[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 1.191

1.191   ! elad        1: /*     $NetBSD: uipc_socket.c,v 1.190 2009/09/11 22:06:29 dyoung 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>
1.191   ! elad       66: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.190 2009/09/11 22:06:29 dyoung 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:
1.64      thorpej   101: #include <uvm/uvm.h>
                    102:
1.77      thorpej   103: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
                    104: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37      thorpej   105:
1.142     dyoung    106: extern const struct fileops socketops;
                    107:
1.54      lukem     108: extern int     somaxconn;                      /* patchable (XXX sysctl) */
                    109: int            somaxconn = SOMAXCONN;
1.160     ad        110: kmutex_t       *softnet_lock;
1.49      jonathan  111:
1.64      thorpej   112: #ifdef SOSEND_COUNTERS
                    113: #include <sys/device.h>
                    114:
1.113     thorpej   115: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   116:     NULL, "sosend", "loan big");
1.113     thorpej   117: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   118:     NULL, "sosend", "copy big");
1.113     thorpej   119: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   120:     NULL, "sosend", "copy small");
1.113     thorpej   121: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   122:     NULL, "sosend", "kva limit");
                    123:
                    124: #define        SOSEND_COUNTER_INCR(ev)         (ev)->ev_count++
                    125:
1.101     matt      126: EVCNT_ATTACH_STATIC(sosend_loan_big);
                    127: EVCNT_ATTACH_STATIC(sosend_copy_big);
                    128: EVCNT_ATTACH_STATIC(sosend_copy_small);
                    129: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64      thorpej   130: #else
                    131:
                    132: #define        SOSEND_COUNTER_INCR(ev)         /* nothing */
                    133:
                    134: #endif /* SOSEND_COUNTERS */
                    135:
1.119     yamt      136: static struct callback_entry sokva_reclaimerentry;
1.1       cgd       137:
1.167     ad        138: #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR)
1.121     yamt      139: int sock_loan_thresh = -1;
1.71      thorpej   140: #else
1.121     yamt      141: int sock_loan_thresh = 4096;
1.65      thorpej   142: #endif
1.64      thorpej   143:
1.136     ad        144: static kmutex_t so_pendfree_lock;
1.113     thorpej   145: static struct mbuf *so_pendfree;
1.64      thorpej   146:
1.84      ragge     147: #ifndef SOMAXKVA
                    148: #define        SOMAXKVA (16 * 1024 * 1024)
                    149: #endif
                    150: int somaxkva = SOMAXKVA;
1.113     thorpej   151: static int socurkva;
1.136     ad        152: static kcondvar_t socurkva_cv;
1.64      thorpej   153:
1.191   ! elad      154: static kauth_listener_t socket_listener;
        !           155:
1.64      thorpej   156: #define        SOCK_LOAN_CHUNK         65536
                    157:
1.117     yamt      158: static size_t sodopendfree(void);
                    159: static size_t sodopendfreel(void);
1.93      yamt      160:
1.178     pooka     161: static void sysctl_kern_somaxkva_setup(void);
                    162: static struct sysctllog *socket_sysctllog;
                    163:
1.113     thorpej   164: static vsize_t
1.129     yamt      165: sokvareserve(struct socket *so, vsize_t len)
1.80      yamt      166: {
1.98      christos  167:        int error;
1.80      yamt      168:
1.136     ad        169:        mutex_enter(&so_pendfree_lock);
1.80      yamt      170:        while (socurkva + len > somaxkva) {
1.93      yamt      171:                size_t freed;
                    172:
                    173:                /*
                    174:                 * try to do pendfree.
                    175:                 */
                    176:
1.117     yamt      177:                freed = sodopendfreel();
1.93      yamt      178:
                    179:                /*
                    180:                 * if some kva was freed, try again.
                    181:                 */
                    182:
                    183:                if (freed)
1.80      yamt      184:                        continue;
1.93      yamt      185:
1.80      yamt      186:                SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136     ad        187:                error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98      christos  188:                if (error) {
                    189:                        len = 0;
                    190:                        break;
                    191:                }
1.80      yamt      192:        }
1.93      yamt      193:        socurkva += len;
1.136     ad        194:        mutex_exit(&so_pendfree_lock);
1.98      christos  195:        return len;
1.95      yamt      196: }
                    197:
1.113     thorpej   198: static void
1.95      yamt      199: sokvaunreserve(vsize_t len)
                    200: {
                    201:
1.136     ad        202:        mutex_enter(&so_pendfree_lock);
1.95      yamt      203:        socurkva -= len;
1.136     ad        204:        cv_broadcast(&socurkva_cv);
                    205:        mutex_exit(&so_pendfree_lock);
1.95      yamt      206: }
                    207:
                    208: /*
                    209:  * sokvaalloc: allocate kva for loan.
                    210:  */
                    211:
                    212: vaddr_t
                    213: sokvaalloc(vsize_t len, struct socket *so)
                    214: {
                    215:        vaddr_t lva;
                    216:
                    217:        /*
                    218:         * reserve kva.
                    219:         */
                    220:
1.98      christos  221:        if (sokvareserve(so, len) == 0)
                    222:                return 0;
1.93      yamt      223:
                    224:        /*
                    225:         * allocate kva.
                    226:         */
1.80      yamt      227:
1.109     yamt      228:        lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95      yamt      229:        if (lva == 0) {
                    230:                sokvaunreserve(len);
1.80      yamt      231:                return (0);
1.95      yamt      232:        }
1.80      yamt      233:
                    234:        return lva;
                    235: }
                    236:
1.93      yamt      237: /*
                    238:  * sokvafree: free kva for loan.
                    239:  */
                    240:
1.80      yamt      241: void
                    242: sokvafree(vaddr_t sva, vsize_t len)
                    243: {
1.93      yamt      244:
                    245:        /*
                    246:         * free kva.
                    247:         */
1.80      yamt      248:
1.109     yamt      249:        uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93      yamt      250:
                    251:        /*
                    252:         * unreserve kva.
                    253:         */
                    254:
1.95      yamt      255:        sokvaunreserve(len);
1.80      yamt      256: }
                    257:
1.64      thorpej   258: static void
1.134     christos  259: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64      thorpej   260: {
1.156     yamt      261:        vaddr_t sva, eva;
1.64      thorpej   262:        vsize_t len;
1.156     yamt      263:        int npgs;
                    264:
                    265:        KASSERT(pgs != NULL);
1.64      thorpej   266:
                    267:        eva = round_page((vaddr_t) buf + size);
                    268:        sva = trunc_page((vaddr_t) buf);
                    269:        len = eva - sva;
                    270:        npgs = len >> PAGE_SHIFT;
                    271:
                    272:        pmap_kremove(sva, len);
                    273:        pmap_update(pmap_kernel());
                    274:        uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80      yamt      275:        sokvafree(sva, len);
1.64      thorpej   276: }
                    277:
                    278: static size_t
1.152     matt      279: sodopendfree(void)
1.64      thorpej   280: {
1.93      yamt      281:        size_t rv;
1.64      thorpej   282:
1.160     ad        283:        if (__predict_true(so_pendfree == NULL))
                    284:                return 0;
                    285:
1.136     ad        286:        mutex_enter(&so_pendfree_lock);
1.117     yamt      287:        rv = sodopendfreel();
1.136     ad        288:        mutex_exit(&so_pendfree_lock);
1.93      yamt      289:
                    290:        return rv;
                    291: }
                    292:
                    293: /*
                    294:  * sodopendfreel: free mbufs on "pendfree" list.
1.136     ad        295:  * unlock and relock so_pendfree_lock when freeing mbufs.
1.93      yamt      296:  *
1.136     ad        297:  * => called with so_pendfree_lock held.
1.93      yamt      298:  */
                    299:
                    300: static size_t
1.152     matt      301: sodopendfreel(void)
1.93      yamt      302: {
1.137     ad        303:        struct mbuf *m, *next;
1.93      yamt      304:        size_t rv = 0;
                    305:
1.136     ad        306:        KASSERT(mutex_owned(&so_pendfree_lock));
1.64      thorpej   307:
1.137     ad        308:        while (so_pendfree != NULL) {
1.64      thorpej   309:                m = so_pendfree;
1.93      yamt      310:                so_pendfree = NULL;
1.136     ad        311:                mutex_exit(&so_pendfree_lock);
1.93      yamt      312:
                    313:                for (; m != NULL; m = next) {
                    314:                        next = m->m_next;
1.156     yamt      315:                        KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
                    316:                        KASSERT(m->m_ext.ext_refcnt == 0);
1.93      yamt      317:
                    318:                        rv += m->m_ext.ext_size;
1.156     yamt      319:                        sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
1.93      yamt      320:                            m->m_ext.ext_size);
1.145     ad        321:                        pool_cache_put(mb_cache, m);
1.93      yamt      322:                }
1.64      thorpej   323:
1.136     ad        324:                mutex_enter(&so_pendfree_lock);
1.64      thorpej   325:        }
                    326:
                    327:        return (rv);
                    328: }
                    329:
1.80      yamt      330: void
1.134     christos  331: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64      thorpej   332: {
                    333:
1.156     yamt      334:        KASSERT(m != NULL);
1.64      thorpej   335:
1.93      yamt      336:        /*
                    337:         * postpone freeing mbuf.
                    338:         *
                    339:         * we can't do it in interrupt context
                    340:         * because we need to put kva back to kernel_map.
                    341:         */
                    342:
1.136     ad        343:        mutex_enter(&so_pendfree_lock);
1.92      yamt      344:        m->m_next = so_pendfree;
                    345:        so_pendfree = m;
1.136     ad        346:        cv_broadcast(&socurkva_cv);
                    347:        mutex_exit(&so_pendfree_lock);
1.64      thorpej   348: }
                    349:
                    350: static long
                    351: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
                    352: {
                    353:        struct iovec *iov = uio->uio_iov;
                    354:        vaddr_t sva, eva;
                    355:        vsize_t len;
1.156     yamt      356:        vaddr_t lva;
                    357:        int npgs, error;
                    358:        vaddr_t va;
                    359:        int i;
1.64      thorpej   360:
1.116     yamt      361:        if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64      thorpej   362:                return (0);
                    363:
                    364:        if (iov->iov_len < (size_t) space)
                    365:                space = iov->iov_len;
                    366:        if (space > SOCK_LOAN_CHUNK)
                    367:                space = SOCK_LOAN_CHUNK;
                    368:
                    369:        eva = round_page((vaddr_t) iov->iov_base + space);
                    370:        sva = trunc_page((vaddr_t) iov->iov_base);
                    371:        len = eva - sva;
                    372:        npgs = len >> PAGE_SHIFT;
                    373:
1.79      thorpej   374:        KASSERT(npgs <= M_EXT_MAXPAGES);
                    375:
1.80      yamt      376:        lva = sokvaalloc(len, so);
1.64      thorpej   377:        if (lva == 0)
1.80      yamt      378:                return 0;
1.64      thorpej   379:
1.116     yamt      380:        error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79      thorpej   381:            m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64      thorpej   382:        if (error) {
1.80      yamt      383:                sokvafree(lva, len);
1.64      thorpej   384:                return (0);
                    385:        }
                    386:
                    387:        for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79      thorpej   388:                pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
                    389:                    VM_PROT_READ);
1.64      thorpej   390:        pmap_update(pmap_kernel());
                    391:
                    392:        lva += (vaddr_t) iov->iov_base & PAGE_MASK;
                    393:
1.134     christos  394:        MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79      thorpej   395:        m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64      thorpej   396:
                    397:        uio->uio_resid -= space;
                    398:        /* uio_offset not updated, not set/used for write(2) */
1.134     christos  399:        uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64      thorpej   400:        uio->uio_iov->iov_len -= space;
                    401:        if (uio->uio_iov->iov_len == 0) {
                    402:                uio->uio_iov++;
                    403:                uio->uio_iovcnt--;
                    404:        }
                    405:
                    406:        return (space);
                    407: }
                    408:
1.119     yamt      409: static int
1.129     yamt      410: sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg)
1.119     yamt      411: {
                    412:
                    413:        KASSERT(ce == &sokva_reclaimerentry);
                    414:        KASSERT(obj == NULL);
                    415:
                    416:        sodopendfree();
                    417:        if (!vm_map_starved_p(kernel_map)) {
                    418:                return CALLBACK_CHAIN_ABORT;
                    419:        }
                    420:        return CALLBACK_CHAIN_CONTINUE;
                    421: }
                    422:
1.142     dyoung    423: struct mbuf *
1.147     dyoung    424: getsombuf(struct socket *so, int type)
1.142     dyoung    425: {
                    426:        struct mbuf *m;
                    427:
1.147     dyoung    428:        m = m_get(M_WAIT, type);
1.142     dyoung    429:        MCLAIM(m, so->so_mowner);
                    430:        return m;
                    431: }
                    432:
1.191   ! elad      433: static int
        !           434: socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
        !           435:     void *arg0, void *arg1, void *arg2, void *arg3)
        !           436: {
        !           437:        int result;
        !           438:        enum kauth_network_req req;
        !           439:
        !           440:        result = KAUTH_RESULT_DEFER;
        !           441:        req = (enum kauth_network_req)arg0;
        !           442:
        !           443:        if (action != KAUTH_NETWORK_SOCKET)
        !           444:                return result;
        !           445:
        !           446:        switch (req) {
        !           447:        case KAUTH_REQ_NETWORK_SOCKET_DROP: {
        !           448:                /* Normal users can only drop their own connections. */
        !           449:                struct socket *so = (struct socket *)arg1;
        !           450:                uid_t sockuid = so->so_uidinfo->ui_uid;
        !           451:
        !           452:                if (sockuid == kauth_cred_getuid(cred) ||
        !           453:                    sockuid == kauth_cred_geteuid(cred))
        !           454:                        result = KAUTH_RESULT_ALLOW;
        !           455:
        !           456:                break;
        !           457:                }
        !           458:
        !           459:        case KAUTH_REQ_NETWORK_SOCKET_OPEN:
        !           460:                /* We allow "raw" routing/bluetooth sockets to anyone. */
        !           461:                if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_BLUETOOTH)
        !           462:                        result = KAUTH_RESULT_ALLOW;
        !           463:                else {
        !           464:                        /* Privileged, let secmodel handle this. */
        !           465:                        if ((u_long)arg2 == SOCK_RAW)
        !           466:                                break;
        !           467:                }
        !           468:
        !           469:                result = KAUTH_RESULT_ALLOW;
        !           470:
        !           471:                break;
        !           472:
        !           473:        default:
        !           474:                break;
        !           475:        }
        !           476:
        !           477:        return result;
        !           478: }
        !           479:
1.119     yamt      480: void
                    481: soinit(void)
                    482: {
                    483:
1.178     pooka     484:        sysctl_kern_somaxkva_setup();
                    485:
1.148     ad        486:        mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.160     ad        487:        softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.136     ad        488:        cv_init(&socurkva_cv, "sokva");
1.166     ad        489:        soinit2();
1.136     ad        490:
1.119     yamt      491:        /* Set the initial adjusted socket buffer size. */
                    492:        if (sb_max_set(sb_max))
                    493:                panic("bad initial sb_max value: %lu", sb_max);
                    494:
                    495:        callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback,
                    496:            &sokva_reclaimerentry, NULL, sokva_reclaim_callback);
1.191   ! elad      497:
        !           498:        socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
        !           499:            socket_listener_cb, NULL);
1.119     yamt      500: }
                    501:
1.1       cgd       502: /*
                    503:  * Socket operation routines.
                    504:  * These routines are called by the routines in
                    505:  * sys_socket.c or from a system process, and
                    506:  * implement the semantics of socket operations by
                    507:  * switching out to the protocol specific routines.
                    508:  */
                    509: /*ARGSUSED*/
1.3       andrew    510: int
1.160     ad        511: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l,
                    512:         struct socket *lockso)
1.1       cgd       513: {
1.99      matt      514:        const struct protosw    *prp;
1.54      lukem     515:        struct socket   *so;
1.115     yamt      516:        uid_t           uid;
1.160     ad        517:        int             error;
                    518:        kmutex_t        *lock;
1.1       cgd       519:
1.132     elad      520:        error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
                    521:            KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
                    522:            KAUTH_ARG(proto));
1.140     dyoung    523:        if (error != 0)
                    524:                return error;
1.127     elad      525:
1.1       cgd       526:        if (proto)
                    527:                prp = pffindproto(dom, proto, type);
                    528:        else
                    529:                prp = pffindtype(dom, type);
1.140     dyoung    530:        if (prp == NULL) {
1.120     ginsbach  531:                /* no support for domain */
                    532:                if (pffinddomain(dom) == 0)
1.140     dyoung    533:                        return EAFNOSUPPORT;
1.120     ginsbach  534:                /* no support for socket type */
                    535:                if (proto == 0 && type != 0)
1.140     dyoung    536:                        return EPROTOTYPE;
                    537:                return EPROTONOSUPPORT;
1.120     ginsbach  538:        }
1.140     dyoung    539:        if (prp->pr_usrreq == NULL)
                    540:                return EPROTONOSUPPORT;
1.1       cgd       541:        if (prp->pr_type != type)
1.140     dyoung    542:                return EPROTOTYPE;
1.160     ad        543:
                    544:        so = soget(true);
1.1       cgd       545:        so->so_type = type;
                    546:        so->so_proto = prp;
1.33      matt      547:        so->so_send = sosend;
                    548:        so->so_receive = soreceive;
1.78      matt      549: #ifdef MBUFTRACE
                    550:        so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
                    551:        so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
                    552:        so->so_mowner = &prp->pr_domain->dom_mowner;
                    553: #endif
1.191   ! elad      554:        /* so->so_cred = kauth_cred_dup(l->l_cred); */
1.138     rmind     555:        uid = kauth_cred_geteuid(l->l_cred);
1.115     yamt      556:        so->so_uidinfo = uid_find(uid);
1.168     yamt      557:        so->so_egid = kauth_cred_getegid(l->l_cred);
                    558:        so->so_cpid = l->l_proc->p_pid;
1.160     ad        559:        if (lockso != NULL) {
                    560:                /* Caller wants us to share a lock. */
                    561:                lock = lockso->so_lock;
                    562:                so->so_lock = lock;
                    563:                mutex_obj_hold(lock);
                    564:                mutex_enter(lock);
                    565:        } else {
                    566:                /* Lock assigned and taken during PRU_ATTACH. */
                    567:        }
1.140     dyoung    568:        error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
                    569:            (struct mbuf *)(long)proto, NULL, l);
1.160     ad        570:        KASSERT(solocked(so));
1.140     dyoung    571:        if (error != 0) {
1.1       cgd       572:                so->so_state |= SS_NOFDREF;
                    573:                sofree(so);
1.140     dyoung    574:                return error;
1.1       cgd       575:        }
1.160     ad        576:        sounlock(so);
1.1       cgd       577:        *aso = so;
1.140     dyoung    578:        return 0;
1.1       cgd       579: }
                    580:
1.142     dyoung    581: /* On success, write file descriptor to fdout and return zero.  On
                    582:  * failure, return non-zero; *fdout will be undefined.
                    583:  */
                    584: int
                    585: fsocreate(int domain, struct socket **sop, int type, int protocol,
                    586:     struct lwp *l, int *fdout)
                    587: {
                    588:        struct socket   *so;
                    589:        struct file     *fp;
                    590:        int             fd, error;
                    591:
1.155     ad        592:        if ((error = fd_allocfile(&fp, &fd)) != 0)
1.142     dyoung    593:                return (error);
                    594:        fp->f_flag = FREAD|FWRITE;
                    595:        fp->f_type = DTYPE_SOCKET;
                    596:        fp->f_ops = &socketops;
1.160     ad        597:        error = socreate(domain, &so, type, protocol, l, NULL);
1.142     dyoung    598:        if (error != 0) {
1.155     ad        599:                fd_abort(curproc, fp, fd);
1.142     dyoung    600:        } else {
                    601:                if (sop != NULL)
                    602:                        *sop = so;
                    603:                fp->f_data = so;
1.155     ad        604:                fd_affix(curproc, fp, fd);
1.142     dyoung    605:                *fdout = fd;
                    606:        }
                    607:        return error;
                    608: }
                    609:
1.3       andrew    610: int
1.190     dyoung    611: sofamily(const struct socket *so)
                    612: {
                    613:        const struct protosw *pr;
                    614:        const struct domain *dom;
                    615:
                    616:        if ((pr = so->so_proto) == NULL)
                    617:                return AF_UNSPEC;
                    618:        if ((dom = pr->pr_domain) == NULL)
                    619:                return AF_UNSPEC;
                    620:        return dom->dom_family;
                    621: }
                    622:
                    623: int
1.114     christos  624: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       625: {
1.160     ad        626:        int     error;
1.1       cgd       627:
1.160     ad        628:        solock(so);
1.140     dyoung    629:        error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.160     ad        630:        sounlock(so);
1.140     dyoung    631:        return error;
1.1       cgd       632: }
                    633:
1.3       andrew    634: int
1.150     elad      635: solisten(struct socket *so, int backlog, struct lwp *l)
1.1       cgd       636: {
1.160     ad        637:        int     error;
1.1       cgd       638:
1.160     ad        639:        solock(so);
1.158     ad        640:        if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
1.163     ad        641:            SS_ISDISCONNECTING)) != 0) {
                    642:                sounlock(so);
1.158     ad        643:                return (EOPNOTSUPP);
1.163     ad        644:        }
1.140     dyoung    645:        error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
1.150     elad      646:            NULL, NULL, l);
1.140     dyoung    647:        if (error != 0) {
1.160     ad        648:                sounlock(so);
1.140     dyoung    649:                return error;
1.1       cgd       650:        }
1.63      matt      651:        if (TAILQ_EMPTY(&so->so_q))
1.1       cgd       652:                so->so_options |= SO_ACCEPTCONN;
                    653:        if (backlog < 0)
                    654:                backlog = 0;
1.49      jonathan  655:        so->so_qlimit = min(backlog, somaxconn);
1.160     ad        656:        sounlock(so);
1.140     dyoung    657:        return 0;
1.1       cgd       658: }
                    659:
1.21      christos  660: void
1.54      lukem     661: sofree(struct socket *so)
1.1       cgd       662: {
1.161     ad        663:        u_int refs;
1.1       cgd       664:
1.160     ad        665:        KASSERT(solocked(so));
                    666:
                    667:        if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) {
                    668:                sounlock(so);
1.1       cgd       669:                return;
1.160     ad        670:        }
1.43      mycroft   671:        if (so->so_head) {
                    672:                /*
                    673:                 * We must not decommission a socket that's on the accept(2)
                    674:                 * queue.  If we do, then accept(2) may hang after select(2)
                    675:                 * indicated that the listening socket was ready.
                    676:                 */
1.160     ad        677:                if (!soqremque(so, 0)) {
                    678:                        sounlock(so);
1.43      mycroft   679:                        return;
1.160     ad        680:                }
1.43      mycroft   681:        }
1.98      christos  682:        if (so->so_rcv.sb_hiwat)
1.110     christos  683:                (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98      christos  684:                    RLIM_INFINITY);
                    685:        if (so->so_snd.sb_hiwat)
1.110     christos  686:                (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98      christos  687:                    RLIM_INFINITY);
                    688:        sbrelease(&so->so_snd, so);
1.160     ad        689:        KASSERT(!cv_has_waiters(&so->so_cv));
                    690:        KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv));
                    691:        KASSERT(!cv_has_waiters(&so->so_snd.sb_cv));
1.1       cgd       692:        sorflush(so);
1.161     ad        693:        refs = so->so_aborting; /* XXX */
1.177     ad        694:        /* Remove acccept filter if one is present. */
1.170     tls       695:        if (so->so_accf != NULL)
1.177     ad        696:                (void)accept_filt_clear(so);
1.191   ! elad      697:        /* kauth_cred_free(so->so_cred); */
1.160     ad        698:        sounlock(so);
1.161     ad        699:        if (refs == 0)          /* XXX */
                    700:                soput(so);
1.1       cgd       701: }
                    702:
                    703: /*
                    704:  * Close a socket on last file table reference removal.
                    705:  * Initiate disconnect if connected.
                    706:  * Free socket when disconnect complete.
                    707:  */
1.3       andrew    708: int
1.54      lukem     709: soclose(struct socket *so)
1.1       cgd       710: {
1.54      lukem     711:        struct socket   *so2;
1.160     ad        712:        int             error;
                    713:        int             error2;
1.1       cgd       714:
1.54      lukem     715:        error = 0;
1.160     ad        716:        solock(so);
1.1       cgd       717:        if (so->so_options & SO_ACCEPTCONN) {
1.172     ad        718:                for (;;) {
                    719:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.160     ad        720:                                KASSERT(solocked2(so, so2));
                    721:                                (void) soqremque(so2, 0);
                    722:                                /* soabort drops the lock. */
                    723:                                (void) soabort(so2);
                    724:                                solock(so);
1.172     ad        725:                                continue;
1.160     ad        726:                        }
1.172     ad        727:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.160     ad        728:                                KASSERT(solocked2(so, so2));
                    729:                                (void) soqremque(so2, 1);
                    730:                                /* soabort drops the lock. */
                    731:                                (void) soabort(so2);
                    732:                                solock(so);
1.172     ad        733:                                continue;
1.160     ad        734:                        }
1.172     ad        735:                        break;
                    736:                }
1.1       cgd       737:        }
                    738:        if (so->so_pcb == 0)
                    739:                goto discard;
                    740:        if (so->so_state & SS_ISCONNECTED) {
                    741:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    742:                        error = sodisconnect(so);
                    743:                        if (error)
                    744:                                goto drop;
                    745:                }
                    746:                if (so->so_options & SO_LINGER) {
1.151     ad        747:                        if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio)
1.1       cgd       748:                                goto drop;
1.21      christos  749:                        while (so->so_state & SS_ISCONNECTED) {
1.185     yamt      750:                                error = sowait(so, true, so->so_linger * hz);
1.21      christos  751:                                if (error)
1.1       cgd       752:                                        break;
1.21      christos  753:                        }
1.1       cgd       754:                }
                    755:        }
1.54      lukem     756:  drop:
1.1       cgd       757:        if (so->so_pcb) {
1.160     ad        758:                error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140     dyoung    759:                    NULL, NULL, NULL, NULL);
1.1       cgd       760:                if (error == 0)
                    761:                        error = error2;
                    762:        }
1.54      lukem     763:  discard:
1.1       cgd       764:        if (so->so_state & SS_NOFDREF)
                    765:                panic("soclose: NOFDREF");
                    766:        so->so_state |= SS_NOFDREF;
                    767:        sofree(so);
                    768:        return (error);
                    769: }
                    770:
                    771: /*
1.160     ad        772:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       773:  */
1.3       andrew    774: int
1.54      lukem     775: soabort(struct socket *so)
1.1       cgd       776: {
1.161     ad        777:        u_int refs;
1.139     yamt      778:        int error;
1.160     ad        779:
                    780:        KASSERT(solocked(so));
                    781:        KASSERT(so->so_head == NULL);
1.1       cgd       782:
1.161     ad        783:        so->so_aborting++;              /* XXX */
1.140     dyoung    784:        error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
                    785:            NULL, NULL, NULL);
1.161     ad        786:        refs = --so->so_aborting;       /* XXX */
1.164     drochner  787:        if (error || (refs == 0)) {
1.139     yamt      788:                sofree(so);
1.160     ad        789:        } else {
                    790:                sounlock(so);
1.139     yamt      791:        }
                    792:        return error;
1.1       cgd       793: }
                    794:
1.3       andrew    795: int
1.54      lukem     796: soaccept(struct socket *so, struct mbuf *nam)
1.1       cgd       797: {
1.160     ad        798:        int     error;
                    799:
                    800:        KASSERT(solocked(so));
1.1       cgd       801:
1.54      lukem     802:        error = 0;
1.1       cgd       803:        if ((so->so_state & SS_NOFDREF) == 0)
                    804:                panic("soaccept: !NOFDREF");
                    805:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   806:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    807:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41      mycroft   808:                error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140     dyoung    809:                    NULL, nam, NULL, NULL);
1.41      mycroft   810:        else
1.53      itojun    811:                error = ECONNABORTED;
1.52      itojun    812:
1.1       cgd       813:        return (error);
                    814: }
                    815:
1.3       andrew    816: int
1.114     christos  817: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       818: {
1.160     ad        819:        int             error;
                    820:
                    821:        KASSERT(solocked(so));
1.1       cgd       822:
                    823:        if (so->so_options & SO_ACCEPTCONN)
                    824:                return (EOPNOTSUPP);
                    825:        /*
                    826:         * If protocol is connection-based, can only connect once.
                    827:         * Otherwise, if connected, try to disconnect first.
                    828:         * This allows user to disconnect by connecting to, e.g.,
                    829:         * a null address.
                    830:         */
                    831:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    832:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
                    833:            (error = sodisconnect(so))))
                    834:                error = EISCONN;
                    835:        else
                    836:                error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140     dyoung    837:                    NULL, nam, NULL, l);
1.1       cgd       838:        return (error);
                    839: }
                    840:
1.3       andrew    841: int
1.54      lukem     842: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       843: {
1.160     ad        844:        int     error;
                    845:
                    846:        KASSERT(solocked2(so1, so2));
1.1       cgd       847:
1.22      mycroft   848:        error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140     dyoung    849:            NULL, (struct mbuf *)so2, NULL, NULL);
1.1       cgd       850:        return (error);
                    851: }
                    852:
1.3       andrew    853: int
1.54      lukem     854: sodisconnect(struct socket *so)
1.1       cgd       855: {
1.160     ad        856:        int     error;
                    857:
                    858:        KASSERT(solocked(so));
1.1       cgd       859:
                    860:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    861:                error = ENOTCONN;
1.160     ad        862:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       863:                error = EALREADY;
1.160     ad        864:        } else {
                    865:                error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
                    866:                    NULL, NULL, NULL, NULL);
1.1       cgd       867:        }
1.117     yamt      868:        sodopendfree();
1.1       cgd       869:        return (error);
                    870: }
                    871:
1.15      mycroft   872: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       873: /*
                    874:  * Send on a socket.
                    875:  * If send must go all at once and message is larger than
                    876:  * send buffering, then hard error.
                    877:  * Lock against other senders.
                    878:  * If must go all at once and not enough room now, then
                    879:  * inform user that this would block and do nothing.
                    880:  * Otherwise, if nonblocking, send as much as possible.
                    881:  * The data to be sent is described by "uio" if nonzero,
                    882:  * otherwise by the mbuf chain "top" (which must be null
                    883:  * if uio is not).  Data provided in mbuf chain must be small
                    884:  * enough to send all at once.
                    885:  *
                    886:  * Returns nonzero on error, timeout or signal; callers
                    887:  * must check for short counts if EINTR/ERESTART are returned.
                    888:  * Data and control buffers are freed on return.
                    889:  */
1.3       andrew    890: int
1.54      lukem     891: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114     christos  892:        struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       893: {
1.54      lukem     894:        struct mbuf     **mp, *m;
1.114     christos  895:        struct proc     *p;
1.58      jdolecek  896:        long            space, len, resid, clen, mlen;
                    897:        int             error, s, dontroute, atomic;
1.54      lukem     898:
1.114     christos  899:        p = l->l_proc;
1.117     yamt      900:        sodopendfree();
1.160     ad        901:        clen = 0;
1.64      thorpej   902:
1.160     ad        903:        /*
                    904:         * solock() provides atomicity of access.  splsoftnet() prevents
                    905:         * protocol processing soft interrupts from interrupting us and
                    906:         * blocking (expensive).
                    907:         */
                    908:        s = splsoftnet();
                    909:        solock(so);
1.54      lukem     910:        atomic = sosendallatonce(so) || top;
1.1       cgd       911:        if (uio)
                    912:                resid = uio->uio_resid;
                    913:        else
                    914:                resid = top->m_pkthdr.len;
1.7       cgd       915:        /*
                    916:         * In theory resid should be unsigned.
                    917:         * However, space must be signed, as it might be less than 0
                    918:         * if we over-committed, and we must use a signed comparison
                    919:         * of space and resid.  On the other hand, a negative resid
                    920:         * causes us to loop sending 0-length segments to the protocol.
                    921:         */
1.29      mycroft   922:        if (resid < 0) {
                    923:                error = EINVAL;
                    924:                goto out;
                    925:        }
1.1       cgd       926:        dontroute =
                    927:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    928:            (so->so_proto->pr_flags & PR_ATOMIC);
1.165     christos  929:        l->l_ru.ru_msgsnd++;
1.1       cgd       930:        if (control)
                    931:                clen = control->m_len;
1.54      lukem     932:  restart:
1.21      christos  933:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       934:                goto out;
                    935:        do {
1.160     ad        936:                if (so->so_state & SS_CANTSENDMORE) {
                    937:                        error = EPIPE;
                    938:                        goto release;
                    939:                }
1.48      thorpej   940:                if (so->so_error) {
                    941:                        error = so->so_error;
                    942:                        so->so_error = 0;
                    943:                        goto release;
                    944:                }
1.1       cgd       945:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    946:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
                    947:                                if ((so->so_state & SS_ISCONFIRMING) == 0 &&
1.160     ad        948:                                    !(resid == 0 && clen != 0)) {
                    949:                                        error = ENOTCONN;
                    950:                                        goto release;
                    951:                                }
                    952:                        } else if (addr == 0) {
                    953:                                error = EDESTADDRREQ;
                    954:                                goto release;
                    955:                        }
1.1       cgd       956:                }
                    957:                space = sbspace(&so->so_snd);
                    958:                if (flags & MSG_OOB)
                    959:                        space += 1024;
1.21      christos  960:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        961:                    clen > so->so_snd.sb_hiwat) {
                    962:                        error = EMSGSIZE;
                    963:                        goto release;
                    964:                }
1.96      mycroft   965:                if (space < resid + clen &&
1.1       cgd       966:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.160     ad        967:                        if (so->so_nbio) {
                    968:                                error = EWOULDBLOCK;
                    969:                                goto release;
                    970:                        }
1.1       cgd       971:                        sbunlock(&so->so_snd);
                    972:                        error = sbwait(&so->so_snd);
                    973:                        if (error)
                    974:                                goto out;
                    975:                        goto restart;
                    976:                }
                    977:                mp = &top;
                    978:                space -= clen;
                    979:                do {
1.45      tv        980:                        if (uio == NULL) {
                    981:                                /*
                    982:                                 * Data is prepackaged in "top".
                    983:                                 */
                    984:                                resid = 0;
                    985:                                if (flags & MSG_EOR)
                    986:                                        top->m_flags |= M_EOR;
                    987:                        } else do {
1.160     ad        988:                                sounlock(so);
                    989:                                splx(s);
1.144     dyoung    990:                                if (top == NULL) {
1.78      matt      991:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv        992:                                        mlen = MHLEN;
                    993:                                        m->m_pkthdr.len = 0;
1.140     dyoung    994:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv        995:                                } else {
1.78      matt      996:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv        997:                                        mlen = MLEN;
                    998:                                }
1.78      matt      999:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt     1000:                                if (sock_loan_thresh >= 0 &&
                   1001:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                   1002:                                    space >= sock_loan_thresh &&
1.64      thorpej  1003:                                    (len = sosend_loan(so, uio, m,
                   1004:                                                       space)) != 0) {
                   1005:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                   1006:                                        space -= len;
                   1007:                                        goto have_data;
                   1008:                                }
1.45      tv       1009:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej  1010:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78      matt     1011:                                        m_clget(m, M_WAIT);
1.45      tv       1012:                                        if ((m->m_flags & M_EXT) == 0)
                   1013:                                                goto nopages;
                   1014:                                        mlen = MCLBYTES;
                   1015:                                        if (atomic && top == 0) {
1.58      jdolecek 1016:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem    1017:                                                    resid);
1.45      tv       1018:                                                m->m_data += max_hdr;
                   1019:                                        } else
1.58      jdolecek 1020:                                                len = lmin(MCLBYTES, resid);
1.45      tv       1021:                                        space -= len;
                   1022:                                } else {
1.64      thorpej  1023:  nopages:
                   1024:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek 1025:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv       1026:                                        space -= len;
                   1027:                                        /*
                   1028:                                         * For datagram protocols, leave room
                   1029:                                         * for protocol headers in first mbuf.
                   1030:                                         */
                   1031:                                        if (atomic && top == 0 && len < mlen)
                   1032:                                                MH_ALIGN(m, len);
                   1033:                                }
1.144     dyoung   1034:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej  1035:  have_data:
1.45      tv       1036:                                resid = uio->uio_resid;
                   1037:                                m->m_len = len;
                   1038:                                *mp = m;
                   1039:                                top->m_pkthdr.len += len;
1.160     ad       1040:                                s = splsoftnet();
                   1041:                                solock(so);
1.144     dyoung   1042:                                if (error != 0)
1.45      tv       1043:                                        goto release;
                   1044:                                mp = &m->m_next;
                   1045:                                if (resid <= 0) {
                   1046:                                        if (flags & MSG_EOR)
                   1047:                                                top->m_flags |= M_EOR;
                   1048:                                        break;
                   1049:                                }
                   1050:                        } while (space > 0 && atomic);
1.108     perry    1051:
1.160     ad       1052:                        if (so->so_state & SS_CANTSENDMORE) {
                   1053:                                error = EPIPE;
                   1054:                                goto release;
                   1055:                        }
1.45      tv       1056:                        if (dontroute)
                   1057:                                so->so_options |= SO_DONTROUTE;
                   1058:                        if (resid > 0)
                   1059:                                so->so_state |= SS_MORETOCOME;
1.46      sommerfe 1060:                        error = (*so->so_proto->pr_usrreq)(so,
                   1061:                            (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.160     ad       1062:                            top, addr, control, curlwp);
1.45      tv       1063:                        if (dontroute)
                   1064:                                so->so_options &= ~SO_DONTROUTE;
                   1065:                        if (resid > 0)
                   1066:                                so->so_state &= ~SS_MORETOCOME;
                   1067:                        clen = 0;
1.144     dyoung   1068:                        control = NULL;
                   1069:                        top = NULL;
1.45      tv       1070:                        mp = &top;
1.144     dyoung   1071:                        if (error != 0)
1.1       cgd      1072:                                goto release;
                   1073:                } while (resid && space > 0);
                   1074:        } while (resid);
                   1075:
1.54      lukem    1076:  release:
1.1       cgd      1077:        sbunlock(&so->so_snd);
1.54      lukem    1078:  out:
1.160     ad       1079:        sounlock(so);
                   1080:        splx(s);
1.1       cgd      1081:        if (top)
                   1082:                m_freem(top);
                   1083:        if (control)
                   1084:                m_freem(control);
                   1085:        return (error);
                   1086: }
                   1087:
                   1088: /*
1.159     ad       1089:  * Following replacement or removal of the first mbuf on the first
                   1090:  * mbuf chain of a socket buffer, push necessary state changes back
                   1091:  * into the socket buffer so that other consumers see the values
                   1092:  * consistently.  'nextrecord' is the callers locally stored value of
                   1093:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1094:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1095:  */
                   1096: static void
                   1097: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1098: {
                   1099:
1.160     ad       1100:        KASSERT(solocked(sb->sb_so));
                   1101:
1.159     ad       1102:        /*
                   1103:         * First, update for the new value of nextrecord.  If necessary,
                   1104:         * make it the first record.
                   1105:         */
                   1106:        if (sb->sb_mb != NULL)
                   1107:                sb->sb_mb->m_nextpkt = nextrecord;
                   1108:        else
                   1109:                sb->sb_mb = nextrecord;
                   1110:
                   1111:         /*
                   1112:          * Now update any dependent socket buffer fields to reflect
                   1113:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1114:          * the addition of a second clause that takes care of the
                   1115:          * case where sb_mb has been updated, but remains the last
                   1116:          * record.
                   1117:          */
                   1118:         if (sb->sb_mb == NULL) {
                   1119:                 sb->sb_mbtail = NULL;
                   1120:                 sb->sb_lastrecord = NULL;
                   1121:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1122:                 sb->sb_lastrecord = sb->sb_mb;
                   1123: }
                   1124:
                   1125: /*
1.1       cgd      1126:  * Implement receive operations on a socket.
                   1127:  * We depend on the way that records are added to the sockbuf
                   1128:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1129:  * must begin with an address if the protocol so specifies,
                   1130:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1131:  * and then zero or more mbufs of data.
                   1132:  * In order to avoid blocking network interrupts for the entire time here,
                   1133:  * we splx() while doing the actual copy to user space.
                   1134:  * Although the sockbuf is locked, new data may still be appended,
                   1135:  * and thus we must maintain consistency of the sockbuf during that time.
                   1136:  *
                   1137:  * The caller may receive the data as a single mbuf chain by supplying
                   1138:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1139:  * only for the count in uio_resid.
                   1140:  */
1.3       andrew   1141: int
1.54      lukem    1142: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1143:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1144: {
1.116     yamt     1145:        struct lwp *l = curlwp;
1.160     ad       1146:        struct mbuf     *m, **mp, *mt;
1.146     dyoung   1147:        int atomic, flags, len, error, s, offset, moff, type, orig_resid;
1.99      matt     1148:        const struct protosw    *pr;
1.54      lukem    1149:        struct mbuf     *nextrecord;
1.67      he       1150:        int             mbuf_removed = 0;
1.146     dyoung   1151:        const struct domain *dom;
1.64      thorpej  1152:
1.54      lukem    1153:        pr = so->so_proto;
1.146     dyoung   1154:        atomic = pr->pr_flags & PR_ATOMIC;
                   1155:        dom = pr->pr_domain;
1.1       cgd      1156:        mp = mp0;
1.54      lukem    1157:        type = 0;
                   1158:        orig_resid = uio->uio_resid;
1.102     jonathan 1159:
1.144     dyoung   1160:        if (paddr != NULL)
                   1161:                *paddr = NULL;
                   1162:        if (controlp != NULL)
                   1163:                *controlp = NULL;
                   1164:        if (flagsp != NULL)
1.1       cgd      1165:                flags = *flagsp &~ MSG_EOR;
                   1166:        else
                   1167:                flags = 0;
1.66      enami    1168:
                   1169:        if ((flags & MSG_DONTWAIT) == 0)
1.117     yamt     1170:                sodopendfree();
1.66      enami    1171:
1.1       cgd      1172:        if (flags & MSG_OOB) {
                   1173:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1174:                solock(so);
1.17      cgd      1175:                error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140     dyoung   1176:                    (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.160     ad       1177:                sounlock(so);
1.1       cgd      1178:                if (error)
                   1179:                        goto bad;
                   1180:                do {
1.134     christos 1181:                        error = uiomove(mtod(m, void *),
1.1       cgd      1182:                            (int) min(uio->uio_resid, m->m_len), uio);
                   1183:                        m = m_free(m);
1.144     dyoung   1184:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1185:  bad:
1.144     dyoung   1186:                if (m != NULL)
1.1       cgd      1187:                        m_freem(m);
1.144     dyoung   1188:                return error;
1.1       cgd      1189:        }
1.144     dyoung   1190:        if (mp != NULL)
1.140     dyoung   1191:                *mp = NULL;
1.160     ad       1192:
                   1193:        /*
                   1194:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1195:         * protocol processing soft interrupts from interrupting us and
                   1196:         * blocking (expensive).
                   1197:         */
                   1198:        s = splsoftnet();
                   1199:        solock(so);
1.1       cgd      1200:        if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140     dyoung   1201:                (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1       cgd      1202:
1.54      lukem    1203:  restart:
1.160     ad       1204:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1205:                sounlock(so);
                   1206:                splx(s);
1.144     dyoung   1207:                return error;
1.160     ad       1208:        }
1.1       cgd      1209:
                   1210:        m = so->so_rcv.sb_mb;
                   1211:        /*
                   1212:         * If we have less data than requested, block awaiting more
                   1213:         * (subject to any timeout) if:
1.15      mycroft  1214:         *   1. the current count is less than the low water mark,
1.1       cgd      1215:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1216:         *      receive operation at once if we block (resid <= hiwat), or
                   1217:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1218:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1219:         * we have to do the receive in sections, and thus risk returning
                   1220:         * a short count if a timeout or signal occurs after we start.
                   1221:         */
1.144     dyoung   1222:        if (m == NULL ||
                   1223:            ((flags & MSG_DONTWAIT) == 0 &&
                   1224:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1225:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1226:              ((flags & MSG_WAITALL) &&
                   1227:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1228:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1229: #ifdef DIAGNOSTIC
1.144     dyoung   1230:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1231:                        panic("receive 1");
                   1232: #endif
                   1233:                if (so->so_error) {
1.144     dyoung   1234:                        if (m != NULL)
1.15      mycroft  1235:                                goto dontblock;
1.1       cgd      1236:                        error = so->so_error;
                   1237:                        if ((flags & MSG_PEEK) == 0)
                   1238:                                so->so_error = 0;
                   1239:                        goto release;
                   1240:                }
                   1241:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1242:                        if (m != NULL)
1.15      mycroft  1243:                                goto dontblock;
1.1       cgd      1244:                        else
                   1245:                                goto release;
                   1246:                }
1.144     dyoung   1247:                for (; m != NULL; m = m->m_next)
1.1       cgd      1248:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1249:                                m = so->so_rcv.sb_mb;
                   1250:                                goto dontblock;
                   1251:                        }
                   1252:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1253:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1254:                        error = ENOTCONN;
                   1255:                        goto release;
                   1256:                }
                   1257:                if (uio->uio_resid == 0)
                   1258:                        goto release;
1.151     ad       1259:                if (so->so_nbio || (flags & MSG_DONTWAIT)) {
1.1       cgd      1260:                        error = EWOULDBLOCK;
                   1261:                        goto release;
                   1262:                }
1.69      thorpej  1263:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1264:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1265:                sbunlock(&so->so_rcv);
                   1266:                error = sbwait(&so->so_rcv);
1.160     ad       1267:                if (error != 0) {
                   1268:                        sounlock(so);
                   1269:                        splx(s);
1.144     dyoung   1270:                        return error;
1.160     ad       1271:                }
1.1       cgd      1272:                goto restart;
                   1273:        }
1.54      lukem    1274:  dontblock:
1.69      thorpej  1275:        /*
                   1276:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1277:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1278:         * pointer to the next record in the socket buffer.  We must keep the
                   1279:         * various socket buffer pointers and local stack versions of the
                   1280:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1281:         * socket lock, and re-reading them when picking it up.
1.159     ad       1282:         *
                   1283:         * Otherwise, we will race with the network stack appending new data
                   1284:         * or records onto the socket buffer by using inconsistent/stale
                   1285:         * versions of the field, possibly resulting in socket buffer
                   1286:         * corruption.
                   1287:         *
                   1288:         * By holding the high-level sblock(), we prevent simultaneous
                   1289:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1290:         */
1.144     dyoung   1291:        if (l != NULL)
1.157     ad       1292:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1293:        KASSERT(m == so->so_rcv.sb_mb);
                   1294:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1295:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1296:        nextrecord = m->m_nextpkt;
                   1297:        if (pr->pr_flags & PR_ADDR) {
                   1298: #ifdef DIAGNOSTIC
                   1299:                if (m->m_type != MT_SONAME)
                   1300:                        panic("receive 1a");
                   1301: #endif
1.3       andrew   1302:                orig_resid = 0;
1.1       cgd      1303:                if (flags & MSG_PEEK) {
                   1304:                        if (paddr)
                   1305:                                *paddr = m_copy(m, 0, m->m_len);
                   1306:                        m = m->m_next;
                   1307:                } else {
                   1308:                        sbfree(&so->so_rcv, m);
1.67      he       1309:                        mbuf_removed = 1;
1.144     dyoung   1310:                        if (paddr != NULL) {
1.1       cgd      1311:                                *paddr = m;
                   1312:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1313:                                m->m_next = NULL;
1.1       cgd      1314:                                m = so->so_rcv.sb_mb;
                   1315:                        } else {
                   1316:                                MFREE(m, so->so_rcv.sb_mb);
                   1317:                                m = so->so_rcv.sb_mb;
                   1318:                        }
1.159     ad       1319:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1320:                }
                   1321:        }
1.159     ad       1322:
                   1323:        /*
                   1324:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1325:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1326:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1327:         * perform externalization (or freeing if controlp == NULL).
                   1328:         */
                   1329:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1330:                struct mbuf *cm = NULL, *cmn;
                   1331:                struct mbuf **cme = &cm;
                   1332:
                   1333:                do {
                   1334:                        if (flags & MSG_PEEK) {
                   1335:                                if (controlp != NULL) {
                   1336:                                        *controlp = m_copy(m, 0, m->m_len);
                   1337:                                        controlp = &(*controlp)->m_next;
                   1338:                                }
                   1339:                                m = m->m_next;
                   1340:                        } else {
                   1341:                                sbfree(&so->so_rcv, m);
1.1       cgd      1342:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1343:                                m->m_next = NULL;
1.159     ad       1344:                                *cme = m;
                   1345:                                cme = &(*cme)->m_next;
1.1       cgd      1346:                                m = so->so_rcv.sb_mb;
1.159     ad       1347:                        }
                   1348:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1349:                if ((flags & MSG_PEEK) == 0)
                   1350:                        sbsync(&so->so_rcv, nextrecord);
                   1351:                for (; cm != NULL; cm = cmn) {
                   1352:                        cmn = cm->m_next;
                   1353:                        cm->m_next = NULL;
                   1354:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1355:                        if (controlp != NULL) {
                   1356:                                if (dom->dom_externalize != NULL &&
                   1357:                                    type == SCM_RIGHTS) {
1.160     ad       1358:                                        sounlock(so);
1.159     ad       1359:                                        splx(s);
                   1360:                                        error = (*dom->dom_externalize)(cm, l);
                   1361:                                        s = splsoftnet();
1.160     ad       1362:                                        solock(so);
1.159     ad       1363:                                }
                   1364:                                *controlp = cm;
                   1365:                                while (*controlp != NULL)
                   1366:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1367:                        } else {
1.106     itojun   1368:                                /*
                   1369:                                 * Dispose of any SCM_RIGHTS message that went
                   1370:                                 * through the read path rather than recv.
                   1371:                                 */
1.159     ad       1372:                                if (dom->dom_dispose != NULL &&
                   1373:                                    type == SCM_RIGHTS) {
1.160     ad       1374:                                        sounlock(so);
1.159     ad       1375:                                        (*dom->dom_dispose)(cm);
1.160     ad       1376:                                        solock(so);
1.159     ad       1377:                                }
                   1378:                                m_freem(cm);
1.1       cgd      1379:                        }
                   1380:                }
1.159     ad       1381:                if (m != NULL)
                   1382:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1383:                else
                   1384:                        nextrecord = so->so_rcv.sb_mb;
                   1385:                orig_resid = 0;
1.1       cgd      1386:        }
1.69      thorpej  1387:
1.159     ad       1388:        /* If m is non-NULL, we have some data to read. */
                   1389:        if (__predict_true(m != NULL)) {
1.1       cgd      1390:                type = m->m_type;
                   1391:                if (type == MT_OOBDATA)
                   1392:                        flags |= MSG_OOB;
                   1393:        }
1.69      thorpej  1394:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1395:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1396:
1.1       cgd      1397:        moff = 0;
                   1398:        offset = 0;
1.144     dyoung   1399:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1400:                if (m->m_type == MT_OOBDATA) {
                   1401:                        if (type != MT_OOBDATA)
                   1402:                                break;
                   1403:                } else if (type == MT_OOBDATA)
                   1404:                        break;
                   1405: #ifdef DIAGNOSTIC
                   1406:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1407:                        panic("receive 3");
                   1408: #endif
                   1409:                so->so_state &= ~SS_RCVATMARK;
                   1410:                len = uio->uio_resid;
                   1411:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1412:                        len = so->so_oobmark - offset;
                   1413:                if (len > m->m_len - moff)
                   1414:                        len = m->m_len - moff;
                   1415:                /*
                   1416:                 * If mp is set, just pass back the mbufs.
                   1417:                 * Otherwise copy them out via the uio, then free.
                   1418:                 * Sockbuf must be consistent here (points to current mbuf,
                   1419:                 * it points to next record) when we drop priority;
                   1420:                 * we must note any additions to the sockbuf when we
                   1421:                 * block interrupts again.
                   1422:                 */
1.144     dyoung   1423:                if (mp == NULL) {
1.69      thorpej  1424:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1425:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1426:                        sounlock(so);
1.1       cgd      1427:                        splx(s);
1.134     christos 1428:                        error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20      mycroft  1429:                        s = splsoftnet();
1.160     ad       1430:                        solock(so);
1.144     dyoung   1431:                        if (error != 0) {
1.67      he       1432:                                /*
                   1433:                                 * If any part of the record has been removed
                   1434:                                 * (such as the MT_SONAME mbuf, which will
                   1435:                                 * happen when PR_ADDR, and thus also
                   1436:                                 * PR_ATOMIC, is set), then drop the entire
                   1437:                                 * record to maintain the atomicity of the
                   1438:                                 * receive operation.
                   1439:                                 *
                   1440:                                 * This avoids a later panic("receive 1a")
                   1441:                                 * when compiled with DIAGNOSTIC.
                   1442:                                 */
1.146     dyoung   1443:                                if (m && mbuf_removed && atomic)
1.67      he       1444:                                        (void) sbdroprecord(&so->so_rcv);
                   1445:
1.57      jdolecek 1446:                                goto release;
1.67      he       1447:                        }
1.1       cgd      1448:                } else
                   1449:                        uio->uio_resid -= len;
                   1450:                if (len == m->m_len - moff) {
                   1451:                        if (m->m_flags & M_EOR)
                   1452:                                flags |= MSG_EOR;
                   1453:                        if (flags & MSG_PEEK) {
                   1454:                                m = m->m_next;
                   1455:                                moff = 0;
                   1456:                        } else {
                   1457:                                nextrecord = m->m_nextpkt;
                   1458:                                sbfree(&so->so_rcv, m);
                   1459:                                if (mp) {
                   1460:                                        *mp = m;
                   1461:                                        mp = &m->m_next;
                   1462:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1463:                                        *mp = NULL;
1.1       cgd      1464:                                } else {
                   1465:                                        MFREE(m, so->so_rcv.sb_mb);
                   1466:                                        m = so->so_rcv.sb_mb;
                   1467:                                }
1.69      thorpej  1468:                                /*
                   1469:                                 * If m != NULL, we also know that
                   1470:                                 * so->so_rcv.sb_mb != NULL.
                   1471:                                 */
                   1472:                                KASSERT(so->so_rcv.sb_mb == m);
                   1473:                                if (m) {
1.1       cgd      1474:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1475:                                        if (nextrecord == NULL)
                   1476:                                                so->so_rcv.sb_lastrecord = m;
                   1477:                                } else {
                   1478:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1479:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1480:                                }
                   1481:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1482:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1483:                        }
1.144     dyoung   1484:                } else if (flags & MSG_PEEK)
                   1485:                        moff += len;
                   1486:                else {
1.160     ad       1487:                        if (mp != NULL) {
                   1488:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1489:                                if (__predict_false(mt == NULL)) {
                   1490:                                        sounlock(so);
                   1491:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1492:                                        solock(so);
                   1493:                                }
                   1494:                                *mp = mt;
                   1495:                        }
1.144     dyoung   1496:                        m->m_data += len;
                   1497:                        m->m_len -= len;
                   1498:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1499:                }
                   1500:                if (so->so_oobmark) {
                   1501:                        if ((flags & MSG_PEEK) == 0) {
                   1502:                                so->so_oobmark -= len;
                   1503:                                if (so->so_oobmark == 0) {
                   1504:                                        so->so_state |= SS_RCVATMARK;
                   1505:                                        break;
                   1506:                                }
1.7       cgd      1507:                        } else {
1.1       cgd      1508:                                offset += len;
1.7       cgd      1509:                                if (offset == so->so_oobmark)
                   1510:                                        break;
                   1511:                        }
1.1       cgd      1512:                }
                   1513:                if (flags & MSG_EOR)
                   1514:                        break;
                   1515:                /*
                   1516:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1517:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1518:                 * termination.  If a signal/timeout occurs, return
                   1519:                 * with a short count but without error.
                   1520:                 * Keep sockbuf locked against other readers.
                   1521:                 */
1.144     dyoung   1522:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1523:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1524:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1525:                                break;
1.68      matt     1526:                        /*
                   1527:                         * If we are peeking and the socket receive buffer is
                   1528:                         * full, stop since we can't get more data to peek at.
                   1529:                         */
                   1530:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1531:                                break;
                   1532:                        /*
                   1533:                         * If we've drained the socket buffer, tell the
                   1534:                         * protocol in case it needs to do something to
                   1535:                         * get it filled again.
                   1536:                         */
                   1537:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
                   1538:                                (*pr->pr_usrreq)(so, PRU_RCVD,
1.140     dyoung   1539:                                    NULL, (struct mbuf *)(long)flags, NULL, l);
1.69      thorpej  1540:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1541:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1       cgd      1542:                        error = sbwait(&so->so_rcv);
1.144     dyoung   1543:                        if (error != 0) {
1.1       cgd      1544:                                sbunlock(&so->so_rcv);
1.160     ad       1545:                                sounlock(so);
1.1       cgd      1546:                                splx(s);
1.144     dyoung   1547:                                return 0;
1.1       cgd      1548:                        }
1.21      christos 1549:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1550:                                nextrecord = m->m_nextpkt;
                   1551:                }
                   1552:        }
1.3       andrew   1553:
1.146     dyoung   1554:        if (m && atomic) {
1.3       andrew   1555:                flags |= MSG_TRUNC;
                   1556:                if ((flags & MSG_PEEK) == 0)
                   1557:                        (void) sbdroprecord(&so->so_rcv);
                   1558:        }
1.1       cgd      1559:        if ((flags & MSG_PEEK) == 0) {
1.144     dyoung   1560:                if (m == NULL) {
1.69      thorpej  1561:                        /*
1.70      thorpej  1562:                         * First part is an inline SB_EMPTY_FIXUP().  Second
1.69      thorpej  1563:                         * part makes sure sb_lastrecord is up-to-date if
                   1564:                         * there is still data in the socket buffer.
                   1565:                         */
1.1       cgd      1566:                        so->so_rcv.sb_mb = nextrecord;
1.69      thorpej  1567:                        if (so->so_rcv.sb_mb == NULL) {
                   1568:                                so->so_rcv.sb_mbtail = NULL;
                   1569:                                so->so_rcv.sb_lastrecord = NULL;
                   1570:                        } else if (nextrecord->m_nextpkt == NULL)
                   1571:                                so->so_rcv.sb_lastrecord = nextrecord;
                   1572:                }
                   1573:                SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
                   1574:                SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1       cgd      1575:                if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140     dyoung   1576:                        (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
                   1577:                            (struct mbuf *)(long)flags, NULL, l);
1.1       cgd      1578:        }
1.3       andrew   1579:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1580:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1581:                sbunlock(&so->so_rcv);
                   1582:                goto restart;
                   1583:        }
1.108     perry    1584:
1.144     dyoung   1585:        if (flagsp != NULL)
1.1       cgd      1586:                *flagsp |= flags;
1.54      lukem    1587:  release:
1.1       cgd      1588:        sbunlock(&so->so_rcv);
1.160     ad       1589:        sounlock(so);
1.1       cgd      1590:        splx(s);
1.144     dyoung   1591:        return error;
1.1       cgd      1592: }
                   1593:
1.14      mycroft  1594: int
1.54      lukem    1595: soshutdown(struct socket *so, int how)
1.1       cgd      1596: {
1.99      matt     1597:        const struct protosw    *pr;
1.160     ad       1598:        int     error;
                   1599:
                   1600:        KASSERT(solocked(so));
1.34      kleink   1601:
1.54      lukem    1602:        pr = so->so_proto;
1.34      kleink   1603:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1604:                return (EINVAL);
1.1       cgd      1605:
1.160     ad       1606:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1607:                sorflush(so);
1.160     ad       1608:                error = 0;
                   1609:        }
1.34      kleink   1610:        if (how == SHUT_WR || how == SHUT_RDWR)
1.160     ad       1611:                error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1.140     dyoung   1612:                    NULL, NULL, NULL);
1.160     ad       1613:
                   1614:        return error;
1.1       cgd      1615: }
                   1616:
1.188     ad       1617: int
                   1618: sodrain(struct socket *so)
                   1619: {
                   1620:        int error;
                   1621:
                   1622:        solock(so);
                   1623:        so->so_state |= SS_ISDRAINING;
                   1624:        cv_broadcast(&so->so_cv);
                   1625:        error = soshutdown(so, SHUT_RDWR);
                   1626:        sounlock(so);
                   1627:
                   1628:        return error;
                   1629: }
                   1630:
1.14      mycroft  1631: void
1.54      lukem    1632: sorflush(struct socket *so)
1.1       cgd      1633: {
1.54      lukem    1634:        struct sockbuf  *sb, asb;
1.99      matt     1635:        const struct protosw    *pr;
1.160     ad       1636:
                   1637:        KASSERT(solocked(so));
1.1       cgd      1638:
1.54      lukem    1639:        sb = &so->so_rcv;
                   1640:        pr = so->so_proto;
1.160     ad       1641:        socantrcvmore(so);
1.1       cgd      1642:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1643:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1644:        sbunlock(sb);
                   1645:        asb = *sb;
1.86      wrstuden 1646:        /*
                   1647:         * Clear most of the sockbuf structure, but leave some of the
                   1648:         * fields valid.
                   1649:         */
                   1650:        memset(&sb->sb_startzero, 0,
                   1651:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1652:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1653:                sounlock(so);
1.1       cgd      1654:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1655:                solock(so);
                   1656:        }
1.98      christos 1657:        sbrelease(&asb, so);
1.1       cgd      1658: }
                   1659:
1.171     plunky   1660: /*
                   1661:  * internal set SOL_SOCKET options
                   1662:  */
1.142     dyoung   1663: static int
1.171     plunky   1664: sosetopt1(struct socket *so, const struct sockopt *sopt)
1.1       cgd      1665: {
1.182     christos 1666:        int error = EINVAL, optval, opt;
1.171     plunky   1667:        struct linger l;
                   1668:        struct timeval tv;
1.142     dyoung   1669:
1.179     christos 1670:        switch ((opt = sopt->sopt_name)) {
1.142     dyoung   1671:
1.170     tls      1672:        case SO_ACCEPTFILTER:
1.177     ad       1673:                error = accept_filt_setopt(so, sopt);
                   1674:                KASSERT(solocked(so));
1.170     tls      1675:                break;
                   1676:
1.171     plunky   1677:        case SO_LINGER:
                   1678:                error = sockopt_get(sopt, &l, sizeof(l));
1.177     ad       1679:                solock(so);
1.171     plunky   1680:                if (error)
1.177     ad       1681:                        break;
1.171     plunky   1682:                if (l.l_linger < 0 || l.l_linger > USHRT_MAX ||
1.177     ad       1683:                    l.l_linger > (INT_MAX / hz)) {
                   1684:                        error = EDOM;
                   1685:                        break;
                   1686:                }
1.171     plunky   1687:                so->so_linger = l.l_linger;
                   1688:                if (l.l_onoff)
                   1689:                        so->so_options |= SO_LINGER;
                   1690:                else
                   1691:                        so->so_options &= ~SO_LINGER;
1.177     ad       1692:                break;
1.1       cgd      1693:
1.142     dyoung   1694:        case SO_DEBUG:
                   1695:        case SO_KEEPALIVE:
                   1696:        case SO_DONTROUTE:
                   1697:        case SO_USELOOPBACK:
                   1698:        case SO_BROADCAST:
                   1699:        case SO_REUSEADDR:
                   1700:        case SO_REUSEPORT:
                   1701:        case SO_OOBINLINE:
                   1702:        case SO_TIMESTAMP:
1.184     christos 1703: #ifdef SO_OTIMESTAMP
                   1704:        case SO_OTIMESTAMP:
                   1705: #endif
1.171     plunky   1706:                error = sockopt_getint(sopt, &optval);
1.177     ad       1707:                solock(so);
1.171     plunky   1708:                if (error)
1.177     ad       1709:                        break;
1.171     plunky   1710:                if (optval)
1.179     christos 1711:                        so->so_options |= opt;
1.142     dyoung   1712:                else
1.179     christos 1713:                        so->so_options &= ~opt;
1.142     dyoung   1714:                break;
                   1715:
                   1716:        case SO_SNDBUF:
                   1717:        case SO_RCVBUF:
                   1718:        case SO_SNDLOWAT:
                   1719:        case SO_RCVLOWAT:
1.171     plunky   1720:                error = sockopt_getint(sopt, &optval);
1.177     ad       1721:                solock(so);
1.171     plunky   1722:                if (error)
1.177     ad       1723:                        break;
1.1       cgd      1724:
1.142     dyoung   1725:                /*
                   1726:                 * Values < 1 make no sense for any of these
                   1727:                 * options, so disallow them.
                   1728:                 */
1.177     ad       1729:                if (optval < 1) {
                   1730:                        error = EINVAL;
                   1731:                        break;
                   1732:                }
1.1       cgd      1733:
1.179     christos 1734:                switch (opt) {
1.171     plunky   1735:                case SO_SNDBUF:
1.177     ad       1736:                        if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) {
                   1737:                                error = ENOBUFS;
                   1738:                                break;
                   1739:                        }
1.171     plunky   1740:                        so->so_snd.sb_flags &= ~SB_AUTOSIZE;
                   1741:                        break;
1.1       cgd      1742:
                   1743:                case SO_RCVBUF:
1.177     ad       1744:                        if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) {
                   1745:                                error = ENOBUFS;
                   1746:                                break;
                   1747:                        }
1.171     plunky   1748:                        so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1.142     dyoung   1749:                        break;
                   1750:
                   1751:                /*
                   1752:                 * Make sure the low-water is never greater than
                   1753:                 * the high-water.
                   1754:                 */
1.1       cgd      1755:                case SO_SNDLOWAT:
1.171     plunky   1756:                        if (optval > so->so_snd.sb_hiwat)
                   1757:                                optval = so->so_snd.sb_hiwat;
                   1758:
                   1759:                        so->so_snd.sb_lowat = optval;
1.142     dyoung   1760:                        break;
1.171     plunky   1761:
1.1       cgd      1762:                case SO_RCVLOWAT:
1.171     plunky   1763:                        if (optval > so->so_rcv.sb_hiwat)
                   1764:                                optval = so->so_rcv.sb_hiwat;
                   1765:
                   1766:                        so->so_rcv.sb_lowat = optval;
1.142     dyoung   1767:                        break;
                   1768:                }
                   1769:                break;
1.28      thorpej  1770:
1.179     christos 1771: #ifdef COMPAT_50
                   1772:        case SO_OSNDTIMEO:
                   1773:        case SO_ORCVTIMEO: {
                   1774:                struct timeval50 otv;
                   1775:                error = sockopt_get(sopt, &otv, sizeof(otv));
1.186     pooka    1776:                if (error) {
                   1777:                        solock(so);
1.183     christos 1778:                        break;
1.186     pooka    1779:                }
1.179     christos 1780:                timeval50_to_timeval(&otv, &tv);
                   1781:                opt = opt == SO_OSNDTIMEO ? SO_SNDTIMEO : SO_RCVTIMEO;
1.182     christos 1782:                error = 0;
1.179     christos 1783:                /*FALLTHROUGH*/
                   1784:        }
                   1785: #endif /* COMPAT_50 */
                   1786:
1.142     dyoung   1787:        case SO_SNDTIMEO:
                   1788:        case SO_RCVTIMEO:
1.182     christos 1789:                if (error)
1.179     christos 1790:                        error = sockopt_get(sopt, &tv, sizeof(tv));
1.177     ad       1791:                solock(so);
1.171     plunky   1792:                if (error)
1.177     ad       1793:                        break;
1.171     plunky   1794:
1.177     ad       1795:                if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) {
                   1796:                        error = EDOM;
                   1797:                        break;
                   1798:                }
1.28      thorpej  1799:
1.171     plunky   1800:                optval = tv.tv_sec * hz + tv.tv_usec / tick;
                   1801:                if (optval == 0 && tv.tv_usec != 0)
                   1802:                        optval = 1;
1.28      thorpej  1803:
1.179     christos 1804:                switch (opt) {
1.142     dyoung   1805:                case SO_SNDTIMEO:
1.171     plunky   1806:                        so->so_snd.sb_timeo = optval;
1.1       cgd      1807:                        break;
                   1808:                case SO_RCVTIMEO:
1.171     plunky   1809:                        so->so_rcv.sb_timeo = optval;
1.142     dyoung   1810:                        break;
                   1811:                }
                   1812:                break;
1.1       cgd      1813:
1.142     dyoung   1814:        default:
1.177     ad       1815:                solock(so);
                   1816:                error = ENOPROTOOPT;
                   1817:                break;
1.142     dyoung   1818:        }
1.177     ad       1819:        KASSERT(solocked(so));
                   1820:        return error;
1.142     dyoung   1821: }
1.1       cgd      1822:
1.142     dyoung   1823: int
1.171     plunky   1824: sosetopt(struct socket *so, struct sockopt *sopt)
1.142     dyoung   1825: {
                   1826:        int error, prerr;
1.1       cgd      1827:
1.177     ad       1828:        if (sopt->sopt_level == SOL_SOCKET) {
1.171     plunky   1829:                error = sosetopt1(so, sopt);
1.177     ad       1830:                KASSERT(solocked(so));
                   1831:        } else {
1.142     dyoung   1832:                error = ENOPROTOOPT;
1.177     ad       1833:                solock(so);
                   1834:        }
1.1       cgd      1835:
1.142     dyoung   1836:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1837:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1838:                /* give the protocol stack a shot */
1.171     plunky   1839:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt);
1.142     dyoung   1840:                if (prerr == 0)
                   1841:                        error = 0;
                   1842:                else if (prerr != ENOPROTOOPT)
                   1843:                        error = prerr;
1.171     plunky   1844:        }
1.160     ad       1845:        sounlock(so);
1.142     dyoung   1846:        return error;
1.1       cgd      1847: }
                   1848:
1.171     plunky   1849: /*
                   1850:  * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt()
                   1851:  */
                   1852: int
                   1853: so_setsockopt(struct lwp *l, struct socket *so, int level, int name,
                   1854:     const void *val, size_t valsize)
                   1855: {
                   1856:        struct sockopt sopt;
                   1857:        int error;
                   1858:
                   1859:        KASSERT(valsize == 0 || val != NULL);
                   1860:
                   1861:        sockopt_init(&sopt, level, name, valsize);
                   1862:        sockopt_set(&sopt, val, valsize);
                   1863:
                   1864:        error = sosetopt(so, &sopt);
                   1865:
                   1866:        sockopt_destroy(&sopt);
                   1867:
                   1868:        return error;
                   1869: }
                   1870:
                   1871: /*
                   1872:  * internal get SOL_SOCKET options
                   1873:  */
                   1874: static int
                   1875: sogetopt1(struct socket *so, struct sockopt *sopt)
                   1876: {
1.179     christos 1877:        int error, optval, opt;
1.171     plunky   1878:        struct linger l;
                   1879:        struct timeval tv;
                   1880:
1.179     christos 1881:        switch ((opt = sopt->sopt_name)) {
1.171     plunky   1882:
                   1883:        case SO_ACCEPTFILTER:
1.177     ad       1884:                error = accept_filt_getopt(so, sopt);
1.171     plunky   1885:                break;
                   1886:
                   1887:        case SO_LINGER:
                   1888:                l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0;
                   1889:                l.l_linger = so->so_linger;
                   1890:
                   1891:                error = sockopt_set(sopt, &l, sizeof(l));
                   1892:                break;
                   1893:
                   1894:        case SO_USELOOPBACK:
                   1895:        case SO_DONTROUTE:
                   1896:        case SO_DEBUG:
                   1897:        case SO_KEEPALIVE:
                   1898:        case SO_REUSEADDR:
                   1899:        case SO_REUSEPORT:
                   1900:        case SO_BROADCAST:
                   1901:        case SO_OOBINLINE:
                   1902:        case SO_TIMESTAMP:
1.184     christos 1903: #ifdef SO_OTIMESTAMP
                   1904:        case SO_OTIMESTAMP:
                   1905: #endif
1.179     christos 1906:                error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0);
1.171     plunky   1907:                break;
                   1908:
                   1909:        case SO_TYPE:
                   1910:                error = sockopt_setint(sopt, so->so_type);
                   1911:                break;
                   1912:
                   1913:        case SO_ERROR:
                   1914:                error = sockopt_setint(sopt, so->so_error);
                   1915:                so->so_error = 0;
                   1916:                break;
                   1917:
                   1918:        case SO_SNDBUF:
                   1919:                error = sockopt_setint(sopt, so->so_snd.sb_hiwat);
                   1920:                break;
                   1921:
                   1922:        case SO_RCVBUF:
                   1923:                error = sockopt_setint(sopt, so->so_rcv.sb_hiwat);
                   1924:                break;
                   1925:
                   1926:        case SO_SNDLOWAT:
                   1927:                error = sockopt_setint(sopt, so->so_snd.sb_lowat);
                   1928:                break;
                   1929:
                   1930:        case SO_RCVLOWAT:
                   1931:                error = sockopt_setint(sopt, so->so_rcv.sb_lowat);
                   1932:                break;
                   1933:
1.179     christos 1934: #ifdef COMPAT_50
                   1935:        case SO_OSNDTIMEO:
                   1936:        case SO_ORCVTIMEO: {
                   1937:                struct timeval50 otv;
                   1938:
                   1939:                optval = (opt == SO_OSNDTIMEO ?
                   1940:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1941:
                   1942:                otv.tv_sec = optval / hz;
                   1943:                otv.tv_usec = (optval % hz) * tick;
                   1944:
                   1945:                error = sockopt_set(sopt, &otv, sizeof(otv));
                   1946:                break;
                   1947:        }
                   1948: #endif /* COMPAT_50 */
                   1949:
1.171     plunky   1950:        case SO_SNDTIMEO:
                   1951:        case SO_RCVTIMEO:
1.179     christos 1952:                optval = (opt == SO_SNDTIMEO ?
1.171     plunky   1953:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1954:
                   1955:                tv.tv_sec = optval / hz;
                   1956:                tv.tv_usec = (optval % hz) * tick;
                   1957:
                   1958:                error = sockopt_set(sopt, &tv, sizeof(tv));
                   1959:                break;
                   1960:
                   1961:        case SO_OVERFLOWED:
                   1962:                error = sockopt_setint(sopt, so->so_rcv.sb_overflowed);
                   1963:                break;
                   1964:
                   1965:        default:
                   1966:                error = ENOPROTOOPT;
                   1967:                break;
                   1968:        }
                   1969:
                   1970:        return (error);
                   1971: }
                   1972:
1.14      mycroft  1973: int
1.171     plunky   1974: sogetopt(struct socket *so, struct sockopt *sopt)
1.1       cgd      1975: {
1.160     ad       1976:        int             error;
1.1       cgd      1977:
1.160     ad       1978:        solock(so);
1.171     plunky   1979:        if (sopt->sopt_level != SOL_SOCKET) {
1.1       cgd      1980:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       1981:                        error = ((*so->so_proto->pr_ctloutput)
1.171     plunky   1982:                            (PRCO_GETOPT, so, sopt));
1.1       cgd      1983:                } else
1.160     ad       1984:                        error = (ENOPROTOOPT);
1.1       cgd      1985:        } else {
1.171     plunky   1986:                error = sogetopt1(so, sopt);
                   1987:        }
                   1988:        sounlock(so);
                   1989:        return (error);
                   1990: }
                   1991:
                   1992: /*
                   1993:  * alloc sockopt data buffer buffer
                   1994:  *     - will be released at destroy
                   1995:  */
1.176     plunky   1996: static int
                   1997: sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag)
1.171     plunky   1998: {
                   1999:
                   2000:        KASSERT(sopt->sopt_size == 0);
                   2001:
1.176     plunky   2002:        if (len > sizeof(sopt->sopt_buf)) {
                   2003:                sopt->sopt_data = kmem_zalloc(len, kmflag);
                   2004:                if (sopt->sopt_data == NULL)
                   2005:                        return ENOMEM;
                   2006:        } else
1.171     plunky   2007:                sopt->sopt_data = sopt->sopt_buf;
                   2008:
                   2009:        sopt->sopt_size = len;
1.176     plunky   2010:        return 0;
1.171     plunky   2011: }
                   2012:
                   2013: /*
                   2014:  * initialise sockopt storage
1.176     plunky   2015:  *     - MAY sleep during allocation
1.171     plunky   2016:  */
                   2017: void
                   2018: sockopt_init(struct sockopt *sopt, int level, int name, size_t size)
                   2019: {
1.1       cgd      2020:
1.171     plunky   2021:        memset(sopt, 0, sizeof(*sopt));
1.1       cgd      2022:
1.171     plunky   2023:        sopt->sopt_level = level;
                   2024:        sopt->sopt_name = name;
1.176     plunky   2025:        (void)sockopt_alloc(sopt, size, KM_SLEEP);
1.171     plunky   2026: }
                   2027:
                   2028: /*
                   2029:  * destroy sockopt storage
                   2030:  *     - will release any held memory references
                   2031:  */
                   2032: void
                   2033: sockopt_destroy(struct sockopt *sopt)
                   2034: {
                   2035:
                   2036:        if (sopt->sopt_data != sopt->sopt_buf)
1.173     plunky   2037:                kmem_free(sopt->sopt_data, sopt->sopt_size);
1.171     plunky   2038:
                   2039:        memset(sopt, 0, sizeof(*sopt));
                   2040: }
                   2041:
                   2042: /*
                   2043:  * set sockopt value
                   2044:  *     - value is copied into sockopt
1.176     plunky   2045:  *     - memory is allocated when necessary, will not sleep
1.171     plunky   2046:  */
                   2047: int
                   2048: sockopt_set(struct sockopt *sopt, const void *buf, size_t len)
                   2049: {
1.176     plunky   2050:        int error;
1.171     plunky   2051:
1.176     plunky   2052:        if (sopt->sopt_size == 0) {
                   2053:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2054:                if (error)
                   2055:                        return error;
                   2056:        }
1.171     plunky   2057:
                   2058:        KASSERT(sopt->sopt_size == len);
                   2059:        memcpy(sopt->sopt_data, buf, len);
                   2060:        return 0;
                   2061: }
                   2062:
                   2063: /*
                   2064:  * common case of set sockopt integer value
                   2065:  */
                   2066: int
                   2067: sockopt_setint(struct sockopt *sopt, int val)
                   2068: {
                   2069:
                   2070:        return sockopt_set(sopt, &val, sizeof(int));
                   2071: }
                   2072:
                   2073: /*
                   2074:  * get sockopt value
                   2075:  *     - correct size must be given
                   2076:  */
                   2077: int
                   2078: sockopt_get(const struct sockopt *sopt, void *buf, size_t len)
                   2079: {
1.170     tls      2080:
1.171     plunky   2081:        if (sopt->sopt_size != len)
                   2082:                return EINVAL;
1.1       cgd      2083:
1.171     plunky   2084:        memcpy(buf, sopt->sopt_data, len);
                   2085:        return 0;
                   2086: }
1.1       cgd      2087:
1.171     plunky   2088: /*
                   2089:  * common case of get sockopt integer value
                   2090:  */
                   2091: int
                   2092: sockopt_getint(const struct sockopt *sopt, int *valp)
                   2093: {
1.1       cgd      2094:
1.171     plunky   2095:        return sockopt_get(sopt, valp, sizeof(int));
                   2096: }
1.1       cgd      2097:
1.171     plunky   2098: /*
                   2099:  * set sockopt value from mbuf
                   2100:  *     - ONLY for legacy code
                   2101:  *     - mbuf is released by sockopt
1.176     plunky   2102:  *     - will not sleep
1.171     plunky   2103:  */
                   2104: int
                   2105: sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m)
                   2106: {
                   2107:        size_t len;
1.176     plunky   2108:        int error;
1.1       cgd      2109:
1.171     plunky   2110:        len = m_length(m);
1.1       cgd      2111:
1.176     plunky   2112:        if (sopt->sopt_size == 0) {
                   2113:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2114:                if (error)
                   2115:                        return error;
                   2116:        }
1.1       cgd      2117:
1.171     plunky   2118:        KASSERT(sopt->sopt_size == len);
                   2119:        m_copydata(m, 0, len, sopt->sopt_data);
                   2120:        m_freem(m);
1.1       cgd      2121:
1.171     plunky   2122:        return 0;
                   2123: }
1.1       cgd      2124:
1.171     plunky   2125: /*
                   2126:  * get sockopt value into mbuf
                   2127:  *     - ONLY for legacy code
                   2128:  *     - mbuf to be released by the caller
1.176     plunky   2129:  *     - will not sleep
1.171     plunky   2130:  */
                   2131: struct mbuf *
                   2132: sockopt_getmbuf(const struct sockopt *sopt)
                   2133: {
                   2134:        struct mbuf *m;
1.107     darrenr  2135:
1.176     plunky   2136:        if (sopt->sopt_size > MCLBYTES)
                   2137:                return NULL;
                   2138:
                   2139:        m = m_get(M_DONTWAIT, MT_SOOPTS);
1.171     plunky   2140:        if (m == NULL)
                   2141:                return NULL;
                   2142:
1.176     plunky   2143:        if (sopt->sopt_size > MLEN) {
                   2144:                MCLGET(m, M_DONTWAIT);
                   2145:                if ((m->m_flags & M_EXT) == 0) {
                   2146:                        m_free(m);
                   2147:                        return NULL;
                   2148:                }
1.1       cgd      2149:        }
1.176     plunky   2150:
                   2151:        memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size);
                   2152:        m->m_len = sopt->sopt_size;
1.160     ad       2153:
1.171     plunky   2154:        return m;
1.1       cgd      2155: }
                   2156:
1.14      mycroft  2157: void
1.54      lukem    2158: sohasoutofband(struct socket *so)
1.1       cgd      2159: {
1.153     rmind    2160:
1.90      christos 2161:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.189     ad       2162:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT);
1.1       cgd      2163: }
1.72      jdolecek 2164:
                   2165: static void
                   2166: filt_sordetach(struct knote *kn)
                   2167: {
                   2168:        struct socket   *so;
                   2169:
1.155     ad       2170:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2171:        solock(so);
1.73      christos 2172:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2173:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 2174:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       2175:        sounlock(so);
1.72      jdolecek 2176: }
                   2177:
                   2178: /*ARGSUSED*/
                   2179: static int
1.129     yamt     2180: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 2181: {
                   2182:        struct socket   *so;
1.160     ad       2183:        int rv;
1.72      jdolecek 2184:
1.155     ad       2185:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2186:        if (hint != NOTE_SUBMIT)
                   2187:                solock(so);
1.72      jdolecek 2188:        kn->kn_data = so->so_rcv.sb_cc;
                   2189:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    2190:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2191:                kn->kn_fflags = so->so_error;
1.160     ad       2192:                rv = 1;
                   2193:        } else if (so->so_error)        /* temporary udp error */
                   2194:                rv = 1;
                   2195:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2196:                rv = (kn->kn_data >= kn->kn_sdata);
                   2197:        else
                   2198:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   2199:        if (hint != NOTE_SUBMIT)
                   2200:                sounlock(so);
                   2201:        return rv;
1.72      jdolecek 2202: }
                   2203:
                   2204: static void
                   2205: filt_sowdetach(struct knote *kn)
                   2206: {
                   2207:        struct socket   *so;
                   2208:
1.155     ad       2209:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2210:        solock(so);
1.73      christos 2211:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2212:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 2213:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       2214:        sounlock(so);
1.72      jdolecek 2215: }
                   2216:
                   2217: /*ARGSUSED*/
                   2218: static int
1.129     yamt     2219: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 2220: {
                   2221:        struct socket   *so;
1.160     ad       2222:        int rv;
1.72      jdolecek 2223:
1.155     ad       2224:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2225:        if (hint != NOTE_SUBMIT)
                   2226:                solock(so);
1.72      jdolecek 2227:        kn->kn_data = sbspace(&so->so_snd);
                   2228:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    2229:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2230:                kn->kn_fflags = so->so_error;
1.160     ad       2231:                rv = 1;
                   2232:        } else if (so->so_error)        /* temporary udp error */
                   2233:                rv = 1;
                   2234:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 2235:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       2236:                rv = 0;
                   2237:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2238:                rv = (kn->kn_data >= kn->kn_sdata);
                   2239:        else
                   2240:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   2241:        if (hint != NOTE_SUBMIT)
                   2242:                sounlock(so);
                   2243:        return rv;
1.72      jdolecek 2244: }
                   2245:
                   2246: /*ARGSUSED*/
                   2247: static int
1.129     yamt     2248: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 2249: {
                   2250:        struct socket   *so;
1.160     ad       2251:        int rv;
1.72      jdolecek 2252:
1.155     ad       2253:        so = ((file_t *)kn->kn_obj)->f_data;
1.72      jdolecek 2254:
                   2255:        /*
                   2256:         * Set kn_data to number of incoming connections, not
                   2257:         * counting partial (incomplete) connections.
1.108     perry    2258:         */
1.160     ad       2259:        if (hint != NOTE_SUBMIT)
                   2260:                solock(so);
1.72      jdolecek 2261:        kn->kn_data = so->so_qlen;
1.160     ad       2262:        rv = (kn->kn_data > 0);
                   2263:        if (hint != NOTE_SUBMIT)
                   2264:                sounlock(so);
                   2265:        return rv;
1.72      jdolecek 2266: }
                   2267:
                   2268: static const struct filterops solisten_filtops =
                   2269:        { 1, NULL, filt_sordetach, filt_solisten };
                   2270: static const struct filterops soread_filtops =
                   2271:        { 1, NULL, filt_sordetach, filt_soread };
                   2272: static const struct filterops sowrite_filtops =
                   2273:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   2274:
                   2275: int
1.129     yamt     2276: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 2277: {
                   2278:        struct socket   *so;
                   2279:        struct sockbuf  *sb;
                   2280:
1.155     ad       2281:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2282:        solock(so);
1.72      jdolecek 2283:        switch (kn->kn_filter) {
                   2284:        case EVFILT_READ:
                   2285:                if (so->so_options & SO_ACCEPTCONN)
                   2286:                        kn->kn_fop = &solisten_filtops;
                   2287:                else
                   2288:                        kn->kn_fop = &soread_filtops;
                   2289:                sb = &so->so_rcv;
                   2290:                break;
                   2291:        case EVFILT_WRITE:
                   2292:                kn->kn_fop = &sowrite_filtops;
                   2293:                sb = &so->so_snd;
                   2294:                break;
                   2295:        default:
1.160     ad       2296:                sounlock(so);
1.149     pooka    2297:                return (EINVAL);
1.72      jdolecek 2298:        }
1.73      christos 2299:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 2300:        sb->sb_flags |= SB_KNOTE;
1.160     ad       2301:        sounlock(so);
1.72      jdolecek 2302:        return (0);
                   2303: }
                   2304:
1.154     ad       2305: static int
                   2306: sodopoll(struct socket *so, int events)
                   2307: {
                   2308:        int revents;
                   2309:
                   2310:        revents = 0;
                   2311:
                   2312:        if (events & (POLLIN | POLLRDNORM))
                   2313:                if (soreadable(so))
                   2314:                        revents |= events & (POLLIN | POLLRDNORM);
                   2315:
                   2316:        if (events & (POLLOUT | POLLWRNORM))
                   2317:                if (sowritable(so))
                   2318:                        revents |= events & (POLLOUT | POLLWRNORM);
                   2319:
                   2320:        if (events & (POLLPRI | POLLRDBAND))
                   2321:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   2322:                        revents |= events & (POLLPRI | POLLRDBAND);
                   2323:
                   2324:        return revents;
                   2325: }
                   2326:
                   2327: int
                   2328: sopoll(struct socket *so, int events)
                   2329: {
                   2330:        int revents = 0;
                   2331:
1.160     ad       2332: #ifndef DIAGNOSTIC
                   2333:        /*
                   2334:         * Do a quick, unlocked check in expectation that the socket
                   2335:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   2336:         * as the solocked() assertions will fail.
                   2337:         */
1.154     ad       2338:        if ((revents = sodopoll(so, events)) != 0)
                   2339:                return revents;
1.160     ad       2340: #endif
1.154     ad       2341:
1.160     ad       2342:        solock(so);
1.154     ad       2343:        if ((revents = sodopoll(so, events)) == 0) {
                   2344:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   2345:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       2346:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       2347:                }
                   2348:
                   2349:                if (events & (POLLOUT | POLLWRNORM)) {
                   2350:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2351:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2352:                }
                   2353:        }
1.160     ad       2354:        sounlock(so);
1.154     ad       2355:
                   2356:        return revents;
                   2357: }
                   2358:
                   2359:
1.94      yamt     2360: #include <sys/sysctl.h>
                   2361:
                   2362: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
                   2363:
                   2364: /*
                   2365:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2366:  * value is not too small.
                   2367:  * (XXX should we maybe make sure it's not too large as well?)
                   2368:  */
                   2369: static int
                   2370: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2371: {
                   2372:        int error, new_somaxkva;
                   2373:        struct sysctlnode node;
                   2374:
                   2375:        new_somaxkva = somaxkva;
                   2376:        node = *rnode;
                   2377:        node.sysctl_data = &new_somaxkva;
                   2378:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2379:        if (error || newp == NULL)
                   2380:                return (error);
                   2381:
                   2382:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2383:                return (EINVAL);
                   2384:
1.136     ad       2385:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2386:        somaxkva = new_somaxkva;
1.136     ad       2387:        cv_broadcast(&socurkva_cv);
                   2388:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2389:
                   2390:        return (error);
                   2391: }
                   2392:
1.178     pooka    2393: static void
1.187     cegger   2394: sysctl_kern_somaxkva_setup(void)
1.94      yamt     2395: {
                   2396:
1.178     pooka    2397:        KASSERT(socket_sysctllog == NULL);
                   2398:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2399:                       CTLFLAG_PERMANENT,
                   2400:                       CTLTYPE_NODE, "kern", NULL,
                   2401:                       NULL, 0, NULL, 0,
                   2402:                       CTL_KERN, CTL_EOL);
                   2403:
1.178     pooka    2404:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2405:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2406:                       CTLTYPE_INT, "somaxkva",
                   2407:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2408:                                    "used for socket buffers"),
1.94      yamt     2409:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2410:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
                   2411: }

CVSweb <webmaster@jp.NetBSD.org>