[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 1.204

1.204   ! christos    1: /*     $NetBSD: uipc_socket.c,v 1.203 2011/02/01 01:39:20 matt Exp $   */
1.64      thorpej     2:
                      3: /*-
1.188     ad          4:  * Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc.
1.64      thorpej     5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
1.188     ad          8:  * by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran.
1.64      thorpej     9:  *
                     10:  * Redistribution and use in source and binary forms, with or without
                     11:  * modification, are permitted provided that the following conditions
                     12:  * are met:
                     13:  * 1. Redistributions of source code must retain the above copyright
                     14:  *    notice, this list of conditions and the following disclaimer.
                     15:  * 2. Redistributions in binary form must reproduce the above copyright
                     16:  *    notice, this list of conditions and the following disclaimer in the
                     17:  *    documentation and/or other materials provided with the distribution.
                     18:  *
                     19:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     20:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     21:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     22:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     23:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     24:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     25:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     26:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     27:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     28:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     29:  * POSSIBILITY OF SUCH DAMAGE.
                     30:  */
1.16      cgd        31:
1.1       cgd        32: /*
1.159     ad         33:  * Copyright (c) 2004 The FreeBSD Foundation
                     34:  * Copyright (c) 2004 Robert Watson
1.15      mycroft    35:  * Copyright (c) 1982, 1986, 1988, 1990, 1993
                     36:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        37:  *
                     38:  * Redistribution and use in source and binary forms, with or without
                     39:  * modification, are permitted provided that the following conditions
                     40:  * are met:
                     41:  * 1. Redistributions of source code must retain the above copyright
                     42:  *    notice, this list of conditions and the following disclaimer.
                     43:  * 2. Redistributions in binary form must reproduce the above copyright
                     44:  *    notice, this list of conditions and the following disclaimer in the
                     45:  *    documentation and/or other materials provided with the distribution.
1.85      agc        46:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        47:  *    may be used to endorse or promote products derived from this software
                     48:  *    without specific prior written permission.
                     49:  *
                     50:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     51:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     52:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     53:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     54:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     55:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     56:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     57:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     58:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     59:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     60:  * SUCH DAMAGE.
                     61:  *
1.32      fvdl       62:  *     @(#)uipc_socket.c       8.6 (Berkeley) 5/2/95
1.1       cgd        63:  */
1.59      lukem      64:
                     65: #include <sys/cdefs.h>
1.204   ! christos   66: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.203 2011/02/01 01:39:20 matt Exp $");
1.64      thorpej    67:
1.179     christos   68: #include "opt_compat_netbsd.h"
1.64      thorpej    69: #include "opt_sock_counters.h"
                     70: #include "opt_sosend_loan.h"
1.81      martin     71: #include "opt_mbuftrace.h"
1.84      ragge      72: #include "opt_somaxkva.h"
1.167     ad         73: #include "opt_multiprocessor.h"        /* XXX */
1.1       cgd        74:
1.9       mycroft    75: #include <sys/param.h>
                     76: #include <sys/systm.h>
                     77: #include <sys/proc.h>
                     78: #include <sys/file.h>
1.142     dyoung     79: #include <sys/filedesc.h>
1.173     plunky     80: #include <sys/kmem.h>
1.9       mycroft    81: #include <sys/mbuf.h>
                     82: #include <sys/domain.h>
                     83: #include <sys/kernel.h>
                     84: #include <sys/protosw.h>
                     85: #include <sys/socket.h>
                     86: #include <sys/socketvar.h>
1.21      christos   87: #include <sys/signalvar.h>
1.9       mycroft    88: #include <sys/resourcevar.h>
1.174     pooka      89: #include <sys/uidinfo.h>
1.72      jdolecek   90: #include <sys/event.h>
1.89      christos   91: #include <sys/poll.h>
1.118     elad       92: #include <sys/kauth.h>
1.136     ad         93: #include <sys/mutex.h>
                     94: #include <sys/condvar.h>
1.37      thorpej    95:
1.179     christos   96: #ifdef COMPAT_50
                     97: #include <compat/sys/time.h>
1.184     christos   98: #include <compat/sys/socket.h>
1.179     christos   99: #endif
                    100:
1.202     uebayasi  101: #include <uvm/uvm_extern.h>
                    102: #include <uvm/uvm_loan.h>
                    103: #include <uvm/uvm_page.h>
1.64      thorpej   104:
1.77      thorpej   105: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
                    106: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37      thorpej   107:
1.142     dyoung    108: extern const struct fileops socketops;
                    109:
1.54      lukem     110: extern int     somaxconn;                      /* patchable (XXX sysctl) */
                    111: int            somaxconn = SOMAXCONN;
1.160     ad        112: kmutex_t       *softnet_lock;
1.49      jonathan  113:
1.64      thorpej   114: #ifdef SOSEND_COUNTERS
                    115: #include <sys/device.h>
                    116:
1.113     thorpej   117: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   118:     NULL, "sosend", "loan big");
1.113     thorpej   119: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   120:     NULL, "sosend", "copy big");
1.113     thorpej   121: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   122:     NULL, "sosend", "copy small");
1.113     thorpej   123: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   124:     NULL, "sosend", "kva limit");
                    125:
                    126: #define        SOSEND_COUNTER_INCR(ev)         (ev)->ev_count++
                    127:
1.101     matt      128: EVCNT_ATTACH_STATIC(sosend_loan_big);
                    129: EVCNT_ATTACH_STATIC(sosend_copy_big);
                    130: EVCNT_ATTACH_STATIC(sosend_copy_small);
                    131: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64      thorpej   132: #else
                    133:
                    134: #define        SOSEND_COUNTER_INCR(ev)         /* nothing */
                    135:
                    136: #endif /* SOSEND_COUNTERS */
                    137:
1.119     yamt      138: static struct callback_entry sokva_reclaimerentry;
1.1       cgd       139:
1.167     ad        140: #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR)
1.121     yamt      141: int sock_loan_thresh = -1;
1.71      thorpej   142: #else
1.121     yamt      143: int sock_loan_thresh = 4096;
1.65      thorpej   144: #endif
1.64      thorpej   145:
1.136     ad        146: static kmutex_t so_pendfree_lock;
1.113     thorpej   147: static struct mbuf *so_pendfree;
1.64      thorpej   148:
1.84      ragge     149: #ifndef SOMAXKVA
                    150: #define        SOMAXKVA (16 * 1024 * 1024)
                    151: #endif
                    152: int somaxkva = SOMAXKVA;
1.113     thorpej   153: static int socurkva;
1.136     ad        154: static kcondvar_t socurkva_cv;
1.64      thorpej   155:
1.191     elad      156: static kauth_listener_t socket_listener;
                    157:
1.64      thorpej   158: #define        SOCK_LOAN_CHUNK         65536
                    159:
1.117     yamt      160: static size_t sodopendfree(void);
                    161: static size_t sodopendfreel(void);
1.93      yamt      162:
1.178     pooka     163: static void sysctl_kern_somaxkva_setup(void);
                    164: static struct sysctllog *socket_sysctllog;
                    165:
1.113     thorpej   166: static vsize_t
1.129     yamt      167: sokvareserve(struct socket *so, vsize_t len)
1.80      yamt      168: {
1.98      christos  169:        int error;
1.80      yamt      170:
1.136     ad        171:        mutex_enter(&so_pendfree_lock);
1.80      yamt      172:        while (socurkva + len > somaxkva) {
1.93      yamt      173:                size_t freed;
                    174:
                    175:                /*
                    176:                 * try to do pendfree.
                    177:                 */
                    178:
1.117     yamt      179:                freed = sodopendfreel();
1.93      yamt      180:
                    181:                /*
                    182:                 * if some kva was freed, try again.
                    183:                 */
                    184:
                    185:                if (freed)
1.80      yamt      186:                        continue;
1.93      yamt      187:
1.80      yamt      188:                SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136     ad        189:                error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98      christos  190:                if (error) {
                    191:                        len = 0;
                    192:                        break;
                    193:                }
1.80      yamt      194:        }
1.93      yamt      195:        socurkva += len;
1.136     ad        196:        mutex_exit(&so_pendfree_lock);
1.98      christos  197:        return len;
1.95      yamt      198: }
                    199:
1.113     thorpej   200: static void
1.95      yamt      201: sokvaunreserve(vsize_t len)
                    202: {
                    203:
1.136     ad        204:        mutex_enter(&so_pendfree_lock);
1.95      yamt      205:        socurkva -= len;
1.136     ad        206:        cv_broadcast(&socurkva_cv);
                    207:        mutex_exit(&so_pendfree_lock);
1.95      yamt      208: }
                    209:
                    210: /*
                    211:  * sokvaalloc: allocate kva for loan.
                    212:  */
                    213:
                    214: vaddr_t
                    215: sokvaalloc(vsize_t len, struct socket *so)
                    216: {
                    217:        vaddr_t lva;
                    218:
                    219:        /*
                    220:         * reserve kva.
                    221:         */
                    222:
1.98      christos  223:        if (sokvareserve(so, len) == 0)
                    224:                return 0;
1.93      yamt      225:
                    226:        /*
                    227:         * allocate kva.
                    228:         */
1.80      yamt      229:
1.109     yamt      230:        lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95      yamt      231:        if (lva == 0) {
                    232:                sokvaunreserve(len);
1.80      yamt      233:                return (0);
1.95      yamt      234:        }
1.80      yamt      235:
                    236:        return lva;
                    237: }
                    238:
1.93      yamt      239: /*
                    240:  * sokvafree: free kva for loan.
                    241:  */
                    242:
1.80      yamt      243: void
                    244: sokvafree(vaddr_t sva, vsize_t len)
                    245: {
1.93      yamt      246:
                    247:        /*
                    248:         * free kva.
                    249:         */
1.80      yamt      250:
1.109     yamt      251:        uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93      yamt      252:
                    253:        /*
                    254:         * unreserve kva.
                    255:         */
                    256:
1.95      yamt      257:        sokvaunreserve(len);
1.80      yamt      258: }
                    259:
1.64      thorpej   260: static void
1.134     christos  261: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64      thorpej   262: {
1.156     yamt      263:        vaddr_t sva, eva;
1.64      thorpej   264:        vsize_t len;
1.156     yamt      265:        int npgs;
                    266:
                    267:        KASSERT(pgs != NULL);
1.64      thorpej   268:
                    269:        eva = round_page((vaddr_t) buf + size);
                    270:        sva = trunc_page((vaddr_t) buf);
                    271:        len = eva - sva;
                    272:        npgs = len >> PAGE_SHIFT;
                    273:
                    274:        pmap_kremove(sva, len);
                    275:        pmap_update(pmap_kernel());
                    276:        uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80      yamt      277:        sokvafree(sva, len);
1.64      thorpej   278: }
                    279:
                    280: static size_t
1.152     matt      281: sodopendfree(void)
1.64      thorpej   282: {
1.93      yamt      283:        size_t rv;
1.64      thorpej   284:
1.160     ad        285:        if (__predict_true(so_pendfree == NULL))
                    286:                return 0;
                    287:
1.136     ad        288:        mutex_enter(&so_pendfree_lock);
1.117     yamt      289:        rv = sodopendfreel();
1.136     ad        290:        mutex_exit(&so_pendfree_lock);
1.93      yamt      291:
                    292:        return rv;
                    293: }
                    294:
                    295: /*
                    296:  * sodopendfreel: free mbufs on "pendfree" list.
1.136     ad        297:  * unlock and relock so_pendfree_lock when freeing mbufs.
1.93      yamt      298:  *
1.136     ad        299:  * => called with so_pendfree_lock held.
1.93      yamt      300:  */
                    301:
                    302: static size_t
1.152     matt      303: sodopendfreel(void)
1.93      yamt      304: {
1.137     ad        305:        struct mbuf *m, *next;
1.93      yamt      306:        size_t rv = 0;
                    307:
1.136     ad        308:        KASSERT(mutex_owned(&so_pendfree_lock));
1.64      thorpej   309:
1.137     ad        310:        while (so_pendfree != NULL) {
1.64      thorpej   311:                m = so_pendfree;
1.93      yamt      312:                so_pendfree = NULL;
1.136     ad        313:                mutex_exit(&so_pendfree_lock);
1.93      yamt      314:
                    315:                for (; m != NULL; m = next) {
                    316:                        next = m->m_next;
1.156     yamt      317:                        KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
                    318:                        KASSERT(m->m_ext.ext_refcnt == 0);
1.93      yamt      319:
                    320:                        rv += m->m_ext.ext_size;
1.156     yamt      321:                        sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
1.93      yamt      322:                            m->m_ext.ext_size);
1.145     ad        323:                        pool_cache_put(mb_cache, m);
1.93      yamt      324:                }
1.64      thorpej   325:
1.136     ad        326:                mutex_enter(&so_pendfree_lock);
1.64      thorpej   327:        }
                    328:
                    329:        return (rv);
                    330: }
                    331:
1.80      yamt      332: void
1.134     christos  333: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64      thorpej   334: {
                    335:
1.156     yamt      336:        KASSERT(m != NULL);
1.64      thorpej   337:
1.93      yamt      338:        /*
                    339:         * postpone freeing mbuf.
                    340:         *
                    341:         * we can't do it in interrupt context
                    342:         * because we need to put kva back to kernel_map.
                    343:         */
                    344:
1.136     ad        345:        mutex_enter(&so_pendfree_lock);
1.92      yamt      346:        m->m_next = so_pendfree;
                    347:        so_pendfree = m;
1.136     ad        348:        cv_broadcast(&socurkva_cv);
                    349:        mutex_exit(&so_pendfree_lock);
1.64      thorpej   350: }
                    351:
                    352: static long
                    353: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
                    354: {
                    355:        struct iovec *iov = uio->uio_iov;
                    356:        vaddr_t sva, eva;
                    357:        vsize_t len;
1.156     yamt      358:        vaddr_t lva;
                    359:        int npgs, error;
                    360:        vaddr_t va;
                    361:        int i;
1.64      thorpej   362:
1.116     yamt      363:        if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64      thorpej   364:                return (0);
                    365:
                    366:        if (iov->iov_len < (size_t) space)
                    367:                space = iov->iov_len;
                    368:        if (space > SOCK_LOAN_CHUNK)
                    369:                space = SOCK_LOAN_CHUNK;
                    370:
                    371:        eva = round_page((vaddr_t) iov->iov_base + space);
                    372:        sva = trunc_page((vaddr_t) iov->iov_base);
                    373:        len = eva - sva;
                    374:        npgs = len >> PAGE_SHIFT;
                    375:
1.79      thorpej   376:        KASSERT(npgs <= M_EXT_MAXPAGES);
                    377:
1.80      yamt      378:        lva = sokvaalloc(len, so);
1.64      thorpej   379:        if (lva == 0)
1.80      yamt      380:                return 0;
1.64      thorpej   381:
1.116     yamt      382:        error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79      thorpej   383:            m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64      thorpej   384:        if (error) {
1.80      yamt      385:                sokvafree(lva, len);
1.64      thorpej   386:                return (0);
                    387:        }
                    388:
                    389:        for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79      thorpej   390:                pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
1.194     cegger    391:                    VM_PROT_READ, 0);
1.64      thorpej   392:        pmap_update(pmap_kernel());
                    393:
                    394:        lva += (vaddr_t) iov->iov_base & PAGE_MASK;
                    395:
1.134     christos  396:        MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79      thorpej   397:        m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64      thorpej   398:
                    399:        uio->uio_resid -= space;
                    400:        /* uio_offset not updated, not set/used for write(2) */
1.134     christos  401:        uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64      thorpej   402:        uio->uio_iov->iov_len -= space;
                    403:        if (uio->uio_iov->iov_len == 0) {
                    404:                uio->uio_iov++;
                    405:                uio->uio_iovcnt--;
                    406:        }
                    407:
                    408:        return (space);
                    409: }
                    410:
1.119     yamt      411: static int
1.129     yamt      412: sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg)
1.119     yamt      413: {
                    414:
                    415:        KASSERT(ce == &sokva_reclaimerentry);
                    416:        KASSERT(obj == NULL);
                    417:
                    418:        sodopendfree();
                    419:        if (!vm_map_starved_p(kernel_map)) {
                    420:                return CALLBACK_CHAIN_ABORT;
                    421:        }
                    422:        return CALLBACK_CHAIN_CONTINUE;
                    423: }
                    424:
1.142     dyoung    425: struct mbuf *
1.147     dyoung    426: getsombuf(struct socket *so, int type)
1.142     dyoung    427: {
                    428:        struct mbuf *m;
                    429:
1.147     dyoung    430:        m = m_get(M_WAIT, type);
1.142     dyoung    431:        MCLAIM(m, so->so_mowner);
                    432:        return m;
                    433: }
                    434:
1.191     elad      435: static int
                    436: socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
                    437:     void *arg0, void *arg1, void *arg2, void *arg3)
                    438: {
                    439:        int result;
                    440:        enum kauth_network_req req;
                    441:
                    442:        result = KAUTH_RESULT_DEFER;
                    443:        req = (enum kauth_network_req)arg0;
                    444:
1.193     elad      445:        if ((action != KAUTH_NETWORK_SOCKET) &&
                    446:            (action != KAUTH_NETWORK_BIND))
1.191     elad      447:                return result;
                    448:
                    449:        switch (req) {
1.193     elad      450:        case KAUTH_REQ_NETWORK_BIND_PORT:
                    451:                result = KAUTH_RESULT_ALLOW;
                    452:                break;
                    453:
1.191     elad      454:        case KAUTH_REQ_NETWORK_SOCKET_DROP: {
                    455:                /* Normal users can only drop their own connections. */
                    456:                struct socket *so = (struct socket *)arg1;
                    457:
1.199     elad      458:                if (proc_uidmatch(cred, so->so_cred))
1.191     elad      459:                        result = KAUTH_RESULT_ALLOW;
                    460:
                    461:                break;
                    462:                }
                    463:
                    464:        case KAUTH_REQ_NETWORK_SOCKET_OPEN:
                    465:                /* We allow "raw" routing/bluetooth sockets to anyone. */
1.203     matt      466:                if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_OROUTE
                    467:                    || (u_long)arg1 == PF_BLUETOOTH) {
1.191     elad      468:                        result = KAUTH_RESULT_ALLOW;
1.203     matt      469:                } else {
1.191     elad      470:                        /* Privileged, let secmodel handle this. */
                    471:                        if ((u_long)arg2 == SOCK_RAW)
                    472:                                break;
                    473:                }
                    474:
                    475:                result = KAUTH_RESULT_ALLOW;
                    476:
                    477:                break;
                    478:
1.192     elad      479:        case KAUTH_REQ_NETWORK_SOCKET_CANSEE:
                    480:                result = KAUTH_RESULT_ALLOW;
                    481:
                    482:                break;
                    483:
1.191     elad      484:        default:
                    485:                break;
                    486:        }
                    487:
                    488:        return result;
                    489: }
                    490:
1.119     yamt      491: void
                    492: soinit(void)
                    493: {
                    494:
1.178     pooka     495:        sysctl_kern_somaxkva_setup();
                    496:
1.148     ad        497:        mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.160     ad        498:        softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.136     ad        499:        cv_init(&socurkva_cv, "sokva");
1.166     ad        500:        soinit2();
1.136     ad        501:
1.119     yamt      502:        /* Set the initial adjusted socket buffer size. */
                    503:        if (sb_max_set(sb_max))
                    504:                panic("bad initial sb_max value: %lu", sb_max);
                    505:
                    506:        callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback,
                    507:            &sokva_reclaimerentry, NULL, sokva_reclaim_callback);
1.191     elad      508:
                    509:        socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
                    510:            socket_listener_cb, NULL);
1.119     yamt      511: }
                    512:
1.1       cgd       513: /*
                    514:  * Socket operation routines.
                    515:  * These routines are called by the routines in
                    516:  * sys_socket.c or from a system process, and
                    517:  * implement the semantics of socket operations by
                    518:  * switching out to the protocol specific routines.
                    519:  */
                    520: /*ARGSUSED*/
1.3       andrew    521: int
1.160     ad        522: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l,
                    523:         struct socket *lockso)
1.1       cgd       524: {
1.99      matt      525:        const struct protosw    *prp;
1.54      lukem     526:        struct socket   *so;
1.115     yamt      527:        uid_t           uid;
1.160     ad        528:        int             error;
                    529:        kmutex_t        *lock;
1.1       cgd       530:
1.132     elad      531:        error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
                    532:            KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
                    533:            KAUTH_ARG(proto));
1.140     dyoung    534:        if (error != 0)
                    535:                return error;
1.127     elad      536:
1.1       cgd       537:        if (proto)
                    538:                prp = pffindproto(dom, proto, type);
                    539:        else
                    540:                prp = pffindtype(dom, type);
1.140     dyoung    541:        if (prp == NULL) {
1.120     ginsbach  542:                /* no support for domain */
                    543:                if (pffinddomain(dom) == 0)
1.140     dyoung    544:                        return EAFNOSUPPORT;
1.120     ginsbach  545:                /* no support for socket type */
                    546:                if (proto == 0 && type != 0)
1.140     dyoung    547:                        return EPROTOTYPE;
                    548:                return EPROTONOSUPPORT;
1.120     ginsbach  549:        }
1.140     dyoung    550:        if (prp->pr_usrreq == NULL)
                    551:                return EPROTONOSUPPORT;
1.1       cgd       552:        if (prp->pr_type != type)
1.140     dyoung    553:                return EPROTOTYPE;
1.160     ad        554:
                    555:        so = soget(true);
1.1       cgd       556:        so->so_type = type;
                    557:        so->so_proto = prp;
1.33      matt      558:        so->so_send = sosend;
                    559:        so->so_receive = soreceive;
1.78      matt      560: #ifdef MBUFTRACE
                    561:        so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
                    562:        so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
                    563:        so->so_mowner = &prp->pr_domain->dom_mowner;
                    564: #endif
1.138     rmind     565:        uid = kauth_cred_geteuid(l->l_cred);
1.115     yamt      566:        so->so_uidinfo = uid_find(uid);
1.168     yamt      567:        so->so_cpid = l->l_proc->p_pid;
1.160     ad        568:        if (lockso != NULL) {
                    569:                /* Caller wants us to share a lock. */
                    570:                lock = lockso->so_lock;
                    571:                so->so_lock = lock;
                    572:                mutex_obj_hold(lock);
                    573:                mutex_enter(lock);
                    574:        } else {
                    575:                /* Lock assigned and taken during PRU_ATTACH. */
                    576:        }
1.140     dyoung    577:        error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
                    578:            (struct mbuf *)(long)proto, NULL, l);
1.160     ad        579:        KASSERT(solocked(so));
1.140     dyoung    580:        if (error != 0) {
1.1       cgd       581:                so->so_state |= SS_NOFDREF;
                    582:                sofree(so);
1.140     dyoung    583:                return error;
1.1       cgd       584:        }
1.198     elad      585:        so->so_cred = kauth_cred_dup(l->l_cred);
1.160     ad        586:        sounlock(so);
1.1       cgd       587:        *aso = so;
1.140     dyoung    588:        return 0;
1.1       cgd       589: }
                    590:
1.142     dyoung    591: /* On success, write file descriptor to fdout and return zero.  On
                    592:  * failure, return non-zero; *fdout will be undefined.
                    593:  */
                    594: int
                    595: fsocreate(int domain, struct socket **sop, int type, int protocol,
                    596:     struct lwp *l, int *fdout)
                    597: {
                    598:        struct socket   *so;
                    599:        struct file     *fp;
                    600:        int             fd, error;
1.204   ! christos  601:        int             flags = type & SOCK_FLAGS_MASK;
1.142     dyoung    602:
1.204   ! christos  603:        type &= ~SOCK_FLAGS_MASK;
1.155     ad        604:        if ((error = fd_allocfile(&fp, &fd)) != 0)
1.204   ! christos  605:                return error;
        !           606:        fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0);
        !           607:        fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0);
1.142     dyoung    608:        fp->f_type = DTYPE_SOCKET;
                    609:        fp->f_ops = &socketops;
1.160     ad        610:        error = socreate(domain, &so, type, protocol, l, NULL);
1.142     dyoung    611:        if (error != 0) {
1.155     ad        612:                fd_abort(curproc, fp, fd);
1.142     dyoung    613:        } else {
                    614:                if (sop != NULL)
                    615:                        *sop = so;
                    616:                fp->f_data = so;
1.155     ad        617:                fd_affix(curproc, fp, fd);
1.142     dyoung    618:                *fdout = fd;
                    619:        }
                    620:        return error;
                    621: }
                    622:
1.3       andrew    623: int
1.190     dyoung    624: sofamily(const struct socket *so)
                    625: {
                    626:        const struct protosw *pr;
                    627:        const struct domain *dom;
                    628:
                    629:        if ((pr = so->so_proto) == NULL)
                    630:                return AF_UNSPEC;
                    631:        if ((dom = pr->pr_domain) == NULL)
                    632:                return AF_UNSPEC;
                    633:        return dom->dom_family;
                    634: }
                    635:
                    636: int
1.114     christos  637: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       638: {
1.160     ad        639:        int     error;
1.1       cgd       640:
1.160     ad        641:        solock(so);
1.140     dyoung    642:        error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.160     ad        643:        sounlock(so);
1.140     dyoung    644:        return error;
1.1       cgd       645: }
                    646:
1.3       andrew    647: int
1.150     elad      648: solisten(struct socket *so, int backlog, struct lwp *l)
1.1       cgd       649: {
1.160     ad        650:        int     error;
1.1       cgd       651:
1.160     ad        652:        solock(so);
1.158     ad        653:        if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
1.163     ad        654:            SS_ISDISCONNECTING)) != 0) {
                    655:                sounlock(so);
1.158     ad        656:                return (EOPNOTSUPP);
1.163     ad        657:        }
1.140     dyoung    658:        error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
1.150     elad      659:            NULL, NULL, l);
1.140     dyoung    660:        if (error != 0) {
1.160     ad        661:                sounlock(so);
1.140     dyoung    662:                return error;
1.1       cgd       663:        }
1.63      matt      664:        if (TAILQ_EMPTY(&so->so_q))
1.1       cgd       665:                so->so_options |= SO_ACCEPTCONN;
                    666:        if (backlog < 0)
                    667:                backlog = 0;
1.49      jonathan  668:        so->so_qlimit = min(backlog, somaxconn);
1.160     ad        669:        sounlock(so);
1.140     dyoung    670:        return 0;
1.1       cgd       671: }
                    672:
1.21      christos  673: void
1.54      lukem     674: sofree(struct socket *so)
1.1       cgd       675: {
1.161     ad        676:        u_int refs;
1.1       cgd       677:
1.160     ad        678:        KASSERT(solocked(so));
                    679:
                    680:        if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) {
                    681:                sounlock(so);
1.1       cgd       682:                return;
1.160     ad        683:        }
1.43      mycroft   684:        if (so->so_head) {
                    685:                /*
                    686:                 * We must not decommission a socket that's on the accept(2)
                    687:                 * queue.  If we do, then accept(2) may hang after select(2)
                    688:                 * indicated that the listening socket was ready.
                    689:                 */
1.160     ad        690:                if (!soqremque(so, 0)) {
                    691:                        sounlock(so);
1.43      mycroft   692:                        return;
1.160     ad        693:                }
1.43      mycroft   694:        }
1.98      christos  695:        if (so->so_rcv.sb_hiwat)
1.110     christos  696:                (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98      christos  697:                    RLIM_INFINITY);
                    698:        if (so->so_snd.sb_hiwat)
1.110     christos  699:                (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98      christos  700:                    RLIM_INFINITY);
                    701:        sbrelease(&so->so_snd, so);
1.160     ad        702:        KASSERT(!cv_has_waiters(&so->so_cv));
                    703:        KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv));
                    704:        KASSERT(!cv_has_waiters(&so->so_snd.sb_cv));
1.1       cgd       705:        sorflush(so);
1.161     ad        706:        refs = so->so_aborting; /* XXX */
1.177     ad        707:        /* Remove acccept filter if one is present. */
1.170     tls       708:        if (so->so_accf != NULL)
1.177     ad        709:                (void)accept_filt_clear(so);
1.160     ad        710:        sounlock(so);
1.161     ad        711:        if (refs == 0)          /* XXX */
                    712:                soput(so);
1.1       cgd       713: }
                    714:
                    715: /*
                    716:  * Close a socket on last file table reference removal.
                    717:  * Initiate disconnect if connected.
                    718:  * Free socket when disconnect complete.
                    719:  */
1.3       andrew    720: int
1.54      lukem     721: soclose(struct socket *so)
1.1       cgd       722: {
1.54      lukem     723:        struct socket   *so2;
1.160     ad        724:        int             error;
                    725:        int             error2;
1.1       cgd       726:
1.54      lukem     727:        error = 0;
1.160     ad        728:        solock(so);
1.1       cgd       729:        if (so->so_options & SO_ACCEPTCONN) {
1.172     ad        730:                for (;;) {
                    731:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.160     ad        732:                                KASSERT(solocked2(so, so2));
                    733:                                (void) soqremque(so2, 0);
                    734:                                /* soabort drops the lock. */
                    735:                                (void) soabort(so2);
                    736:                                solock(so);
1.172     ad        737:                                continue;
1.160     ad        738:                        }
1.172     ad        739:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.160     ad        740:                                KASSERT(solocked2(so, so2));
                    741:                                (void) soqremque(so2, 1);
                    742:                                /* soabort drops the lock. */
                    743:                                (void) soabort(so2);
                    744:                                solock(so);
1.172     ad        745:                                continue;
1.160     ad        746:                        }
1.172     ad        747:                        break;
                    748:                }
1.1       cgd       749:        }
                    750:        if (so->so_pcb == 0)
                    751:                goto discard;
                    752:        if (so->so_state & SS_ISCONNECTED) {
                    753:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    754:                        error = sodisconnect(so);
                    755:                        if (error)
                    756:                                goto drop;
                    757:                }
                    758:                if (so->so_options & SO_LINGER) {
1.151     ad        759:                        if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio)
1.1       cgd       760:                                goto drop;
1.21      christos  761:                        while (so->so_state & SS_ISCONNECTED) {
1.185     yamt      762:                                error = sowait(so, true, so->so_linger * hz);
1.21      christos  763:                                if (error)
1.1       cgd       764:                                        break;
1.21      christos  765:                        }
1.1       cgd       766:                }
                    767:        }
1.54      lukem     768:  drop:
1.1       cgd       769:        if (so->so_pcb) {
1.160     ad        770:                error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140     dyoung    771:                    NULL, NULL, NULL, NULL);
1.1       cgd       772:                if (error == 0)
                    773:                        error = error2;
                    774:        }
1.54      lukem     775:  discard:
1.1       cgd       776:        if (so->so_state & SS_NOFDREF)
                    777:                panic("soclose: NOFDREF");
1.198     elad      778:        kauth_cred_free(so->so_cred);
1.1       cgd       779:        so->so_state |= SS_NOFDREF;
                    780:        sofree(so);
                    781:        return (error);
                    782: }
                    783:
                    784: /*
1.160     ad        785:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       786:  */
1.3       andrew    787: int
1.54      lukem     788: soabort(struct socket *so)
1.1       cgd       789: {
1.161     ad        790:        u_int refs;
1.139     yamt      791:        int error;
1.160     ad        792:
                    793:        KASSERT(solocked(so));
                    794:        KASSERT(so->so_head == NULL);
1.1       cgd       795:
1.161     ad        796:        so->so_aborting++;              /* XXX */
1.140     dyoung    797:        error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
                    798:            NULL, NULL, NULL);
1.161     ad        799:        refs = --so->so_aborting;       /* XXX */
1.164     drochner  800:        if (error || (refs == 0)) {
1.139     yamt      801:                sofree(so);
1.160     ad        802:        } else {
                    803:                sounlock(so);
1.139     yamt      804:        }
                    805:        return error;
1.1       cgd       806: }
                    807:
1.3       andrew    808: int
1.54      lukem     809: soaccept(struct socket *so, struct mbuf *nam)
1.1       cgd       810: {
1.160     ad        811:        int     error;
                    812:
                    813:        KASSERT(solocked(so));
1.1       cgd       814:
1.54      lukem     815:        error = 0;
1.1       cgd       816:        if ((so->so_state & SS_NOFDREF) == 0)
                    817:                panic("soaccept: !NOFDREF");
                    818:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   819:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    820:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41      mycroft   821:                error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140     dyoung    822:                    NULL, nam, NULL, NULL);
1.41      mycroft   823:        else
1.53      itojun    824:                error = ECONNABORTED;
1.52      itojun    825:
1.1       cgd       826:        return (error);
                    827: }
                    828:
1.3       andrew    829: int
1.114     christos  830: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       831: {
1.160     ad        832:        int             error;
                    833:
                    834:        KASSERT(solocked(so));
1.1       cgd       835:
                    836:        if (so->so_options & SO_ACCEPTCONN)
                    837:                return (EOPNOTSUPP);
                    838:        /*
                    839:         * If protocol is connection-based, can only connect once.
                    840:         * Otherwise, if connected, try to disconnect first.
                    841:         * This allows user to disconnect by connecting to, e.g.,
                    842:         * a null address.
                    843:         */
                    844:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    845:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
                    846:            (error = sodisconnect(so))))
                    847:                error = EISCONN;
                    848:        else
                    849:                error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140     dyoung    850:                    NULL, nam, NULL, l);
1.1       cgd       851:        return (error);
                    852: }
                    853:
1.3       andrew    854: int
1.54      lukem     855: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       856: {
1.160     ad        857:        int     error;
                    858:
                    859:        KASSERT(solocked2(so1, so2));
1.1       cgd       860:
1.22      mycroft   861:        error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140     dyoung    862:            NULL, (struct mbuf *)so2, NULL, NULL);
1.1       cgd       863:        return (error);
                    864: }
                    865:
1.3       andrew    866: int
1.54      lukem     867: sodisconnect(struct socket *so)
1.1       cgd       868: {
1.160     ad        869:        int     error;
                    870:
                    871:        KASSERT(solocked(so));
1.1       cgd       872:
                    873:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    874:                error = ENOTCONN;
1.160     ad        875:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       876:                error = EALREADY;
1.160     ad        877:        } else {
                    878:                error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
                    879:                    NULL, NULL, NULL, NULL);
1.1       cgd       880:        }
1.117     yamt      881:        sodopendfree();
1.1       cgd       882:        return (error);
                    883: }
                    884:
1.15      mycroft   885: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       886: /*
                    887:  * Send on a socket.
                    888:  * If send must go all at once and message is larger than
                    889:  * send buffering, then hard error.
                    890:  * Lock against other senders.
                    891:  * If must go all at once and not enough room now, then
                    892:  * inform user that this would block and do nothing.
                    893:  * Otherwise, if nonblocking, send as much as possible.
                    894:  * The data to be sent is described by "uio" if nonzero,
                    895:  * otherwise by the mbuf chain "top" (which must be null
                    896:  * if uio is not).  Data provided in mbuf chain must be small
                    897:  * enough to send all at once.
                    898:  *
                    899:  * Returns nonzero on error, timeout or signal; callers
                    900:  * must check for short counts if EINTR/ERESTART are returned.
                    901:  * Data and control buffers are freed on return.
                    902:  */
1.3       andrew    903: int
1.54      lukem     904: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114     christos  905:        struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       906: {
1.54      lukem     907:        struct mbuf     **mp, *m;
1.114     christos  908:        struct proc     *p;
1.58      jdolecek  909:        long            space, len, resid, clen, mlen;
                    910:        int             error, s, dontroute, atomic;
1.196     dsl       911:        short           wakeup_state = 0;
1.54      lukem     912:
1.114     christos  913:        p = l->l_proc;
1.117     yamt      914:        sodopendfree();
1.160     ad        915:        clen = 0;
1.64      thorpej   916:
1.160     ad        917:        /*
                    918:         * solock() provides atomicity of access.  splsoftnet() prevents
                    919:         * protocol processing soft interrupts from interrupting us and
                    920:         * blocking (expensive).
                    921:         */
                    922:        s = splsoftnet();
                    923:        solock(so);
1.54      lukem     924:        atomic = sosendallatonce(so) || top;
1.1       cgd       925:        if (uio)
                    926:                resid = uio->uio_resid;
                    927:        else
                    928:                resid = top->m_pkthdr.len;
1.7       cgd       929:        /*
                    930:         * In theory resid should be unsigned.
                    931:         * However, space must be signed, as it might be less than 0
                    932:         * if we over-committed, and we must use a signed comparison
                    933:         * of space and resid.  On the other hand, a negative resid
                    934:         * causes us to loop sending 0-length segments to the protocol.
                    935:         */
1.29      mycroft   936:        if (resid < 0) {
                    937:                error = EINVAL;
                    938:                goto out;
                    939:        }
1.1       cgd       940:        dontroute =
                    941:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    942:            (so->so_proto->pr_flags & PR_ATOMIC);
1.165     christos  943:        l->l_ru.ru_msgsnd++;
1.1       cgd       944:        if (control)
                    945:                clen = control->m_len;
1.54      lukem     946:  restart:
1.21      christos  947:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       948:                goto out;
                    949:        do {
1.160     ad        950:                if (so->so_state & SS_CANTSENDMORE) {
                    951:                        error = EPIPE;
                    952:                        goto release;
                    953:                }
1.48      thorpej   954:                if (so->so_error) {
                    955:                        error = so->so_error;
                    956:                        so->so_error = 0;
                    957:                        goto release;
                    958:                }
1.1       cgd       959:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    960:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
                    961:                                if ((so->so_state & SS_ISCONFIRMING) == 0 &&
1.160     ad        962:                                    !(resid == 0 && clen != 0)) {
                    963:                                        error = ENOTCONN;
                    964:                                        goto release;
                    965:                                }
                    966:                        } else if (addr == 0) {
                    967:                                error = EDESTADDRREQ;
                    968:                                goto release;
                    969:                        }
1.1       cgd       970:                }
                    971:                space = sbspace(&so->so_snd);
                    972:                if (flags & MSG_OOB)
                    973:                        space += 1024;
1.21      christos  974:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        975:                    clen > so->so_snd.sb_hiwat) {
                    976:                        error = EMSGSIZE;
                    977:                        goto release;
                    978:                }
1.96      mycroft   979:                if (space < resid + clen &&
1.1       cgd       980:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.160     ad        981:                        if (so->so_nbio) {
                    982:                                error = EWOULDBLOCK;
                    983:                                goto release;
                    984:                        }
1.1       cgd       985:                        sbunlock(&so->so_snd);
1.196     dsl       986:                        if (wakeup_state & SS_RESTARTSYS) {
                    987:                                error = ERESTART;
                    988:                                goto out;
                    989:                        }
1.1       cgd       990:                        error = sbwait(&so->so_snd);
                    991:                        if (error)
                    992:                                goto out;
1.196     dsl       993:                        wakeup_state = so->so_state;
1.1       cgd       994:                        goto restart;
                    995:                }
1.196     dsl       996:                wakeup_state = 0;
1.1       cgd       997:                mp = &top;
                    998:                space -= clen;
                    999:                do {
1.45      tv       1000:                        if (uio == NULL) {
                   1001:                                /*
                   1002:                                 * Data is prepackaged in "top".
                   1003:                                 */
                   1004:                                resid = 0;
                   1005:                                if (flags & MSG_EOR)
                   1006:                                        top->m_flags |= M_EOR;
                   1007:                        } else do {
1.160     ad       1008:                                sounlock(so);
                   1009:                                splx(s);
1.144     dyoung   1010:                                if (top == NULL) {
1.78      matt     1011:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv       1012:                                        mlen = MHLEN;
                   1013:                                        m->m_pkthdr.len = 0;
1.140     dyoung   1014:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv       1015:                                } else {
1.78      matt     1016:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv       1017:                                        mlen = MLEN;
                   1018:                                }
1.78      matt     1019:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt     1020:                                if (sock_loan_thresh >= 0 &&
                   1021:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                   1022:                                    space >= sock_loan_thresh &&
1.64      thorpej  1023:                                    (len = sosend_loan(so, uio, m,
                   1024:                                                       space)) != 0) {
                   1025:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                   1026:                                        space -= len;
                   1027:                                        goto have_data;
                   1028:                                }
1.45      tv       1029:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej  1030:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.201     oki      1031:                                        m_clget(m, M_DONTWAIT);
1.45      tv       1032:                                        if ((m->m_flags & M_EXT) == 0)
                   1033:                                                goto nopages;
                   1034:                                        mlen = MCLBYTES;
                   1035:                                        if (atomic && top == 0) {
1.58      jdolecek 1036:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem    1037:                                                    resid);
1.45      tv       1038:                                                m->m_data += max_hdr;
                   1039:                                        } else
1.58      jdolecek 1040:                                                len = lmin(MCLBYTES, resid);
1.45      tv       1041:                                        space -= len;
                   1042:                                } else {
1.64      thorpej  1043:  nopages:
                   1044:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek 1045:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv       1046:                                        space -= len;
                   1047:                                        /*
                   1048:                                         * For datagram protocols, leave room
                   1049:                                         * for protocol headers in first mbuf.
                   1050:                                         */
                   1051:                                        if (atomic && top == 0 && len < mlen)
                   1052:                                                MH_ALIGN(m, len);
                   1053:                                }
1.144     dyoung   1054:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej  1055:  have_data:
1.45      tv       1056:                                resid = uio->uio_resid;
                   1057:                                m->m_len = len;
                   1058:                                *mp = m;
                   1059:                                top->m_pkthdr.len += len;
1.160     ad       1060:                                s = splsoftnet();
                   1061:                                solock(so);
1.144     dyoung   1062:                                if (error != 0)
1.45      tv       1063:                                        goto release;
                   1064:                                mp = &m->m_next;
                   1065:                                if (resid <= 0) {
                   1066:                                        if (flags & MSG_EOR)
                   1067:                                                top->m_flags |= M_EOR;
                   1068:                                        break;
                   1069:                                }
                   1070:                        } while (space > 0 && atomic);
1.108     perry    1071:
1.160     ad       1072:                        if (so->so_state & SS_CANTSENDMORE) {
                   1073:                                error = EPIPE;
                   1074:                                goto release;
                   1075:                        }
1.45      tv       1076:                        if (dontroute)
                   1077:                                so->so_options |= SO_DONTROUTE;
                   1078:                        if (resid > 0)
                   1079:                                so->so_state |= SS_MORETOCOME;
1.46      sommerfe 1080:                        error = (*so->so_proto->pr_usrreq)(so,
                   1081:                            (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.160     ad       1082:                            top, addr, control, curlwp);
1.45      tv       1083:                        if (dontroute)
                   1084:                                so->so_options &= ~SO_DONTROUTE;
                   1085:                        if (resid > 0)
                   1086:                                so->so_state &= ~SS_MORETOCOME;
                   1087:                        clen = 0;
1.144     dyoung   1088:                        control = NULL;
                   1089:                        top = NULL;
1.45      tv       1090:                        mp = &top;
1.144     dyoung   1091:                        if (error != 0)
1.1       cgd      1092:                                goto release;
                   1093:                } while (resid && space > 0);
                   1094:        } while (resid);
                   1095:
1.54      lukem    1096:  release:
1.1       cgd      1097:        sbunlock(&so->so_snd);
1.54      lukem    1098:  out:
1.160     ad       1099:        sounlock(so);
                   1100:        splx(s);
1.1       cgd      1101:        if (top)
                   1102:                m_freem(top);
                   1103:        if (control)
                   1104:                m_freem(control);
                   1105:        return (error);
                   1106: }
                   1107:
                   1108: /*
1.159     ad       1109:  * Following replacement or removal of the first mbuf on the first
                   1110:  * mbuf chain of a socket buffer, push necessary state changes back
                   1111:  * into the socket buffer so that other consumers see the values
                   1112:  * consistently.  'nextrecord' is the callers locally stored value of
                   1113:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1114:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1115:  */
                   1116: static void
                   1117: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1118: {
                   1119:
1.160     ad       1120:        KASSERT(solocked(sb->sb_so));
                   1121:
1.159     ad       1122:        /*
                   1123:         * First, update for the new value of nextrecord.  If necessary,
                   1124:         * make it the first record.
                   1125:         */
                   1126:        if (sb->sb_mb != NULL)
                   1127:                sb->sb_mb->m_nextpkt = nextrecord;
                   1128:        else
                   1129:                sb->sb_mb = nextrecord;
                   1130:
                   1131:         /*
                   1132:          * Now update any dependent socket buffer fields to reflect
                   1133:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1134:          * the addition of a second clause that takes care of the
                   1135:          * case where sb_mb has been updated, but remains the last
                   1136:          * record.
                   1137:          */
                   1138:         if (sb->sb_mb == NULL) {
                   1139:                 sb->sb_mbtail = NULL;
                   1140:                 sb->sb_lastrecord = NULL;
                   1141:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1142:                 sb->sb_lastrecord = sb->sb_mb;
                   1143: }
                   1144:
                   1145: /*
1.1       cgd      1146:  * Implement receive operations on a socket.
                   1147:  * We depend on the way that records are added to the sockbuf
                   1148:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1149:  * must begin with an address if the protocol so specifies,
                   1150:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1151:  * and then zero or more mbufs of data.
                   1152:  * In order to avoid blocking network interrupts for the entire time here,
                   1153:  * we splx() while doing the actual copy to user space.
                   1154:  * Although the sockbuf is locked, new data may still be appended,
                   1155:  * and thus we must maintain consistency of the sockbuf during that time.
                   1156:  *
                   1157:  * The caller may receive the data as a single mbuf chain by supplying
                   1158:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1159:  * only for the count in uio_resid.
                   1160:  */
1.3       andrew   1161: int
1.54      lukem    1162: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1163:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1164: {
1.116     yamt     1165:        struct lwp *l = curlwp;
1.160     ad       1166:        struct mbuf     *m, **mp, *mt;
1.146     dyoung   1167:        int atomic, flags, len, error, s, offset, moff, type, orig_resid;
1.99      matt     1168:        const struct protosw    *pr;
1.54      lukem    1169:        struct mbuf     *nextrecord;
1.67      he       1170:        int             mbuf_removed = 0;
1.146     dyoung   1171:        const struct domain *dom;
1.196     dsl      1172:        short           wakeup_state = 0;
1.64      thorpej  1173:
1.54      lukem    1174:        pr = so->so_proto;
1.146     dyoung   1175:        atomic = pr->pr_flags & PR_ATOMIC;
                   1176:        dom = pr->pr_domain;
1.1       cgd      1177:        mp = mp0;
1.54      lukem    1178:        type = 0;
                   1179:        orig_resid = uio->uio_resid;
1.102     jonathan 1180:
1.144     dyoung   1181:        if (paddr != NULL)
                   1182:                *paddr = NULL;
                   1183:        if (controlp != NULL)
                   1184:                *controlp = NULL;
                   1185:        if (flagsp != NULL)
1.1       cgd      1186:                flags = *flagsp &~ MSG_EOR;
                   1187:        else
                   1188:                flags = 0;
1.66      enami    1189:
                   1190:        if ((flags & MSG_DONTWAIT) == 0)
1.117     yamt     1191:                sodopendfree();
1.66      enami    1192:
1.1       cgd      1193:        if (flags & MSG_OOB) {
                   1194:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1195:                solock(so);
1.17      cgd      1196:                error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140     dyoung   1197:                    (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.160     ad       1198:                sounlock(so);
1.1       cgd      1199:                if (error)
                   1200:                        goto bad;
                   1201:                do {
1.134     christos 1202:                        error = uiomove(mtod(m, void *),
1.1       cgd      1203:                            (int) min(uio->uio_resid, m->m_len), uio);
                   1204:                        m = m_free(m);
1.144     dyoung   1205:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1206:  bad:
1.144     dyoung   1207:                if (m != NULL)
1.1       cgd      1208:                        m_freem(m);
1.144     dyoung   1209:                return error;
1.1       cgd      1210:        }
1.144     dyoung   1211:        if (mp != NULL)
1.140     dyoung   1212:                *mp = NULL;
1.160     ad       1213:
                   1214:        /*
                   1215:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1216:         * protocol processing soft interrupts from interrupting us and
                   1217:         * blocking (expensive).
                   1218:         */
                   1219:        s = splsoftnet();
                   1220:        solock(so);
1.1       cgd      1221:        if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140     dyoung   1222:                (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1       cgd      1223:
1.54      lukem    1224:  restart:
1.160     ad       1225:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1226:                sounlock(so);
                   1227:                splx(s);
1.144     dyoung   1228:                return error;
1.160     ad       1229:        }
1.1       cgd      1230:
                   1231:        m = so->so_rcv.sb_mb;
                   1232:        /*
                   1233:         * If we have less data than requested, block awaiting more
                   1234:         * (subject to any timeout) if:
1.15      mycroft  1235:         *   1. the current count is less than the low water mark,
1.1       cgd      1236:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1237:         *      receive operation at once if we block (resid <= hiwat), or
                   1238:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1239:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1240:         * we have to do the receive in sections, and thus risk returning
                   1241:         * a short count if a timeout or signal occurs after we start.
                   1242:         */
1.144     dyoung   1243:        if (m == NULL ||
                   1244:            ((flags & MSG_DONTWAIT) == 0 &&
                   1245:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1246:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1247:              ((flags & MSG_WAITALL) &&
                   1248:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1249:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1250: #ifdef DIAGNOSTIC
1.144     dyoung   1251:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1252:                        panic("receive 1");
                   1253: #endif
                   1254:                if (so->so_error) {
1.144     dyoung   1255:                        if (m != NULL)
1.15      mycroft  1256:                                goto dontblock;
1.1       cgd      1257:                        error = so->so_error;
                   1258:                        if ((flags & MSG_PEEK) == 0)
                   1259:                                so->so_error = 0;
                   1260:                        goto release;
                   1261:                }
                   1262:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1263:                        if (m != NULL)
1.15      mycroft  1264:                                goto dontblock;
1.1       cgd      1265:                        else
                   1266:                                goto release;
                   1267:                }
1.144     dyoung   1268:                for (; m != NULL; m = m->m_next)
1.1       cgd      1269:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1270:                                m = so->so_rcv.sb_mb;
                   1271:                                goto dontblock;
                   1272:                        }
                   1273:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1274:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1275:                        error = ENOTCONN;
                   1276:                        goto release;
                   1277:                }
                   1278:                if (uio->uio_resid == 0)
                   1279:                        goto release;
1.151     ad       1280:                if (so->so_nbio || (flags & MSG_DONTWAIT)) {
1.1       cgd      1281:                        error = EWOULDBLOCK;
                   1282:                        goto release;
                   1283:                }
1.69      thorpej  1284:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1285:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1286:                sbunlock(&so->so_rcv);
1.196     dsl      1287:                if (wakeup_state & SS_RESTARTSYS)
                   1288:                        error = ERESTART;
                   1289:                else
                   1290:                        error = sbwait(&so->so_rcv);
1.160     ad       1291:                if (error != 0) {
                   1292:                        sounlock(so);
                   1293:                        splx(s);
1.144     dyoung   1294:                        return error;
1.160     ad       1295:                }
1.196     dsl      1296:                wakeup_state = so->so_state;
1.1       cgd      1297:                goto restart;
                   1298:        }
1.54      lukem    1299:  dontblock:
1.69      thorpej  1300:        /*
                   1301:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1302:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1303:         * pointer to the next record in the socket buffer.  We must keep the
                   1304:         * various socket buffer pointers and local stack versions of the
                   1305:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1306:         * socket lock, and re-reading them when picking it up.
1.159     ad       1307:         *
                   1308:         * Otherwise, we will race with the network stack appending new data
                   1309:         * or records onto the socket buffer by using inconsistent/stale
                   1310:         * versions of the field, possibly resulting in socket buffer
                   1311:         * corruption.
                   1312:         *
                   1313:         * By holding the high-level sblock(), we prevent simultaneous
                   1314:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1315:         */
1.144     dyoung   1316:        if (l != NULL)
1.157     ad       1317:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1318:        KASSERT(m == so->so_rcv.sb_mb);
                   1319:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1320:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1321:        nextrecord = m->m_nextpkt;
                   1322:        if (pr->pr_flags & PR_ADDR) {
                   1323: #ifdef DIAGNOSTIC
                   1324:                if (m->m_type != MT_SONAME)
                   1325:                        panic("receive 1a");
                   1326: #endif
1.3       andrew   1327:                orig_resid = 0;
1.1       cgd      1328:                if (flags & MSG_PEEK) {
                   1329:                        if (paddr)
                   1330:                                *paddr = m_copy(m, 0, m->m_len);
                   1331:                        m = m->m_next;
                   1332:                } else {
                   1333:                        sbfree(&so->so_rcv, m);
1.67      he       1334:                        mbuf_removed = 1;
1.144     dyoung   1335:                        if (paddr != NULL) {
1.1       cgd      1336:                                *paddr = m;
                   1337:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1338:                                m->m_next = NULL;
1.1       cgd      1339:                                m = so->so_rcv.sb_mb;
                   1340:                        } else {
                   1341:                                MFREE(m, so->so_rcv.sb_mb);
                   1342:                                m = so->so_rcv.sb_mb;
                   1343:                        }
1.159     ad       1344:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1345:                }
                   1346:        }
1.159     ad       1347:
                   1348:        /*
                   1349:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1350:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1351:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1352:         * perform externalization (or freeing if controlp == NULL).
                   1353:         */
                   1354:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1355:                struct mbuf *cm = NULL, *cmn;
                   1356:                struct mbuf **cme = &cm;
                   1357:
                   1358:                do {
                   1359:                        if (flags & MSG_PEEK) {
                   1360:                                if (controlp != NULL) {
                   1361:                                        *controlp = m_copy(m, 0, m->m_len);
                   1362:                                        controlp = &(*controlp)->m_next;
                   1363:                                }
                   1364:                                m = m->m_next;
                   1365:                        } else {
                   1366:                                sbfree(&so->so_rcv, m);
1.1       cgd      1367:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1368:                                m->m_next = NULL;
1.159     ad       1369:                                *cme = m;
                   1370:                                cme = &(*cme)->m_next;
1.1       cgd      1371:                                m = so->so_rcv.sb_mb;
1.159     ad       1372:                        }
                   1373:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1374:                if ((flags & MSG_PEEK) == 0)
                   1375:                        sbsync(&so->so_rcv, nextrecord);
                   1376:                for (; cm != NULL; cm = cmn) {
                   1377:                        cmn = cm->m_next;
                   1378:                        cm->m_next = NULL;
                   1379:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1380:                        if (controlp != NULL) {
                   1381:                                if (dom->dom_externalize != NULL &&
                   1382:                                    type == SCM_RIGHTS) {
1.160     ad       1383:                                        sounlock(so);
1.159     ad       1384:                                        splx(s);
1.204   ! christos 1385:                                        error = (*dom->dom_externalize)(cm, l,
        !          1386:                                            (flags & MSG_CMSG_CLOEXEC) ?
        !          1387:                                            O_CLOEXEC : 0);
1.159     ad       1388:                                        s = splsoftnet();
1.160     ad       1389:                                        solock(so);
1.159     ad       1390:                                }
                   1391:                                *controlp = cm;
                   1392:                                while (*controlp != NULL)
                   1393:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1394:                        } else {
1.106     itojun   1395:                                /*
                   1396:                                 * Dispose of any SCM_RIGHTS message that went
                   1397:                                 * through the read path rather than recv.
                   1398:                                 */
1.159     ad       1399:                                if (dom->dom_dispose != NULL &&
                   1400:                                    type == SCM_RIGHTS) {
1.160     ad       1401:                                        sounlock(so);
1.159     ad       1402:                                        (*dom->dom_dispose)(cm);
1.160     ad       1403:                                        solock(so);
1.159     ad       1404:                                }
                   1405:                                m_freem(cm);
1.1       cgd      1406:                        }
                   1407:                }
1.159     ad       1408:                if (m != NULL)
                   1409:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1410:                else
                   1411:                        nextrecord = so->so_rcv.sb_mb;
                   1412:                orig_resid = 0;
1.1       cgd      1413:        }
1.69      thorpej  1414:
1.159     ad       1415:        /* If m is non-NULL, we have some data to read. */
                   1416:        if (__predict_true(m != NULL)) {
1.1       cgd      1417:                type = m->m_type;
                   1418:                if (type == MT_OOBDATA)
                   1419:                        flags |= MSG_OOB;
                   1420:        }
1.69      thorpej  1421:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1422:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1423:
1.1       cgd      1424:        moff = 0;
                   1425:        offset = 0;
1.144     dyoung   1426:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1427:                if (m->m_type == MT_OOBDATA) {
                   1428:                        if (type != MT_OOBDATA)
                   1429:                                break;
                   1430:                } else if (type == MT_OOBDATA)
                   1431:                        break;
                   1432: #ifdef DIAGNOSTIC
                   1433:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1434:                        panic("receive 3");
                   1435: #endif
                   1436:                so->so_state &= ~SS_RCVATMARK;
1.196     dsl      1437:                wakeup_state = 0;
1.1       cgd      1438:                len = uio->uio_resid;
                   1439:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1440:                        len = so->so_oobmark - offset;
                   1441:                if (len > m->m_len - moff)
                   1442:                        len = m->m_len - moff;
                   1443:                /*
                   1444:                 * If mp is set, just pass back the mbufs.
                   1445:                 * Otherwise copy them out via the uio, then free.
                   1446:                 * Sockbuf must be consistent here (points to current mbuf,
                   1447:                 * it points to next record) when we drop priority;
                   1448:                 * we must note any additions to the sockbuf when we
                   1449:                 * block interrupts again.
                   1450:                 */
1.144     dyoung   1451:                if (mp == NULL) {
1.69      thorpej  1452:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1453:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1454:                        sounlock(so);
1.1       cgd      1455:                        splx(s);
1.134     christos 1456:                        error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20      mycroft  1457:                        s = splsoftnet();
1.160     ad       1458:                        solock(so);
1.144     dyoung   1459:                        if (error != 0) {
1.67      he       1460:                                /*
                   1461:                                 * If any part of the record has been removed
                   1462:                                 * (such as the MT_SONAME mbuf, which will
                   1463:                                 * happen when PR_ADDR, and thus also
                   1464:                                 * PR_ATOMIC, is set), then drop the entire
                   1465:                                 * record to maintain the atomicity of the
                   1466:                                 * receive operation.
                   1467:                                 *
                   1468:                                 * This avoids a later panic("receive 1a")
                   1469:                                 * when compiled with DIAGNOSTIC.
                   1470:                                 */
1.146     dyoung   1471:                                if (m && mbuf_removed && atomic)
1.67      he       1472:                                        (void) sbdroprecord(&so->so_rcv);
                   1473:
1.57      jdolecek 1474:                                goto release;
1.67      he       1475:                        }
1.1       cgd      1476:                } else
                   1477:                        uio->uio_resid -= len;
                   1478:                if (len == m->m_len - moff) {
                   1479:                        if (m->m_flags & M_EOR)
                   1480:                                flags |= MSG_EOR;
                   1481:                        if (flags & MSG_PEEK) {
                   1482:                                m = m->m_next;
                   1483:                                moff = 0;
                   1484:                        } else {
                   1485:                                nextrecord = m->m_nextpkt;
                   1486:                                sbfree(&so->so_rcv, m);
                   1487:                                if (mp) {
                   1488:                                        *mp = m;
                   1489:                                        mp = &m->m_next;
                   1490:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1491:                                        *mp = NULL;
1.1       cgd      1492:                                } else {
                   1493:                                        MFREE(m, so->so_rcv.sb_mb);
                   1494:                                        m = so->so_rcv.sb_mb;
                   1495:                                }
1.69      thorpej  1496:                                /*
                   1497:                                 * If m != NULL, we also know that
                   1498:                                 * so->so_rcv.sb_mb != NULL.
                   1499:                                 */
                   1500:                                KASSERT(so->so_rcv.sb_mb == m);
                   1501:                                if (m) {
1.1       cgd      1502:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1503:                                        if (nextrecord == NULL)
                   1504:                                                so->so_rcv.sb_lastrecord = m;
                   1505:                                } else {
                   1506:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1507:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1508:                                }
                   1509:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1510:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1511:                        }
1.144     dyoung   1512:                } else if (flags & MSG_PEEK)
                   1513:                        moff += len;
                   1514:                else {
1.160     ad       1515:                        if (mp != NULL) {
                   1516:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1517:                                if (__predict_false(mt == NULL)) {
                   1518:                                        sounlock(so);
                   1519:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1520:                                        solock(so);
                   1521:                                }
                   1522:                                *mp = mt;
                   1523:                        }
1.144     dyoung   1524:                        m->m_data += len;
                   1525:                        m->m_len -= len;
                   1526:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1527:                }
                   1528:                if (so->so_oobmark) {
                   1529:                        if ((flags & MSG_PEEK) == 0) {
                   1530:                                so->so_oobmark -= len;
                   1531:                                if (so->so_oobmark == 0) {
                   1532:                                        so->so_state |= SS_RCVATMARK;
                   1533:                                        break;
                   1534:                                }
1.7       cgd      1535:                        } else {
1.1       cgd      1536:                                offset += len;
1.7       cgd      1537:                                if (offset == so->so_oobmark)
                   1538:                                        break;
                   1539:                        }
1.1       cgd      1540:                }
                   1541:                if (flags & MSG_EOR)
                   1542:                        break;
                   1543:                /*
                   1544:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1545:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1546:                 * termination.  If a signal/timeout occurs, return
                   1547:                 * with a short count but without error.
                   1548:                 * Keep sockbuf locked against other readers.
                   1549:                 */
1.144     dyoung   1550:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1551:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1552:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1553:                                break;
1.68      matt     1554:                        /*
                   1555:                         * If we are peeking and the socket receive buffer is
                   1556:                         * full, stop since we can't get more data to peek at.
                   1557:                         */
                   1558:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1559:                                break;
                   1560:                        /*
                   1561:                         * If we've drained the socket buffer, tell the
                   1562:                         * protocol in case it needs to do something to
                   1563:                         * get it filled again.
                   1564:                         */
                   1565:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
                   1566:                                (*pr->pr_usrreq)(so, PRU_RCVD,
1.140     dyoung   1567:                                    NULL, (struct mbuf *)(long)flags, NULL, l);
1.69      thorpej  1568:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1569:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.196     dsl      1570:                        if (wakeup_state & SS_RESTARTSYS)
                   1571:                                error = ERESTART;
                   1572:                        else
                   1573:                                error = sbwait(&so->so_rcv);
1.144     dyoung   1574:                        if (error != 0) {
1.1       cgd      1575:                                sbunlock(&so->so_rcv);
1.160     ad       1576:                                sounlock(so);
1.1       cgd      1577:                                splx(s);
1.144     dyoung   1578:                                return 0;
1.1       cgd      1579:                        }
1.21      christos 1580:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1581:                                nextrecord = m->m_nextpkt;
1.196     dsl      1582:                        wakeup_state = so->so_state;
1.1       cgd      1583:                }
                   1584:        }
1.3       andrew   1585:
1.146     dyoung   1586:        if (m && atomic) {
1.3       andrew   1587:                flags |= MSG_TRUNC;
                   1588:                if ((flags & MSG_PEEK) == 0)
                   1589:                        (void) sbdroprecord(&so->so_rcv);
                   1590:        }
1.1       cgd      1591:        if ((flags & MSG_PEEK) == 0) {
1.144     dyoung   1592:                if (m == NULL) {
1.69      thorpej  1593:                        /*
1.70      thorpej  1594:                         * First part is an inline SB_EMPTY_FIXUP().  Second
1.69      thorpej  1595:                         * part makes sure sb_lastrecord is up-to-date if
                   1596:                         * there is still data in the socket buffer.
                   1597:                         */
1.1       cgd      1598:                        so->so_rcv.sb_mb = nextrecord;
1.69      thorpej  1599:                        if (so->so_rcv.sb_mb == NULL) {
                   1600:                                so->so_rcv.sb_mbtail = NULL;
                   1601:                                so->so_rcv.sb_lastrecord = NULL;
                   1602:                        } else if (nextrecord->m_nextpkt == NULL)
                   1603:                                so->so_rcv.sb_lastrecord = nextrecord;
                   1604:                }
                   1605:                SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
                   1606:                SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1       cgd      1607:                if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140     dyoung   1608:                        (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
                   1609:                            (struct mbuf *)(long)flags, NULL, l);
1.1       cgd      1610:        }
1.3       andrew   1611:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1612:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1613:                sbunlock(&so->so_rcv);
                   1614:                goto restart;
                   1615:        }
1.108     perry    1616:
1.144     dyoung   1617:        if (flagsp != NULL)
1.1       cgd      1618:                *flagsp |= flags;
1.54      lukem    1619:  release:
1.1       cgd      1620:        sbunlock(&so->so_rcv);
1.160     ad       1621:        sounlock(so);
1.1       cgd      1622:        splx(s);
1.144     dyoung   1623:        return error;
1.1       cgd      1624: }
                   1625:
1.14      mycroft  1626: int
1.54      lukem    1627: soshutdown(struct socket *so, int how)
1.1       cgd      1628: {
1.99      matt     1629:        const struct protosw    *pr;
1.160     ad       1630:        int     error;
                   1631:
                   1632:        KASSERT(solocked(so));
1.34      kleink   1633:
1.54      lukem    1634:        pr = so->so_proto;
1.34      kleink   1635:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1636:                return (EINVAL);
1.1       cgd      1637:
1.160     ad       1638:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1639:                sorflush(so);
1.160     ad       1640:                error = 0;
                   1641:        }
1.34      kleink   1642:        if (how == SHUT_WR || how == SHUT_RDWR)
1.160     ad       1643:                error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1.140     dyoung   1644:                    NULL, NULL, NULL);
1.160     ad       1645:
                   1646:        return error;
1.1       cgd      1647: }
                   1648:
1.195     dsl      1649: void
1.196     dsl      1650: sorestart(struct socket *so)
1.188     ad       1651: {
1.196     dsl      1652:        /*
                   1653:         * An application has called close() on an fd on which another
                   1654:         * of its threads has called a socket system call.
                   1655:         * Mark this and wake everyone up, and code that would block again
                   1656:         * instead returns ERESTART.
                   1657:         * On system call re-entry the fd is validated and EBADF returned.
                   1658:         * Any other fd will block again on the 2nd syscall.
                   1659:         */
1.188     ad       1660:        solock(so);
1.196     dsl      1661:        so->so_state |= SS_RESTARTSYS;
1.188     ad       1662:        cv_broadcast(&so->so_cv);
1.196     dsl      1663:        cv_broadcast(&so->so_snd.sb_cv);
                   1664:        cv_broadcast(&so->so_rcv.sb_cv);
1.188     ad       1665:        sounlock(so);
                   1666: }
                   1667:
1.14      mycroft  1668: void
1.54      lukem    1669: sorflush(struct socket *so)
1.1       cgd      1670: {
1.54      lukem    1671:        struct sockbuf  *sb, asb;
1.99      matt     1672:        const struct protosw    *pr;
1.160     ad       1673:
                   1674:        KASSERT(solocked(so));
1.1       cgd      1675:
1.54      lukem    1676:        sb = &so->so_rcv;
                   1677:        pr = so->so_proto;
1.160     ad       1678:        socantrcvmore(so);
1.1       cgd      1679:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1680:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1681:        sbunlock(sb);
                   1682:        asb = *sb;
1.86      wrstuden 1683:        /*
                   1684:         * Clear most of the sockbuf structure, but leave some of the
                   1685:         * fields valid.
                   1686:         */
                   1687:        memset(&sb->sb_startzero, 0,
                   1688:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1689:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1690:                sounlock(so);
1.1       cgd      1691:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1692:                solock(so);
                   1693:        }
1.98      christos 1694:        sbrelease(&asb, so);
1.1       cgd      1695: }
                   1696:
1.171     plunky   1697: /*
                   1698:  * internal set SOL_SOCKET options
                   1699:  */
1.142     dyoung   1700: static int
1.171     plunky   1701: sosetopt1(struct socket *so, const struct sockopt *sopt)
1.1       cgd      1702: {
1.182     christos 1703:        int error = EINVAL, optval, opt;
1.171     plunky   1704:        struct linger l;
                   1705:        struct timeval tv;
1.142     dyoung   1706:
1.179     christos 1707:        switch ((opt = sopt->sopt_name)) {
1.142     dyoung   1708:
1.170     tls      1709:        case SO_ACCEPTFILTER:
1.177     ad       1710:                error = accept_filt_setopt(so, sopt);
                   1711:                KASSERT(solocked(so));
1.170     tls      1712:                break;
                   1713:
1.171     plunky   1714:        case SO_LINGER:
                   1715:                error = sockopt_get(sopt, &l, sizeof(l));
1.177     ad       1716:                solock(so);
1.171     plunky   1717:                if (error)
1.177     ad       1718:                        break;
1.171     plunky   1719:                if (l.l_linger < 0 || l.l_linger > USHRT_MAX ||
1.177     ad       1720:                    l.l_linger > (INT_MAX / hz)) {
                   1721:                        error = EDOM;
                   1722:                        break;
                   1723:                }
1.171     plunky   1724:                so->so_linger = l.l_linger;
                   1725:                if (l.l_onoff)
                   1726:                        so->so_options |= SO_LINGER;
                   1727:                else
                   1728:                        so->so_options &= ~SO_LINGER;
1.177     ad       1729:                break;
1.1       cgd      1730:
1.142     dyoung   1731:        case SO_DEBUG:
                   1732:        case SO_KEEPALIVE:
                   1733:        case SO_DONTROUTE:
                   1734:        case SO_USELOOPBACK:
                   1735:        case SO_BROADCAST:
                   1736:        case SO_REUSEADDR:
                   1737:        case SO_REUSEPORT:
                   1738:        case SO_OOBINLINE:
                   1739:        case SO_TIMESTAMP:
1.184     christos 1740: #ifdef SO_OTIMESTAMP
                   1741:        case SO_OTIMESTAMP:
                   1742: #endif
1.171     plunky   1743:                error = sockopt_getint(sopt, &optval);
1.177     ad       1744:                solock(so);
1.171     plunky   1745:                if (error)
1.177     ad       1746:                        break;
1.171     plunky   1747:                if (optval)
1.179     christos 1748:                        so->so_options |= opt;
1.142     dyoung   1749:                else
1.179     christos 1750:                        so->so_options &= ~opt;
1.142     dyoung   1751:                break;
                   1752:
                   1753:        case SO_SNDBUF:
                   1754:        case SO_RCVBUF:
                   1755:        case SO_SNDLOWAT:
                   1756:        case SO_RCVLOWAT:
1.171     plunky   1757:                error = sockopt_getint(sopt, &optval);
1.177     ad       1758:                solock(so);
1.171     plunky   1759:                if (error)
1.177     ad       1760:                        break;
1.1       cgd      1761:
1.142     dyoung   1762:                /*
                   1763:                 * Values < 1 make no sense for any of these
                   1764:                 * options, so disallow them.
                   1765:                 */
1.177     ad       1766:                if (optval < 1) {
                   1767:                        error = EINVAL;
                   1768:                        break;
                   1769:                }
1.1       cgd      1770:
1.179     christos 1771:                switch (opt) {
1.171     plunky   1772:                case SO_SNDBUF:
1.177     ad       1773:                        if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) {
                   1774:                                error = ENOBUFS;
                   1775:                                break;
                   1776:                        }
1.171     plunky   1777:                        so->so_snd.sb_flags &= ~SB_AUTOSIZE;
                   1778:                        break;
1.1       cgd      1779:
                   1780:                case SO_RCVBUF:
1.177     ad       1781:                        if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) {
                   1782:                                error = ENOBUFS;
                   1783:                                break;
                   1784:                        }
1.171     plunky   1785:                        so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1.142     dyoung   1786:                        break;
                   1787:
                   1788:                /*
                   1789:                 * Make sure the low-water is never greater than
                   1790:                 * the high-water.
                   1791:                 */
1.1       cgd      1792:                case SO_SNDLOWAT:
1.171     plunky   1793:                        if (optval > so->so_snd.sb_hiwat)
                   1794:                                optval = so->so_snd.sb_hiwat;
                   1795:
                   1796:                        so->so_snd.sb_lowat = optval;
1.142     dyoung   1797:                        break;
1.171     plunky   1798:
1.1       cgd      1799:                case SO_RCVLOWAT:
1.171     plunky   1800:                        if (optval > so->so_rcv.sb_hiwat)
                   1801:                                optval = so->so_rcv.sb_hiwat;
                   1802:
                   1803:                        so->so_rcv.sb_lowat = optval;
1.142     dyoung   1804:                        break;
                   1805:                }
                   1806:                break;
1.28      thorpej  1807:
1.179     christos 1808: #ifdef COMPAT_50
                   1809:        case SO_OSNDTIMEO:
                   1810:        case SO_ORCVTIMEO: {
                   1811:                struct timeval50 otv;
                   1812:                error = sockopt_get(sopt, &otv, sizeof(otv));
1.186     pooka    1813:                if (error) {
                   1814:                        solock(so);
1.183     christos 1815:                        break;
1.186     pooka    1816:                }
1.179     christos 1817:                timeval50_to_timeval(&otv, &tv);
                   1818:                opt = opt == SO_OSNDTIMEO ? SO_SNDTIMEO : SO_RCVTIMEO;
1.182     christos 1819:                error = 0;
1.179     christos 1820:                /*FALLTHROUGH*/
                   1821:        }
                   1822: #endif /* COMPAT_50 */
                   1823:
1.142     dyoung   1824:        case SO_SNDTIMEO:
                   1825:        case SO_RCVTIMEO:
1.182     christos 1826:                if (error)
1.179     christos 1827:                        error = sockopt_get(sopt, &tv, sizeof(tv));
1.177     ad       1828:                solock(so);
1.171     plunky   1829:                if (error)
1.177     ad       1830:                        break;
1.171     plunky   1831:
1.177     ad       1832:                if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) {
                   1833:                        error = EDOM;
                   1834:                        break;
                   1835:                }
1.28      thorpej  1836:
1.171     plunky   1837:                optval = tv.tv_sec * hz + tv.tv_usec / tick;
                   1838:                if (optval == 0 && tv.tv_usec != 0)
                   1839:                        optval = 1;
1.28      thorpej  1840:
1.179     christos 1841:                switch (opt) {
1.142     dyoung   1842:                case SO_SNDTIMEO:
1.171     plunky   1843:                        so->so_snd.sb_timeo = optval;
1.1       cgd      1844:                        break;
                   1845:                case SO_RCVTIMEO:
1.171     plunky   1846:                        so->so_rcv.sb_timeo = optval;
1.142     dyoung   1847:                        break;
                   1848:                }
                   1849:                break;
1.1       cgd      1850:
1.142     dyoung   1851:        default:
1.177     ad       1852:                solock(so);
                   1853:                error = ENOPROTOOPT;
                   1854:                break;
1.142     dyoung   1855:        }
1.177     ad       1856:        KASSERT(solocked(so));
                   1857:        return error;
1.142     dyoung   1858: }
1.1       cgd      1859:
1.142     dyoung   1860: int
1.171     plunky   1861: sosetopt(struct socket *so, struct sockopt *sopt)
1.142     dyoung   1862: {
                   1863:        int error, prerr;
1.1       cgd      1864:
1.177     ad       1865:        if (sopt->sopt_level == SOL_SOCKET) {
1.171     plunky   1866:                error = sosetopt1(so, sopt);
1.177     ad       1867:                KASSERT(solocked(so));
                   1868:        } else {
1.142     dyoung   1869:                error = ENOPROTOOPT;
1.177     ad       1870:                solock(so);
                   1871:        }
1.1       cgd      1872:
1.142     dyoung   1873:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1874:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1875:                /* give the protocol stack a shot */
1.171     plunky   1876:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt);
1.142     dyoung   1877:                if (prerr == 0)
                   1878:                        error = 0;
                   1879:                else if (prerr != ENOPROTOOPT)
                   1880:                        error = prerr;
1.171     plunky   1881:        }
1.160     ad       1882:        sounlock(so);
1.142     dyoung   1883:        return error;
1.1       cgd      1884: }
                   1885:
1.171     plunky   1886: /*
                   1887:  * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt()
                   1888:  */
                   1889: int
                   1890: so_setsockopt(struct lwp *l, struct socket *so, int level, int name,
                   1891:     const void *val, size_t valsize)
                   1892: {
                   1893:        struct sockopt sopt;
                   1894:        int error;
                   1895:
                   1896:        KASSERT(valsize == 0 || val != NULL);
                   1897:
                   1898:        sockopt_init(&sopt, level, name, valsize);
                   1899:        sockopt_set(&sopt, val, valsize);
                   1900:
                   1901:        error = sosetopt(so, &sopt);
                   1902:
                   1903:        sockopt_destroy(&sopt);
                   1904:
                   1905:        return error;
                   1906: }
                   1907:
                   1908: /*
                   1909:  * internal get SOL_SOCKET options
                   1910:  */
                   1911: static int
                   1912: sogetopt1(struct socket *so, struct sockopt *sopt)
                   1913: {
1.179     christos 1914:        int error, optval, opt;
1.171     plunky   1915:        struct linger l;
                   1916:        struct timeval tv;
                   1917:
1.179     christos 1918:        switch ((opt = sopt->sopt_name)) {
1.171     plunky   1919:
                   1920:        case SO_ACCEPTFILTER:
1.177     ad       1921:                error = accept_filt_getopt(so, sopt);
1.171     plunky   1922:                break;
                   1923:
                   1924:        case SO_LINGER:
                   1925:                l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0;
                   1926:                l.l_linger = so->so_linger;
                   1927:
                   1928:                error = sockopt_set(sopt, &l, sizeof(l));
                   1929:                break;
                   1930:
                   1931:        case SO_USELOOPBACK:
                   1932:        case SO_DONTROUTE:
                   1933:        case SO_DEBUG:
                   1934:        case SO_KEEPALIVE:
                   1935:        case SO_REUSEADDR:
                   1936:        case SO_REUSEPORT:
                   1937:        case SO_BROADCAST:
                   1938:        case SO_OOBINLINE:
                   1939:        case SO_TIMESTAMP:
1.184     christos 1940: #ifdef SO_OTIMESTAMP
                   1941:        case SO_OTIMESTAMP:
                   1942: #endif
1.179     christos 1943:                error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0);
1.171     plunky   1944:                break;
                   1945:
                   1946:        case SO_TYPE:
                   1947:                error = sockopt_setint(sopt, so->so_type);
                   1948:                break;
                   1949:
                   1950:        case SO_ERROR:
                   1951:                error = sockopt_setint(sopt, so->so_error);
                   1952:                so->so_error = 0;
                   1953:                break;
                   1954:
                   1955:        case SO_SNDBUF:
                   1956:                error = sockopt_setint(sopt, so->so_snd.sb_hiwat);
                   1957:                break;
                   1958:
                   1959:        case SO_RCVBUF:
                   1960:                error = sockopt_setint(sopt, so->so_rcv.sb_hiwat);
                   1961:                break;
                   1962:
                   1963:        case SO_SNDLOWAT:
                   1964:                error = sockopt_setint(sopt, so->so_snd.sb_lowat);
                   1965:                break;
                   1966:
                   1967:        case SO_RCVLOWAT:
                   1968:                error = sockopt_setint(sopt, so->so_rcv.sb_lowat);
                   1969:                break;
                   1970:
1.179     christos 1971: #ifdef COMPAT_50
                   1972:        case SO_OSNDTIMEO:
                   1973:        case SO_ORCVTIMEO: {
                   1974:                struct timeval50 otv;
                   1975:
                   1976:                optval = (opt == SO_OSNDTIMEO ?
                   1977:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1978:
                   1979:                otv.tv_sec = optval / hz;
                   1980:                otv.tv_usec = (optval % hz) * tick;
                   1981:
                   1982:                error = sockopt_set(sopt, &otv, sizeof(otv));
                   1983:                break;
                   1984:        }
                   1985: #endif /* COMPAT_50 */
                   1986:
1.171     plunky   1987:        case SO_SNDTIMEO:
                   1988:        case SO_RCVTIMEO:
1.179     christos 1989:                optval = (opt == SO_SNDTIMEO ?
1.171     plunky   1990:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1991:
                   1992:                tv.tv_sec = optval / hz;
                   1993:                tv.tv_usec = (optval % hz) * tick;
                   1994:
                   1995:                error = sockopt_set(sopt, &tv, sizeof(tv));
                   1996:                break;
                   1997:
                   1998:        case SO_OVERFLOWED:
                   1999:                error = sockopt_setint(sopt, so->so_rcv.sb_overflowed);
                   2000:                break;
                   2001:
                   2002:        default:
                   2003:                error = ENOPROTOOPT;
                   2004:                break;
                   2005:        }
                   2006:
                   2007:        return (error);
                   2008: }
                   2009:
1.14      mycroft  2010: int
1.171     plunky   2011: sogetopt(struct socket *so, struct sockopt *sopt)
1.1       cgd      2012: {
1.160     ad       2013:        int             error;
1.1       cgd      2014:
1.160     ad       2015:        solock(so);
1.171     plunky   2016:        if (sopt->sopt_level != SOL_SOCKET) {
1.1       cgd      2017:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       2018:                        error = ((*so->so_proto->pr_ctloutput)
1.171     plunky   2019:                            (PRCO_GETOPT, so, sopt));
1.1       cgd      2020:                } else
1.160     ad       2021:                        error = (ENOPROTOOPT);
1.1       cgd      2022:        } else {
1.171     plunky   2023:                error = sogetopt1(so, sopt);
                   2024:        }
                   2025:        sounlock(so);
                   2026:        return (error);
                   2027: }
                   2028:
                   2029: /*
                   2030:  * alloc sockopt data buffer buffer
                   2031:  *     - will be released at destroy
                   2032:  */
1.176     plunky   2033: static int
                   2034: sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag)
1.171     plunky   2035: {
                   2036:
                   2037:        KASSERT(sopt->sopt_size == 0);
                   2038:
1.176     plunky   2039:        if (len > sizeof(sopt->sopt_buf)) {
                   2040:                sopt->sopt_data = kmem_zalloc(len, kmflag);
                   2041:                if (sopt->sopt_data == NULL)
                   2042:                        return ENOMEM;
                   2043:        } else
1.171     plunky   2044:                sopt->sopt_data = sopt->sopt_buf;
                   2045:
                   2046:        sopt->sopt_size = len;
1.176     plunky   2047:        return 0;
1.171     plunky   2048: }
                   2049:
                   2050: /*
                   2051:  * initialise sockopt storage
1.176     plunky   2052:  *     - MAY sleep during allocation
1.171     plunky   2053:  */
                   2054: void
                   2055: sockopt_init(struct sockopt *sopt, int level, int name, size_t size)
                   2056: {
1.1       cgd      2057:
1.171     plunky   2058:        memset(sopt, 0, sizeof(*sopt));
1.1       cgd      2059:
1.171     plunky   2060:        sopt->sopt_level = level;
                   2061:        sopt->sopt_name = name;
1.176     plunky   2062:        (void)sockopt_alloc(sopt, size, KM_SLEEP);
1.171     plunky   2063: }
                   2064:
                   2065: /*
                   2066:  * destroy sockopt storage
                   2067:  *     - will release any held memory references
                   2068:  */
                   2069: void
                   2070: sockopt_destroy(struct sockopt *sopt)
                   2071: {
                   2072:
                   2073:        if (sopt->sopt_data != sopt->sopt_buf)
1.173     plunky   2074:                kmem_free(sopt->sopt_data, sopt->sopt_size);
1.171     plunky   2075:
                   2076:        memset(sopt, 0, sizeof(*sopt));
                   2077: }
                   2078:
                   2079: /*
                   2080:  * set sockopt value
                   2081:  *     - value is copied into sockopt
1.176     plunky   2082:  *     - memory is allocated when necessary, will not sleep
1.171     plunky   2083:  */
                   2084: int
                   2085: sockopt_set(struct sockopt *sopt, const void *buf, size_t len)
                   2086: {
1.176     plunky   2087:        int error;
1.171     plunky   2088:
1.176     plunky   2089:        if (sopt->sopt_size == 0) {
                   2090:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2091:                if (error)
                   2092:                        return error;
                   2093:        }
1.171     plunky   2094:
                   2095:        KASSERT(sopt->sopt_size == len);
                   2096:        memcpy(sopt->sopt_data, buf, len);
                   2097:        return 0;
                   2098: }
                   2099:
                   2100: /*
                   2101:  * common case of set sockopt integer value
                   2102:  */
                   2103: int
                   2104: sockopt_setint(struct sockopt *sopt, int val)
                   2105: {
                   2106:
                   2107:        return sockopt_set(sopt, &val, sizeof(int));
                   2108: }
                   2109:
                   2110: /*
                   2111:  * get sockopt value
                   2112:  *     - correct size must be given
                   2113:  */
                   2114: int
                   2115: sockopt_get(const struct sockopt *sopt, void *buf, size_t len)
                   2116: {
1.170     tls      2117:
1.171     plunky   2118:        if (sopt->sopt_size != len)
                   2119:                return EINVAL;
1.1       cgd      2120:
1.171     plunky   2121:        memcpy(buf, sopt->sopt_data, len);
                   2122:        return 0;
                   2123: }
1.1       cgd      2124:
1.171     plunky   2125: /*
                   2126:  * common case of get sockopt integer value
                   2127:  */
                   2128: int
                   2129: sockopt_getint(const struct sockopt *sopt, int *valp)
                   2130: {
1.1       cgd      2131:
1.171     plunky   2132:        return sockopt_get(sopt, valp, sizeof(int));
                   2133: }
1.1       cgd      2134:
1.171     plunky   2135: /*
                   2136:  * set sockopt value from mbuf
                   2137:  *     - ONLY for legacy code
                   2138:  *     - mbuf is released by sockopt
1.176     plunky   2139:  *     - will not sleep
1.171     plunky   2140:  */
                   2141: int
                   2142: sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m)
                   2143: {
                   2144:        size_t len;
1.176     plunky   2145:        int error;
1.1       cgd      2146:
1.171     plunky   2147:        len = m_length(m);
1.1       cgd      2148:
1.176     plunky   2149:        if (sopt->sopt_size == 0) {
                   2150:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2151:                if (error)
                   2152:                        return error;
                   2153:        }
1.1       cgd      2154:
1.171     plunky   2155:        KASSERT(sopt->sopt_size == len);
                   2156:        m_copydata(m, 0, len, sopt->sopt_data);
                   2157:        m_freem(m);
1.1       cgd      2158:
1.171     plunky   2159:        return 0;
                   2160: }
1.1       cgd      2161:
1.171     plunky   2162: /*
                   2163:  * get sockopt value into mbuf
                   2164:  *     - ONLY for legacy code
                   2165:  *     - mbuf to be released by the caller
1.176     plunky   2166:  *     - will not sleep
1.171     plunky   2167:  */
                   2168: struct mbuf *
                   2169: sockopt_getmbuf(const struct sockopt *sopt)
                   2170: {
                   2171:        struct mbuf *m;
1.107     darrenr  2172:
1.176     plunky   2173:        if (sopt->sopt_size > MCLBYTES)
                   2174:                return NULL;
                   2175:
                   2176:        m = m_get(M_DONTWAIT, MT_SOOPTS);
1.171     plunky   2177:        if (m == NULL)
                   2178:                return NULL;
                   2179:
1.176     plunky   2180:        if (sopt->sopt_size > MLEN) {
                   2181:                MCLGET(m, M_DONTWAIT);
                   2182:                if ((m->m_flags & M_EXT) == 0) {
                   2183:                        m_free(m);
                   2184:                        return NULL;
                   2185:                }
1.1       cgd      2186:        }
1.176     plunky   2187:
                   2188:        memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size);
                   2189:        m->m_len = sopt->sopt_size;
1.160     ad       2190:
1.171     plunky   2191:        return m;
1.1       cgd      2192: }
                   2193:
1.14      mycroft  2194: void
1.54      lukem    2195: sohasoutofband(struct socket *so)
1.1       cgd      2196: {
1.153     rmind    2197:
1.90      christos 2198:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.189     ad       2199:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT);
1.1       cgd      2200: }
1.72      jdolecek 2201:
                   2202: static void
                   2203: filt_sordetach(struct knote *kn)
                   2204: {
                   2205:        struct socket   *so;
                   2206:
1.155     ad       2207:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2208:        solock(so);
1.73      christos 2209:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2210:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 2211:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       2212:        sounlock(so);
1.72      jdolecek 2213: }
                   2214:
                   2215: /*ARGSUSED*/
                   2216: static int
1.129     yamt     2217: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 2218: {
                   2219:        struct socket   *so;
1.160     ad       2220:        int rv;
1.72      jdolecek 2221:
1.155     ad       2222:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2223:        if (hint != NOTE_SUBMIT)
                   2224:                solock(so);
1.72      jdolecek 2225:        kn->kn_data = so->so_rcv.sb_cc;
                   2226:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    2227:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2228:                kn->kn_fflags = so->so_error;
1.160     ad       2229:                rv = 1;
                   2230:        } else if (so->so_error)        /* temporary udp error */
                   2231:                rv = 1;
                   2232:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2233:                rv = (kn->kn_data >= kn->kn_sdata);
                   2234:        else
                   2235:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   2236:        if (hint != NOTE_SUBMIT)
                   2237:                sounlock(so);
                   2238:        return rv;
1.72      jdolecek 2239: }
                   2240:
                   2241: static void
                   2242: filt_sowdetach(struct knote *kn)
                   2243: {
                   2244:        struct socket   *so;
                   2245:
1.155     ad       2246:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2247:        solock(so);
1.73      christos 2248:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2249:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 2250:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       2251:        sounlock(so);
1.72      jdolecek 2252: }
                   2253:
                   2254: /*ARGSUSED*/
                   2255: static int
1.129     yamt     2256: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 2257: {
                   2258:        struct socket   *so;
1.160     ad       2259:        int rv;
1.72      jdolecek 2260:
1.155     ad       2261:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2262:        if (hint != NOTE_SUBMIT)
                   2263:                solock(so);
1.72      jdolecek 2264:        kn->kn_data = sbspace(&so->so_snd);
                   2265:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    2266:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2267:                kn->kn_fflags = so->so_error;
1.160     ad       2268:                rv = 1;
                   2269:        } else if (so->so_error)        /* temporary udp error */
                   2270:                rv = 1;
                   2271:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 2272:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       2273:                rv = 0;
                   2274:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2275:                rv = (kn->kn_data >= kn->kn_sdata);
                   2276:        else
                   2277:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   2278:        if (hint != NOTE_SUBMIT)
                   2279:                sounlock(so);
                   2280:        return rv;
1.72      jdolecek 2281: }
                   2282:
                   2283: /*ARGSUSED*/
                   2284: static int
1.129     yamt     2285: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 2286: {
                   2287:        struct socket   *so;
1.160     ad       2288:        int rv;
1.72      jdolecek 2289:
1.155     ad       2290:        so = ((file_t *)kn->kn_obj)->f_data;
1.72      jdolecek 2291:
                   2292:        /*
                   2293:         * Set kn_data to number of incoming connections, not
                   2294:         * counting partial (incomplete) connections.
1.108     perry    2295:         */
1.160     ad       2296:        if (hint != NOTE_SUBMIT)
                   2297:                solock(so);
1.72      jdolecek 2298:        kn->kn_data = so->so_qlen;
1.160     ad       2299:        rv = (kn->kn_data > 0);
                   2300:        if (hint != NOTE_SUBMIT)
                   2301:                sounlock(so);
                   2302:        return rv;
1.72      jdolecek 2303: }
                   2304:
                   2305: static const struct filterops solisten_filtops =
                   2306:        { 1, NULL, filt_sordetach, filt_solisten };
                   2307: static const struct filterops soread_filtops =
                   2308:        { 1, NULL, filt_sordetach, filt_soread };
                   2309: static const struct filterops sowrite_filtops =
                   2310:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   2311:
                   2312: int
1.129     yamt     2313: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 2314: {
                   2315:        struct socket   *so;
                   2316:        struct sockbuf  *sb;
                   2317:
1.155     ad       2318:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2319:        solock(so);
1.72      jdolecek 2320:        switch (kn->kn_filter) {
                   2321:        case EVFILT_READ:
                   2322:                if (so->so_options & SO_ACCEPTCONN)
                   2323:                        kn->kn_fop = &solisten_filtops;
                   2324:                else
                   2325:                        kn->kn_fop = &soread_filtops;
                   2326:                sb = &so->so_rcv;
                   2327:                break;
                   2328:        case EVFILT_WRITE:
                   2329:                kn->kn_fop = &sowrite_filtops;
                   2330:                sb = &so->so_snd;
                   2331:                break;
                   2332:        default:
1.160     ad       2333:                sounlock(so);
1.149     pooka    2334:                return (EINVAL);
1.72      jdolecek 2335:        }
1.73      christos 2336:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 2337:        sb->sb_flags |= SB_KNOTE;
1.160     ad       2338:        sounlock(so);
1.72      jdolecek 2339:        return (0);
                   2340: }
                   2341:
1.154     ad       2342: static int
                   2343: sodopoll(struct socket *so, int events)
                   2344: {
                   2345:        int revents;
                   2346:
                   2347:        revents = 0;
                   2348:
                   2349:        if (events & (POLLIN | POLLRDNORM))
                   2350:                if (soreadable(so))
                   2351:                        revents |= events & (POLLIN | POLLRDNORM);
                   2352:
                   2353:        if (events & (POLLOUT | POLLWRNORM))
                   2354:                if (sowritable(so))
                   2355:                        revents |= events & (POLLOUT | POLLWRNORM);
                   2356:
                   2357:        if (events & (POLLPRI | POLLRDBAND))
                   2358:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   2359:                        revents |= events & (POLLPRI | POLLRDBAND);
                   2360:
                   2361:        return revents;
                   2362: }
                   2363:
                   2364: int
                   2365: sopoll(struct socket *so, int events)
                   2366: {
                   2367:        int revents = 0;
                   2368:
1.160     ad       2369: #ifndef DIAGNOSTIC
                   2370:        /*
                   2371:         * Do a quick, unlocked check in expectation that the socket
                   2372:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   2373:         * as the solocked() assertions will fail.
                   2374:         */
1.154     ad       2375:        if ((revents = sodopoll(so, events)) != 0)
                   2376:                return revents;
1.160     ad       2377: #endif
1.154     ad       2378:
1.160     ad       2379:        solock(so);
1.154     ad       2380:        if ((revents = sodopoll(so, events)) == 0) {
                   2381:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   2382:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       2383:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       2384:                }
                   2385:
                   2386:                if (events & (POLLOUT | POLLWRNORM)) {
                   2387:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2388:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2389:                }
                   2390:        }
1.160     ad       2391:        sounlock(so);
1.154     ad       2392:
                   2393:        return revents;
                   2394: }
                   2395:
                   2396:
1.94      yamt     2397: #include <sys/sysctl.h>
                   2398:
                   2399: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
                   2400:
                   2401: /*
                   2402:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2403:  * value is not too small.
                   2404:  * (XXX should we maybe make sure it's not too large as well?)
                   2405:  */
                   2406: static int
                   2407: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2408: {
                   2409:        int error, new_somaxkva;
                   2410:        struct sysctlnode node;
                   2411:
                   2412:        new_somaxkva = somaxkva;
                   2413:        node = *rnode;
                   2414:        node.sysctl_data = &new_somaxkva;
                   2415:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2416:        if (error || newp == NULL)
                   2417:                return (error);
                   2418:
                   2419:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2420:                return (EINVAL);
                   2421:
1.136     ad       2422:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2423:        somaxkva = new_somaxkva;
1.136     ad       2424:        cv_broadcast(&socurkva_cv);
                   2425:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2426:
                   2427:        return (error);
                   2428: }
                   2429:
1.178     pooka    2430: static void
1.187     cegger   2431: sysctl_kern_somaxkva_setup(void)
1.94      yamt     2432: {
                   2433:
1.178     pooka    2434:        KASSERT(socket_sysctllog == NULL);
                   2435:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2436:                       CTLFLAG_PERMANENT,
                   2437:                       CTLTYPE_NODE, "kern", NULL,
                   2438:                       NULL, 0, NULL, 0,
                   2439:                       CTL_KERN, CTL_EOL);
                   2440:
1.178     pooka    2441:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2442:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2443:                       CTLTYPE_INT, "somaxkva",
                   2444:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2445:                                    "used for socket buffers"),
1.94      yamt     2446:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2447:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
                   2448: }

CVSweb <webmaster@jp.NetBSD.org>