     ad       1957: static int
                   1958: sodopoll(struct socket *so, int events)
                   1959: {
                   1960:        int revents;
                   1961:
                   1962:        revents = 0;
                   1963:
                   1964:        if (events & (POLLIN | POLLRDNORM))
                   1965:                if (soreadable(so))
                   1966:                        revents |= events & (POLLIN | POLLRDNORM);
                   1967:
                   1968:        if (events & (POLLOUT | POLLWRNORM))
                   1969:                if (sowritable(so))
                   1970:                        revents |= events & (POLLOUT | POLLWRNORM);
                   1971:
                   1972:        if (events & (POLLPRI | POLLRDBAND))
                   1973:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   1974:                        revents |= events & (POLLPRI | POLLRDBAND);
                   1975:
                   1976:        return revents;
                   1977: }
                   1978:
                   1979: int
                   1980: sopoll(struct socket *so, int events)
                   1981: {
                   1982:        int revents = 0;
                   1983:
1.160     ad       1984: #ifndef DIAGNOSTIC
                   1985:        /*
                   1986:         * Do a quick, unlocked check in expectation that the socket
                   1987:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   1988:         * as the solocked() assertions will fail.
                   1989:         */
1.154     ad       1990:        if ((revents = sodopoll(so, events)) != 0)
                   1991:                return revents;
1.160     ad       1992: #endif
1.154     ad       1993:
1.160     ad       1994:        solock(so);
1.154     ad       1995:        if ((revents = sodopoll(so, events)) == 0) {
                   1996:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   1997:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       1998:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       1999:                }
                   2000:
                   2001:                if (events & (POLLOUT | POLLWRNORM)) {
                   2002:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2003:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2004:                }
                   2005:        }
1.160     ad       2006:        sounlock(so);
1.154     ad       2007:
                   2008:        return revents;
                   2009: }
                   2010:
                   2011:
1.94      yamt     2012: #include <sys/sysctl.h>
                   2013:
                   2014: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
                   2015:
                   2016: /*
                   2017:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2018:  * value is not too small.
                   2019:  * (XXX should we maybe make sure it's not too large as well?)
                   2020:  */
                   2021: static int
                   2022: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2023: {
                   2024:        int error, new_somaxkva;
                   2025:        struct sysctlnode node;
                   2026:
                   2027:        new_somaxkva = somaxkva;
                   2028:        node = *rnode;
                   2029:        node.sysctl_data = &new_somaxkva;
                   2030:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2031:        if (error || newp == NULL)
                   2032:                return (error);
                   2033:
                   2034:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2035:                return (EINVAL);
                   2036:
1.136     ad       2037:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2038:        somaxkva = new_somaxkva;
1.136     ad       2039:        cv_broadcast(&socurkva_cv);
                   2040:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2041:
                   2042:        return (error);
                   2043: }
                   2044:
                   2045: SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup")
                   2046: {
                   2047:
1.97      atatat   2048:        sysctl_createv(clog, 0, NULL, NULL,
                   2049:                       CTLFLAG_PERMANENT,
                   2050:                       CTLTYPE_NODE, "kern", NULL,
                   2051:                       NULL, 0, NULL, 0,
                   2052:                       CTL_KERN, CTL_EOL);
                   2053:
                   2054:        sysctl_createv(clog, 0, NULL, NULL,
                   2055:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2056:                       CTLTYPE_INT, "somaxkva",
                   2057:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2058:                                    "used for socket buffers"),
1.94      yamt     2059:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2060:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
                   2061: }

CVSweb <webmaster@jp.NetBSD.org>