[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 1.233

1.233   ! rtr         1: /*     $NetBSD: uipc_socket.c,v 1.232 2014/08/05 07:55:31 rtr Exp $    */
1.64      thorpej     2:
                      3: /*-
1.188     ad          4:  * Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc.
1.64      thorpej     5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
1.188     ad          8:  * by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran.
1.64      thorpej     9:  *
                     10:  * Redistribution and use in source and binary forms, with or without
                     11:  * modification, are permitted provided that the following conditions
                     12:  * are met:
                     13:  * 1. Redistributions of source code must retain the above copyright
                     14:  *    notice, this list of conditions and the following disclaimer.
                     15:  * 2. Redistributions in binary form must reproduce the above copyright
                     16:  *    notice, this list of conditions and the following disclaimer in the
                     17:  *    documentation and/or other materials provided with the distribution.
                     18:  *
                     19:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     20:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     21:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     22:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     23:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     24:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     25:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     26:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     27:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     28:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     29:  * POSSIBILITY OF SUCH DAMAGE.
                     30:  */
1.16      cgd        31:
1.1       cgd        32: /*
1.159     ad         33:  * Copyright (c) 2004 The FreeBSD Foundation
                     34:  * Copyright (c) 2004 Robert Watson
1.15      mycroft    35:  * Copyright (c) 1982, 1986, 1988, 1990, 1993
                     36:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        37:  *
                     38:  * Redistribution and use in source and binary forms, with or without
                     39:  * modification, are permitted provided that the following conditions
                     40:  * are met:
                     41:  * 1. Redistributions of source code must retain the above copyright
                     42:  *    notice, this list of conditions and the following disclaimer.
                     43:  * 2. Redistributions in binary form must reproduce the above copyright
                     44:  *    notice, this list of conditions and the following disclaimer in the
                     45:  *    documentation and/or other materials provided with the distribution.
1.85      agc        46:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        47:  *    may be used to endorse or promote products derived from this software
                     48:  *    without specific prior written permission.
                     49:  *
                     50:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     51:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     52:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     53:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     54:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     55:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     56:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     57:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     58:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     59:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     60:  * SUCH DAMAGE.
                     61:  *
1.32      fvdl       62:  *     @(#)uipc_socket.c       8.6 (Berkeley) 5/2/95
1.1       cgd        63:  */
1.59      lukem      64:
1.222     rmind      65: /*
                     66:  * Socket operation routines.
                     67:  *
                     68:  * These routines are called by the routines in sys_socket.c or from a
                     69:  * system process, and implement the semantics of socket operations by
                     70:  * switching out to the protocol specific routines.
                     71:  */
                     72:
1.59      lukem      73: #include <sys/cdefs.h>
1.233   ! rtr        74: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.232 2014/08/05 07:55:31 rtr Exp $");
1.64      thorpej    75:
1.179     christos   76: #include "opt_compat_netbsd.h"
1.64      thorpej    77: #include "opt_sock_counters.h"
                     78: #include "opt_sosend_loan.h"
1.81      martin     79: #include "opt_mbuftrace.h"
1.84      ragge      80: #include "opt_somaxkva.h"
1.167     ad         81: #include "opt_multiprocessor.h"        /* XXX */
1.1       cgd        82:
1.9       mycroft    83: #include <sys/param.h>
                     84: #include <sys/systm.h>
                     85: #include <sys/proc.h>
                     86: #include <sys/file.h>
1.142     dyoung     87: #include <sys/filedesc.h>
1.173     plunky     88: #include <sys/kmem.h>
1.9       mycroft    89: #include <sys/mbuf.h>
                     90: #include <sys/domain.h>
                     91: #include <sys/kernel.h>
                     92: #include <sys/protosw.h>
                     93: #include <sys/socket.h>
                     94: #include <sys/socketvar.h>
1.21      christos   95: #include <sys/signalvar.h>
1.9       mycroft    96: #include <sys/resourcevar.h>
1.174     pooka      97: #include <sys/uidinfo.h>
1.72      jdolecek   98: #include <sys/event.h>
1.89      christos   99: #include <sys/poll.h>
1.118     elad      100: #include <sys/kauth.h>
1.136     ad        101: #include <sys/mutex.h>
                    102: #include <sys/condvar.h>
1.205     bouyer    103: #include <sys/kthread.h>
1.37      thorpej   104:
1.179     christos  105: #ifdef COMPAT_50
                    106: #include <compat/sys/time.h>
1.184     christos  107: #include <compat/sys/socket.h>
1.179     christos  108: #endif
                    109:
1.202     uebayasi  110: #include <uvm/uvm_extern.h>
                    111: #include <uvm/uvm_loan.h>
                    112: #include <uvm/uvm_page.h>
1.64      thorpej   113:
1.77      thorpej   114: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37      thorpej   115:
1.142     dyoung    116: extern const struct fileops socketops;
                    117:
1.54      lukem     118: extern int     somaxconn;                      /* patchable (XXX sysctl) */
                    119: int            somaxconn = SOMAXCONN;
1.160     ad        120: kmutex_t       *softnet_lock;
1.49      jonathan  121:
1.64      thorpej   122: #ifdef SOSEND_COUNTERS
                    123: #include <sys/device.h>
                    124:
1.113     thorpej   125: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   126:     NULL, "sosend", "loan big");
1.113     thorpej   127: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   128:     NULL, "sosend", "copy big");
1.113     thorpej   129: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   130:     NULL, "sosend", "copy small");
1.113     thorpej   131: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   132:     NULL, "sosend", "kva limit");
                    133:
                    134: #define        SOSEND_COUNTER_INCR(ev)         (ev)->ev_count++
                    135:
1.101     matt      136: EVCNT_ATTACH_STATIC(sosend_loan_big);
                    137: EVCNT_ATTACH_STATIC(sosend_copy_big);
                    138: EVCNT_ATTACH_STATIC(sosend_copy_small);
                    139: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64      thorpej   140: #else
                    141:
                    142: #define        SOSEND_COUNTER_INCR(ev)         /* nothing */
                    143:
                    144: #endif /* SOSEND_COUNTERS */
                    145:
1.167     ad        146: #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR)
1.121     yamt      147: int sock_loan_thresh = -1;
1.71      thorpej   148: #else
1.121     yamt      149: int sock_loan_thresh = 4096;
1.65      thorpej   150: #endif
1.64      thorpej   151:
1.136     ad        152: static kmutex_t so_pendfree_lock;
1.205     bouyer    153: static struct mbuf *so_pendfree = NULL;
1.64      thorpej   154:
1.84      ragge     155: #ifndef SOMAXKVA
                    156: #define        SOMAXKVA (16 * 1024 * 1024)
                    157: #endif
                    158: int somaxkva = SOMAXKVA;
1.113     thorpej   159: static int socurkva;
1.136     ad        160: static kcondvar_t socurkva_cv;
1.64      thorpej   161:
1.191     elad      162: static kauth_listener_t socket_listener;
                    163:
1.64      thorpej   164: #define        SOCK_LOAN_CHUNK         65536
                    165:
1.205     bouyer    166: static void sopendfree_thread(void *);
                    167: static kcondvar_t pendfree_thread_cv;
                    168: static lwp_t *sopendfree_lwp;
1.93      yamt      169:
1.212     pooka     170: static void sysctl_kern_socket_setup(void);
1.178     pooka     171: static struct sysctllog *socket_sysctllog;
                    172:
1.113     thorpej   173: static vsize_t
1.129     yamt      174: sokvareserve(struct socket *so, vsize_t len)
1.80      yamt      175: {
1.98      christos  176:        int error;
1.80      yamt      177:
1.136     ad        178:        mutex_enter(&so_pendfree_lock);
1.80      yamt      179:        while (socurkva + len > somaxkva) {
                    180:                SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136     ad        181:                error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98      christos  182:                if (error) {
                    183:                        len = 0;
                    184:                        break;
                    185:                }
1.80      yamt      186:        }
1.93      yamt      187:        socurkva += len;
1.136     ad        188:        mutex_exit(&so_pendfree_lock);
1.98      christos  189:        return len;
1.95      yamt      190: }
                    191:
1.113     thorpej   192: static void
1.95      yamt      193: sokvaunreserve(vsize_t len)
                    194: {
                    195:
1.136     ad        196:        mutex_enter(&so_pendfree_lock);
1.95      yamt      197:        socurkva -= len;
1.136     ad        198:        cv_broadcast(&socurkva_cv);
                    199:        mutex_exit(&so_pendfree_lock);
1.95      yamt      200: }
                    201:
                    202: /*
                    203:  * sokvaalloc: allocate kva for loan.
                    204:  */
                    205:
                    206: vaddr_t
1.209     matt      207: sokvaalloc(vaddr_t sva, vsize_t len, struct socket *so)
1.95      yamt      208: {
                    209:        vaddr_t lva;
                    210:
                    211:        /*
                    212:         * reserve kva.
                    213:         */
                    214:
1.98      christos  215:        if (sokvareserve(so, len) == 0)
                    216:                return 0;
1.93      yamt      217:
                    218:        /*
                    219:         * allocate kva.
                    220:         */
1.80      yamt      221:
1.209     matt      222:        lva = uvm_km_alloc(kernel_map, len, atop(sva) & uvmexp.colormask,
                    223:            UVM_KMF_COLORMATCH | UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95      yamt      224:        if (lva == 0) {
                    225:                sokvaunreserve(len);
1.80      yamt      226:                return (0);
1.95      yamt      227:        }
1.80      yamt      228:
                    229:        return lva;
                    230: }
                    231:
1.93      yamt      232: /*
                    233:  * sokvafree: free kva for loan.
                    234:  */
                    235:
1.80      yamt      236: void
                    237: sokvafree(vaddr_t sva, vsize_t len)
                    238: {
1.93      yamt      239:
                    240:        /*
                    241:         * free kva.
                    242:         */
1.80      yamt      243:
1.109     yamt      244:        uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93      yamt      245:
                    246:        /*
                    247:         * unreserve kva.
                    248:         */
                    249:
1.95      yamt      250:        sokvaunreserve(len);
1.80      yamt      251: }
                    252:
1.64      thorpej   253: static void
1.134     christos  254: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64      thorpej   255: {
1.156     yamt      256:        vaddr_t sva, eva;
1.64      thorpej   257:        vsize_t len;
1.156     yamt      258:        int npgs;
                    259:
                    260:        KASSERT(pgs != NULL);
1.64      thorpej   261:
                    262:        eva = round_page((vaddr_t) buf + size);
                    263:        sva = trunc_page((vaddr_t) buf);
                    264:        len = eva - sva;
                    265:        npgs = len >> PAGE_SHIFT;
                    266:
                    267:        pmap_kremove(sva, len);
                    268:        pmap_update(pmap_kernel());
                    269:        uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80      yamt      270:        sokvafree(sva, len);
1.64      thorpej   271: }
                    272:
1.93      yamt      273: /*
1.205     bouyer    274:  * sopendfree_thread: free mbufs on "pendfree" list.
1.136     ad        275:  * unlock and relock so_pendfree_lock when freeing mbufs.
1.93      yamt      276:  */
                    277:
1.205     bouyer    278: static void
                    279: sopendfree_thread(void *v)
1.93      yamt      280: {
1.137     ad        281:        struct mbuf *m, *next;
1.205     bouyer    282:        size_t rv;
1.93      yamt      283:
1.205     bouyer    284:        mutex_enter(&so_pendfree_lock);
1.64      thorpej   285:
1.205     bouyer    286:        for (;;) {
                    287:                rv = 0;
                    288:                while (so_pendfree != NULL) {
                    289:                        m = so_pendfree;
                    290:                        so_pendfree = NULL;
                    291:                        mutex_exit(&so_pendfree_lock);
                    292:
                    293:                        for (; m != NULL; m = next) {
                    294:                                next = m->m_next;
                    295:                                KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
                    296:                                KASSERT(m->m_ext.ext_refcnt == 0);
                    297:
                    298:                                rv += m->m_ext.ext_size;
                    299:                                sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
                    300:                                    m->m_ext.ext_size);
                    301:                                pool_cache_put(mb_cache, m);
                    302:                        }
1.93      yamt      303:
1.205     bouyer    304:                        mutex_enter(&so_pendfree_lock);
1.93      yamt      305:                }
1.205     bouyer    306:                if (rv)
                    307:                        cv_broadcast(&socurkva_cv);
                    308:                cv_wait(&pendfree_thread_cv, &so_pendfree_lock);
1.64      thorpej   309:        }
1.205     bouyer    310:        panic("sopendfree_thread");
                    311:        /* NOTREACHED */
1.64      thorpej   312: }
                    313:
1.80      yamt      314: void
1.134     christos  315: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64      thorpej   316: {
                    317:
1.156     yamt      318:        KASSERT(m != NULL);
1.64      thorpej   319:
1.93      yamt      320:        /*
                    321:         * postpone freeing mbuf.
                    322:         *
                    323:         * we can't do it in interrupt context
                    324:         * because we need to put kva back to kernel_map.
                    325:         */
                    326:
1.136     ad        327:        mutex_enter(&so_pendfree_lock);
1.92      yamt      328:        m->m_next = so_pendfree;
                    329:        so_pendfree = m;
1.205     bouyer    330:        cv_signal(&pendfree_thread_cv);
1.136     ad        331:        mutex_exit(&so_pendfree_lock);
1.64      thorpej   332: }
                    333:
                    334: static long
                    335: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
                    336: {
                    337:        struct iovec *iov = uio->uio_iov;
                    338:        vaddr_t sva, eva;
                    339:        vsize_t len;
1.156     yamt      340:        vaddr_t lva;
                    341:        int npgs, error;
                    342:        vaddr_t va;
                    343:        int i;
1.64      thorpej   344:
1.116     yamt      345:        if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64      thorpej   346:                return (0);
                    347:
                    348:        if (iov->iov_len < (size_t) space)
                    349:                space = iov->iov_len;
                    350:        if (space > SOCK_LOAN_CHUNK)
                    351:                space = SOCK_LOAN_CHUNK;
                    352:
                    353:        eva = round_page((vaddr_t) iov->iov_base + space);
                    354:        sva = trunc_page((vaddr_t) iov->iov_base);
                    355:        len = eva - sva;
                    356:        npgs = len >> PAGE_SHIFT;
                    357:
1.79      thorpej   358:        KASSERT(npgs <= M_EXT_MAXPAGES);
                    359:
1.209     matt      360:        lva = sokvaalloc(sva, len, so);
1.64      thorpej   361:        if (lva == 0)
1.80      yamt      362:                return 0;
1.64      thorpej   363:
1.116     yamt      364:        error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79      thorpej   365:            m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64      thorpej   366:        if (error) {
1.80      yamt      367:                sokvafree(lva, len);
1.64      thorpej   368:                return (0);
                    369:        }
                    370:
                    371:        for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79      thorpej   372:                pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
1.194     cegger    373:                    VM_PROT_READ, 0);
1.64      thorpej   374:        pmap_update(pmap_kernel());
                    375:
                    376:        lva += (vaddr_t) iov->iov_base & PAGE_MASK;
                    377:
1.134     christos  378:        MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79      thorpej   379:        m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64      thorpej   380:
                    381:        uio->uio_resid -= space;
                    382:        /* uio_offset not updated, not set/used for write(2) */
1.134     christos  383:        uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64      thorpej   384:        uio->uio_iov->iov_len -= space;
                    385:        if (uio->uio_iov->iov_len == 0) {
                    386:                uio->uio_iov++;
                    387:                uio->uio_iovcnt--;
                    388:        }
                    389:
                    390:        return (space);
                    391: }
                    392:
1.142     dyoung    393: struct mbuf *
1.147     dyoung    394: getsombuf(struct socket *so, int type)
1.142     dyoung    395: {
                    396:        struct mbuf *m;
                    397:
1.147     dyoung    398:        m = m_get(M_WAIT, type);
1.142     dyoung    399:        MCLAIM(m, so->so_mowner);
                    400:        return m;
                    401: }
                    402:
1.191     elad      403: static int
                    404: socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
                    405:     void *arg0, void *arg1, void *arg2, void *arg3)
                    406: {
                    407:        int result;
                    408:        enum kauth_network_req req;
                    409:
                    410:        result = KAUTH_RESULT_DEFER;
                    411:        req = (enum kauth_network_req)arg0;
                    412:
1.193     elad      413:        if ((action != KAUTH_NETWORK_SOCKET) &&
                    414:            (action != KAUTH_NETWORK_BIND))
1.191     elad      415:                return result;
                    416:
                    417:        switch (req) {
1.193     elad      418:        case KAUTH_REQ_NETWORK_BIND_PORT:
                    419:                result = KAUTH_RESULT_ALLOW;
                    420:                break;
                    421:
1.191     elad      422:        case KAUTH_REQ_NETWORK_SOCKET_DROP: {
                    423:                /* Normal users can only drop their own connections. */
                    424:                struct socket *so = (struct socket *)arg1;
                    425:
1.220     christos  426:                if (so->so_cred && proc_uidmatch(cred, so->so_cred) == 0)
1.191     elad      427:                        result = KAUTH_RESULT_ALLOW;
                    428:
                    429:                break;
                    430:                }
                    431:
                    432:        case KAUTH_REQ_NETWORK_SOCKET_OPEN:
                    433:                /* We allow "raw" routing/bluetooth sockets to anyone. */
1.203     matt      434:                if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_OROUTE
                    435:                    || (u_long)arg1 == PF_BLUETOOTH) {
1.191     elad      436:                        result = KAUTH_RESULT_ALLOW;
1.203     matt      437:                } else {
1.191     elad      438:                        /* Privileged, let secmodel handle this. */
                    439:                        if ((u_long)arg2 == SOCK_RAW)
                    440:                                break;
                    441:                }
                    442:
                    443:                result = KAUTH_RESULT_ALLOW;
                    444:
                    445:                break;
                    446:
1.192     elad      447:        case KAUTH_REQ_NETWORK_SOCKET_CANSEE:
                    448:                result = KAUTH_RESULT_ALLOW;
                    449:
                    450:                break;
                    451:
1.191     elad      452:        default:
                    453:                break;
                    454:        }
                    455:
                    456:        return result;
                    457: }
                    458:
1.119     yamt      459: void
                    460: soinit(void)
                    461: {
                    462:
1.212     pooka     463:        sysctl_kern_socket_setup();
1.178     pooka     464:
1.148     ad        465:        mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.160     ad        466:        softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.136     ad        467:        cv_init(&socurkva_cv, "sokva");
1.205     bouyer    468:        cv_init(&pendfree_thread_cv, "sopendfr");
1.166     ad        469:        soinit2();
1.136     ad        470:
1.119     yamt      471:        /* Set the initial adjusted socket buffer size. */
                    472:        if (sb_max_set(sb_max))
                    473:                panic("bad initial sb_max value: %lu", sb_max);
                    474:
1.191     elad      475:        socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
                    476:            socket_listener_cb, NULL);
1.119     yamt      477: }
                    478:
1.205     bouyer    479: void
                    480: soinit1(void)
                    481: {
                    482:        int error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL,
                    483:            sopendfree_thread, NULL, &sopendfree_lwp, "sopendfree");
                    484:        if (error)
                    485:                panic("soinit1 %d", error);
                    486: }
                    487:
1.1       cgd       488: /*
1.222     rmind     489:  * socreate: create a new socket of the specified type and the protocol.
                    490:  *
                    491:  * => Caller may specify another socket for lock sharing (must not be held).
                    492:  * => Returns the new socket without lock held.
1.224     rmind     493:  */
1.3       andrew    494: int
1.160     ad        495: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l,
                    496:         struct socket *lockso)
1.1       cgd       497: {
1.99      matt      498:        const struct protosw    *prp;
1.54      lukem     499:        struct socket   *so;
1.115     yamt      500:        uid_t           uid;
1.160     ad        501:        int             error;
                    502:        kmutex_t        *lock;
1.1       cgd       503:
1.132     elad      504:        error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
                    505:            KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
                    506:            KAUTH_ARG(proto));
1.140     dyoung    507:        if (error != 0)
                    508:                return error;
1.127     elad      509:
1.1       cgd       510:        if (proto)
                    511:                prp = pffindproto(dom, proto, type);
                    512:        else
                    513:                prp = pffindtype(dom, type);
1.140     dyoung    514:        if (prp == NULL) {
1.120     ginsbach  515:                /* no support for domain */
                    516:                if (pffinddomain(dom) == 0)
1.140     dyoung    517:                        return EAFNOSUPPORT;
1.120     ginsbach  518:                /* no support for socket type */
                    519:                if (proto == 0 && type != 0)
1.140     dyoung    520:                        return EPROTOTYPE;
                    521:                return EPROTONOSUPPORT;
1.120     ginsbach  522:        }
1.223     rmind     523:        if (prp->pr_usrreqs == NULL)
1.140     dyoung    524:                return EPROTONOSUPPORT;
1.1       cgd       525:        if (prp->pr_type != type)
1.140     dyoung    526:                return EPROTOTYPE;
1.160     ad        527:
                    528:        so = soget(true);
1.1       cgd       529:        so->so_type = type;
                    530:        so->so_proto = prp;
1.33      matt      531:        so->so_send = sosend;
                    532:        so->so_receive = soreceive;
1.78      matt      533: #ifdef MBUFTRACE
                    534:        so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
                    535:        so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
                    536:        so->so_mowner = &prp->pr_domain->dom_mowner;
                    537: #endif
1.138     rmind     538:        uid = kauth_cred_geteuid(l->l_cred);
1.115     yamt      539:        so->so_uidinfo = uid_find(uid);
1.168     yamt      540:        so->so_cpid = l->l_proc->p_pid;
1.224     rmind     541:
                    542:        /*
                    543:         * Lock assigned and taken during PCB attach, unless we share
                    544:         * the lock with another socket, e.g. socketpair(2) case.
                    545:         */
                    546:        if (lockso) {
1.160     ad        547:                lock = lockso->so_lock;
                    548:                so->so_lock = lock;
                    549:                mutex_obj_hold(lock);
                    550:                mutex_enter(lock);
                    551:        }
1.224     rmind     552:
                    553:        /* Attach the PCB (returns with the socket lock held). */
                    554:        error = (*prp->pr_usrreqs->pr_attach)(so, proto);
1.160     ad        555:        KASSERT(solocked(so));
1.224     rmind     556:
                    557:        if (error) {
1.222     rmind     558:                KASSERT(so->so_pcb == NULL);
1.1       cgd       559:                so->so_state |= SS_NOFDREF;
                    560:                sofree(so);
1.140     dyoung    561:                return error;
1.1       cgd       562:        }
1.198     elad      563:        so->so_cred = kauth_cred_dup(l->l_cred);
1.160     ad        564:        sounlock(so);
1.224     rmind     565:
1.1       cgd       566:        *aso = so;
1.140     dyoung    567:        return 0;
1.1       cgd       568: }
                    569:
1.222     rmind     570: /*
                    571:  * fsocreate: create a socket and a file descriptor associated with it.
                    572:  *
                    573:  * => On success, write file descriptor to fdout and return zero.
                    574:  * => On failure, return non-zero; *fdout will be undefined.
1.142     dyoung    575:  */
                    576: int
1.222     rmind     577: fsocreate(int domain, struct socket **sop, int type, int proto, int *fdout)
1.142     dyoung    578: {
1.222     rmind     579:        lwp_t *l = curlwp;
                    580:        int error, fd, flags;
                    581:        struct socket *so;
                    582:        struct file *fp;
1.142     dyoung    583:
1.222     rmind     584:        if ((error = fd_allocfile(&fp, &fd)) != 0) {
1.204     christos  585:                return error;
1.222     rmind     586:        }
                    587:        flags = type & SOCK_FLAGS_MASK;
1.204     christos  588:        fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0);
1.207     christos  589:        fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
                    590:            ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
1.142     dyoung    591:        fp->f_type = DTYPE_SOCKET;
                    592:        fp->f_ops = &socketops;
1.222     rmind     593:
                    594:        type &= ~SOCK_FLAGS_MASK;
                    595:        error = socreate(domain, &so, type, proto, l, NULL);
                    596:        if (error) {
1.155     ad        597:                fd_abort(curproc, fp, fd);
1.222     rmind     598:                return error;
                    599:        }
                    600:        if (flags & SOCK_NONBLOCK) {
                    601:                so->so_state |= SS_NBIO;
                    602:        }
                    603:        fp->f_data = so;
                    604:        fd_affix(curproc, fp, fd);
                    605:
                    606:        if (sop != NULL) {
                    607:                *sop = so;
1.142     dyoung    608:        }
1.222     rmind     609:        *fdout = fd;
1.142     dyoung    610:        return error;
                    611: }
                    612:
1.3       andrew    613: int
1.190     dyoung    614: sofamily(const struct socket *so)
                    615: {
                    616:        const struct protosw *pr;
                    617:        const struct domain *dom;
                    618:
                    619:        if ((pr = so->so_proto) == NULL)
                    620:                return AF_UNSPEC;
                    621:        if ((dom = pr->pr_domain) == NULL)
                    622:                return AF_UNSPEC;
                    623:        return dom->dom_family;
                    624: }
                    625:
                    626: int
1.114     christos  627: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       628: {
1.160     ad        629:        int     error;
1.1       cgd       630:
1.160     ad        631:        solock(so);
1.231     rtr       632:        error = (*so->so_proto->pr_usrreqs->pr_bind)(so, nam, l);
1.160     ad        633:        sounlock(so);
1.140     dyoung    634:        return error;
1.1       cgd       635: }
                    636:
1.3       andrew    637: int
1.150     elad      638: solisten(struct socket *so, int backlog, struct lwp *l)
1.1       cgd       639: {
1.160     ad        640:        int     error;
1.1       cgd       641:
1.160     ad        642:        solock(so);
1.158     ad        643:        if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
1.163     ad        644:            SS_ISDISCONNECTING)) != 0) {
1.222     rmind     645:                sounlock(so);
                    646:                return EINVAL;
1.163     ad        647:        }
1.231     rtr       648:        error = (*so->so_proto->pr_usrreqs->pr_listen)(so, l);
1.140     dyoung    649:        if (error != 0) {
1.160     ad        650:                sounlock(so);
1.140     dyoung    651:                return error;
1.1       cgd       652:        }
1.63      matt      653:        if (TAILQ_EMPTY(&so->so_q))
1.1       cgd       654:                so->so_options |= SO_ACCEPTCONN;
                    655:        if (backlog < 0)
                    656:                backlog = 0;
1.49      jonathan  657:        so->so_qlimit = min(backlog, somaxconn);
1.160     ad        658:        sounlock(so);
1.140     dyoung    659:        return 0;
1.1       cgd       660: }
                    661:
1.21      christos  662: void
1.54      lukem     663: sofree(struct socket *so)
1.1       cgd       664: {
1.161     ad        665:        u_int refs;
1.1       cgd       666:
1.160     ad        667:        KASSERT(solocked(so));
                    668:
                    669:        if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) {
                    670:                sounlock(so);
1.1       cgd       671:                return;
1.160     ad        672:        }
1.43      mycroft   673:        if (so->so_head) {
                    674:                /*
                    675:                 * We must not decommission a socket that's on the accept(2)
                    676:                 * queue.  If we do, then accept(2) may hang after select(2)
                    677:                 * indicated that the listening socket was ready.
                    678:                 */
1.160     ad        679:                if (!soqremque(so, 0)) {
                    680:                        sounlock(so);
1.43      mycroft   681:                        return;
1.160     ad        682:                }
1.43      mycroft   683:        }
1.98      christos  684:        if (so->so_rcv.sb_hiwat)
1.110     christos  685:                (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98      christos  686:                    RLIM_INFINITY);
                    687:        if (so->so_snd.sb_hiwat)
1.110     christos  688:                (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98      christos  689:                    RLIM_INFINITY);
                    690:        sbrelease(&so->so_snd, so);
1.160     ad        691:        KASSERT(!cv_has_waiters(&so->so_cv));
                    692:        KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv));
                    693:        KASSERT(!cv_has_waiters(&so->so_snd.sb_cv));
1.1       cgd       694:        sorflush(so);
1.161     ad        695:        refs = so->so_aborting; /* XXX */
1.177     ad        696:        /* Remove acccept filter if one is present. */
1.170     tls       697:        if (so->so_accf != NULL)
1.177     ad        698:                (void)accept_filt_clear(so);
1.160     ad        699:        sounlock(so);
1.161     ad        700:        if (refs == 0)          /* XXX */
                    701:                soput(so);
1.1       cgd       702: }
                    703:
                    704: /*
1.222     rmind     705:  * soclose: close a socket on last file table reference removal.
                    706:  * Initiate disconnect if connected.  Free socket when disconnect complete.
1.1       cgd       707:  */
1.3       andrew    708: int
1.54      lukem     709: soclose(struct socket *so)
1.1       cgd       710: {
1.222     rmind     711:        struct socket *so2;
                    712:        int error = 0;
1.1       cgd       713:
1.160     ad        714:        solock(so);
1.1       cgd       715:        if (so->so_options & SO_ACCEPTCONN) {
1.172     ad        716:                for (;;) {
                    717:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.160     ad        718:                                KASSERT(solocked2(so, so2));
                    719:                                (void) soqremque(so2, 0);
                    720:                                /* soabort drops the lock. */
                    721:                                (void) soabort(so2);
                    722:                                solock(so);
1.172     ad        723:                                continue;
1.160     ad        724:                        }
1.172     ad        725:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.160     ad        726:                                KASSERT(solocked2(so, so2));
                    727:                                (void) soqremque(so2, 1);
                    728:                                /* soabort drops the lock. */
                    729:                                (void) soabort(so2);
                    730:                                solock(so);
1.172     ad        731:                                continue;
1.160     ad        732:                        }
1.172     ad        733:                        break;
                    734:                }
1.1       cgd       735:        }
1.222     rmind     736:        if (so->so_pcb == NULL)
1.1       cgd       737:                goto discard;
                    738:        if (so->so_state & SS_ISCONNECTED) {
                    739:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    740:                        error = sodisconnect(so);
                    741:                        if (error)
                    742:                                goto drop;
                    743:                }
                    744:                if (so->so_options & SO_LINGER) {
1.206     christos  745:                        if ((so->so_state & (SS_ISDISCONNECTING|SS_NBIO)) ==
                    746:                            (SS_ISDISCONNECTING|SS_NBIO))
1.1       cgd       747:                                goto drop;
1.21      christos  748:                        while (so->so_state & SS_ISCONNECTED) {
1.185     yamt      749:                                error = sowait(so, true, so->so_linger * hz);
1.21      christos  750:                                if (error)
1.1       cgd       751:                                        break;
1.21      christos  752:                        }
1.1       cgd       753:                }
                    754:        }
1.54      lukem     755:  drop:
1.1       cgd       756:        if (so->so_pcb) {
1.224     rmind     757:                KASSERT(solocked(so));
                    758:                (*so->so_proto->pr_usrreqs->pr_detach)(so);
1.1       cgd       759:        }
1.54      lukem     760:  discard:
1.222     rmind     761:        KASSERT((so->so_state & SS_NOFDREF) == 0);
1.198     elad      762:        kauth_cred_free(so->so_cred);
1.1       cgd       763:        so->so_state |= SS_NOFDREF;
                    764:        sofree(so);
1.222     rmind     765:        return error;
1.1       cgd       766: }
                    767:
                    768: /*
1.160     ad        769:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       770:  */
1.3       andrew    771: int
1.54      lukem     772: soabort(struct socket *so)
1.1       cgd       773: {
1.161     ad        774:        u_int refs;
1.139     yamt      775:        int error;
1.160     ad        776:
                    777:        KASSERT(solocked(so));
                    778:        KASSERT(so->so_head == NULL);
1.1       cgd       779:
1.161     ad        780:        so->so_aborting++;              /* XXX */
1.230     mrg       781:        error = (*so->so_proto->pr_usrreqs->pr_abort)(so);
1.161     ad        782:        refs = --so->so_aborting;       /* XXX */
1.164     drochner  783:        if (error || (refs == 0)) {
1.139     yamt      784:                sofree(so);
1.160     ad        785:        } else {
                    786:                sounlock(so);
1.139     yamt      787:        }
                    788:        return error;
1.1       cgd       789: }
                    790:
1.3       andrew    791: int
1.54      lukem     792: soaccept(struct socket *so, struct mbuf *nam)
1.1       cgd       793: {
1.222     rmind     794:        int error;
1.160     ad        795:
                    796:        KASSERT(solocked(so));
1.222     rmind     797:        KASSERT((so->so_state & SS_NOFDREF) != 0);
1.1       cgd       798:
                    799:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   800:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    801:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.225     rtr       802:                error = (*so->so_proto->pr_usrreqs->pr_accept)(so, nam);
1.41      mycroft   803:        else
1.53      itojun    804:                error = ECONNABORTED;
1.52      itojun    805:
1.222     rmind     806:        return error;
1.1       cgd       807: }
                    808:
1.3       andrew    809: int
1.114     christos  810: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       811: {
1.222     rmind     812:        int error;
1.160     ad        813:
                    814:        KASSERT(solocked(so));
1.1       cgd       815:
                    816:        if (so->so_options & SO_ACCEPTCONN)
1.222     rmind     817:                return EOPNOTSUPP;
1.1       cgd       818:        /*
                    819:         * If protocol is connection-based, can only connect once.
                    820:         * Otherwise, if connected, try to disconnect first.
                    821:         * This allows user to disconnect by connecting to, e.g.,
                    822:         * a null address.
                    823:         */
                    824:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    825:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
                    826:            (error = sodisconnect(so))))
                    827:                error = EISCONN;
                    828:        else
1.231     rtr       829:                error = (*so->so_proto->pr_usrreqs->pr_connect)(so, nam, l);
1.222     rmind     830:
                    831:        return error;
1.1       cgd       832: }
                    833:
1.3       andrew    834: int
1.54      lukem     835: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       836: {
1.160     ad        837:        KASSERT(solocked2(so1, so2));
1.1       cgd       838:
1.223     rmind     839:        return (*so1->so_proto->pr_usrreqs->pr_generic)(so1,
                    840:            PRU_CONNECT2, NULL, (struct mbuf *)so2, NULL, NULL);
1.1       cgd       841: }
                    842:
1.3       andrew    843: int
1.54      lukem     844: sodisconnect(struct socket *so)
1.1       cgd       845: {
1.160     ad        846:        int     error;
                    847:
                    848:        KASSERT(solocked(so));
1.1       cgd       849:
                    850:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    851:                error = ENOTCONN;
1.160     ad        852:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       853:                error = EALREADY;
1.160     ad        854:        } else {
1.229     rtr       855:                error = (*so->so_proto->pr_usrreqs->pr_disconnect)(so);
1.1       cgd       856:        }
                    857:        return (error);
                    858: }
                    859:
1.15      mycroft   860: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       861: /*
                    862:  * Send on a socket.
                    863:  * If send must go all at once and message is larger than
                    864:  * send buffering, then hard error.
                    865:  * Lock against other senders.
                    866:  * If must go all at once and not enough room now, then
                    867:  * inform user that this would block and do nothing.
                    868:  * Otherwise, if nonblocking, send as much as possible.
                    869:  * The data to be sent is described by "uio" if nonzero,
                    870:  * otherwise by the mbuf chain "top" (which must be null
                    871:  * if uio is not).  Data provided in mbuf chain must be small
                    872:  * enough to send all at once.
                    873:  *
                    874:  * Returns nonzero on error, timeout or signal; callers
                    875:  * must check for short counts if EINTR/ERESTART are returned.
                    876:  * Data and control buffers are freed on return.
                    877:  */
1.3       andrew    878: int
1.54      lukem     879: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114     christos  880:        struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       881: {
1.54      lukem     882:        struct mbuf     **mp, *m;
1.58      jdolecek  883:        long            space, len, resid, clen, mlen;
                    884:        int             error, s, dontroute, atomic;
1.196     dsl       885:        short           wakeup_state = 0;
1.54      lukem     886:
1.160     ad        887:        clen = 0;
1.64      thorpej   888:
1.160     ad        889:        /*
                    890:         * solock() provides atomicity of access.  splsoftnet() prevents
                    891:         * protocol processing soft interrupts from interrupting us and
                    892:         * blocking (expensive).
                    893:         */
                    894:        s = splsoftnet();
                    895:        solock(so);
1.54      lukem     896:        atomic = sosendallatonce(so) || top;
1.1       cgd       897:        if (uio)
                    898:                resid = uio->uio_resid;
                    899:        else
                    900:                resid = top->m_pkthdr.len;
1.7       cgd       901:        /*
                    902:         * In theory resid should be unsigned.
                    903:         * However, space must be signed, as it might be less than 0
                    904:         * if we over-committed, and we must use a signed comparison
                    905:         * of space and resid.  On the other hand, a negative resid
                    906:         * causes us to loop sending 0-length segments to the protocol.
                    907:         */
1.29      mycroft   908:        if (resid < 0) {
                    909:                error = EINVAL;
                    910:                goto out;
                    911:        }
1.1       cgd       912:        dontroute =
                    913:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    914:            (so->so_proto->pr_flags & PR_ATOMIC);
1.165     christos  915:        l->l_ru.ru_msgsnd++;
1.1       cgd       916:        if (control)
                    917:                clen = control->m_len;
1.54      lukem     918:  restart:
1.21      christos  919:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       920:                goto out;
                    921:        do {
1.160     ad        922:                if (so->so_state & SS_CANTSENDMORE) {
                    923:                        error = EPIPE;
                    924:                        goto release;
                    925:                }
1.48      thorpej   926:                if (so->so_error) {
                    927:                        error = so->so_error;
                    928:                        so->so_error = 0;
                    929:                        goto release;
                    930:                }
1.1       cgd       931:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    932:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
1.217     rmind     933:                                if (resid || clen == 0) {
1.160     ad        934:                                        error = ENOTCONN;
                    935:                                        goto release;
                    936:                                }
                    937:                        } else if (addr == 0) {
                    938:                                error = EDESTADDRREQ;
                    939:                                goto release;
                    940:                        }
1.1       cgd       941:                }
                    942:                space = sbspace(&so->so_snd);
                    943:                if (flags & MSG_OOB)
                    944:                        space += 1024;
1.21      christos  945:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        946:                    clen > so->so_snd.sb_hiwat) {
                    947:                        error = EMSGSIZE;
                    948:                        goto release;
                    949:                }
1.96      mycroft   950:                if (space < resid + clen &&
1.1       cgd       951:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.206     christos  952:                        if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) {
1.160     ad        953:                                error = EWOULDBLOCK;
                    954:                                goto release;
                    955:                        }
1.1       cgd       956:                        sbunlock(&so->so_snd);
1.196     dsl       957:                        if (wakeup_state & SS_RESTARTSYS) {
                    958:                                error = ERESTART;
                    959:                                goto out;
                    960:                        }
1.1       cgd       961:                        error = sbwait(&so->so_snd);
                    962:                        if (error)
                    963:                                goto out;
1.196     dsl       964:                        wakeup_state = so->so_state;
1.1       cgd       965:                        goto restart;
                    966:                }
1.196     dsl       967:                wakeup_state = 0;
1.1       cgd       968:                mp = &top;
                    969:                space -= clen;
                    970:                do {
1.45      tv        971:                        if (uio == NULL) {
                    972:                                /*
                    973:                                 * Data is prepackaged in "top".
                    974:                                 */
                    975:                                resid = 0;
                    976:                                if (flags & MSG_EOR)
                    977:                                        top->m_flags |= M_EOR;
                    978:                        } else do {
1.160     ad        979:                                sounlock(so);
                    980:                                splx(s);
1.144     dyoung    981:                                if (top == NULL) {
1.78      matt      982:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv        983:                                        mlen = MHLEN;
                    984:                                        m->m_pkthdr.len = 0;
1.140     dyoung    985:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv        986:                                } else {
1.78      matt      987:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv        988:                                        mlen = MLEN;
                    989:                                }
1.78      matt      990:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt      991:                                if (sock_loan_thresh >= 0 &&
                    992:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                    993:                                    space >= sock_loan_thresh &&
1.64      thorpej   994:                                    (len = sosend_loan(so, uio, m,
                    995:                                                       space)) != 0) {
                    996:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                    997:                                        space -= len;
                    998:                                        goto have_data;
                    999:                                }
1.45      tv       1000:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej  1001:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.201     oki      1002:                                        m_clget(m, M_DONTWAIT);
1.45      tv       1003:                                        if ((m->m_flags & M_EXT) == 0)
                   1004:                                                goto nopages;
                   1005:                                        mlen = MCLBYTES;
                   1006:                                        if (atomic && top == 0) {
1.58      jdolecek 1007:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem    1008:                                                    resid);
1.45      tv       1009:                                                m->m_data += max_hdr;
                   1010:                                        } else
1.58      jdolecek 1011:                                                len = lmin(MCLBYTES, resid);
1.45      tv       1012:                                        space -= len;
                   1013:                                } else {
1.64      thorpej  1014:  nopages:
                   1015:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek 1016:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv       1017:                                        space -= len;
                   1018:                                        /*
                   1019:                                         * For datagram protocols, leave room
                   1020:                                         * for protocol headers in first mbuf.
                   1021:                                         */
                   1022:                                        if (atomic && top == 0 && len < mlen)
                   1023:                                                MH_ALIGN(m, len);
                   1024:                                }
1.144     dyoung   1025:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej  1026:  have_data:
1.45      tv       1027:                                resid = uio->uio_resid;
                   1028:                                m->m_len = len;
                   1029:                                *mp = m;
                   1030:                                top->m_pkthdr.len += len;
1.160     ad       1031:                                s = splsoftnet();
                   1032:                                solock(so);
1.144     dyoung   1033:                                if (error != 0)
1.45      tv       1034:                                        goto release;
                   1035:                                mp = &m->m_next;
                   1036:                                if (resid <= 0) {
                   1037:                                        if (flags & MSG_EOR)
                   1038:                                                top->m_flags |= M_EOR;
                   1039:                                        break;
                   1040:                                }
                   1041:                        } while (space > 0 && atomic);
1.108     perry    1042:
1.160     ad       1043:                        if (so->so_state & SS_CANTSENDMORE) {
                   1044:                                error = EPIPE;
                   1045:                                goto release;
                   1046:                        }
1.45      tv       1047:                        if (dontroute)
                   1048:                                so->so_options |= SO_DONTROUTE;
                   1049:                        if (resid > 0)
                   1050:                                so->so_state |= SS_MORETOCOME;
1.226     rtr      1051:                        if (flags & MSG_OOB)
                   1052:                                error = (*so->so_proto->pr_usrreqs->pr_sendoob)(so,
                   1053:                                    top, control);
                   1054:                        else
1.232     rtr      1055:                                error = (*so->so_proto->pr_usrreqs->pr_send)(so,
                   1056:                                    top, addr, control, l);
1.45      tv       1057:                        if (dontroute)
                   1058:                                so->so_options &= ~SO_DONTROUTE;
                   1059:                        if (resid > 0)
                   1060:                                so->so_state &= ~SS_MORETOCOME;
                   1061:                        clen = 0;
1.144     dyoung   1062:                        control = NULL;
                   1063:                        top = NULL;
1.45      tv       1064:                        mp = &top;
1.144     dyoung   1065:                        if (error != 0)
1.1       cgd      1066:                                goto release;
                   1067:                } while (resid && space > 0);
                   1068:        } while (resid);
                   1069:
1.54      lukem    1070:  release:
1.1       cgd      1071:        sbunlock(&so->so_snd);
1.54      lukem    1072:  out:
1.160     ad       1073:        sounlock(so);
                   1074:        splx(s);
1.1       cgd      1075:        if (top)
                   1076:                m_freem(top);
                   1077:        if (control)
                   1078:                m_freem(control);
                   1079:        return (error);
                   1080: }
                   1081:
                   1082: /*
1.159     ad       1083:  * Following replacement or removal of the first mbuf on the first
                   1084:  * mbuf chain of a socket buffer, push necessary state changes back
                   1085:  * into the socket buffer so that other consumers see the values
                   1086:  * consistently.  'nextrecord' is the callers locally stored value of
                   1087:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1088:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1089:  */
                   1090: static void
                   1091: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1092: {
                   1093:
1.160     ad       1094:        KASSERT(solocked(sb->sb_so));
                   1095:
1.159     ad       1096:        /*
                   1097:         * First, update for the new value of nextrecord.  If necessary,
                   1098:         * make it the first record.
                   1099:         */
                   1100:        if (sb->sb_mb != NULL)
                   1101:                sb->sb_mb->m_nextpkt = nextrecord;
                   1102:        else
                   1103:                sb->sb_mb = nextrecord;
                   1104:
                   1105:         /*
                   1106:          * Now update any dependent socket buffer fields to reflect
                   1107:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1108:          * the addition of a second clause that takes care of the
                   1109:          * case where sb_mb has been updated, but remains the last
                   1110:          * record.
                   1111:          */
                   1112:         if (sb->sb_mb == NULL) {
                   1113:                 sb->sb_mbtail = NULL;
                   1114:                 sb->sb_lastrecord = NULL;
                   1115:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1116:                 sb->sb_lastrecord = sb->sb_mb;
                   1117: }
                   1118:
                   1119: /*
1.1       cgd      1120:  * Implement receive operations on a socket.
                   1121:  * We depend on the way that records are added to the sockbuf
                   1122:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1123:  * must begin with an address if the protocol so specifies,
                   1124:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1125:  * and then zero or more mbufs of data.
                   1126:  * In order to avoid blocking network interrupts for the entire time here,
                   1127:  * we splx() while doing the actual copy to user space.
                   1128:  * Although the sockbuf is locked, new data may still be appended,
                   1129:  * and thus we must maintain consistency of the sockbuf during that time.
                   1130:  *
                   1131:  * The caller may receive the data as a single mbuf chain by supplying
                   1132:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1133:  * only for the count in uio_resid.
                   1134:  */
1.3       andrew   1135: int
1.54      lukem    1136: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1137:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1138: {
1.116     yamt     1139:        struct lwp *l = curlwp;
1.160     ad       1140:        struct mbuf     *m, **mp, *mt;
1.211     chs      1141:        size_t len, offset, moff, orig_resid;
                   1142:        int atomic, flags, error, s, type;
1.99      matt     1143:        const struct protosw    *pr;
1.54      lukem    1144:        struct mbuf     *nextrecord;
1.67      he       1145:        int             mbuf_removed = 0;
1.146     dyoung   1146:        const struct domain *dom;
1.196     dsl      1147:        short           wakeup_state = 0;
1.64      thorpej  1148:
1.54      lukem    1149:        pr = so->so_proto;
1.146     dyoung   1150:        atomic = pr->pr_flags & PR_ATOMIC;
                   1151:        dom = pr->pr_domain;
1.1       cgd      1152:        mp = mp0;
1.54      lukem    1153:        type = 0;
                   1154:        orig_resid = uio->uio_resid;
1.102     jonathan 1155:
1.144     dyoung   1156:        if (paddr != NULL)
                   1157:                *paddr = NULL;
                   1158:        if (controlp != NULL)
                   1159:                *controlp = NULL;
                   1160:        if (flagsp != NULL)
1.1       cgd      1161:                flags = *flagsp &~ MSG_EOR;
                   1162:        else
                   1163:                flags = 0;
1.66      enami    1164:
1.1       cgd      1165:        if (flags & MSG_OOB) {
                   1166:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1167:                solock(so);
1.226     rtr      1168:                error = (*pr->pr_usrreqs->pr_recvoob)(so, m, flags & MSG_PEEK);
1.160     ad       1169:                sounlock(so);
1.1       cgd      1170:                if (error)
                   1171:                        goto bad;
                   1172:                do {
1.134     christos 1173:                        error = uiomove(mtod(m, void *),
1.211     chs      1174:                            MIN(uio->uio_resid, m->m_len), uio);
1.1       cgd      1175:                        m = m_free(m);
1.144     dyoung   1176:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1177:  bad:
1.144     dyoung   1178:                if (m != NULL)
1.1       cgd      1179:                        m_freem(m);
1.144     dyoung   1180:                return error;
1.1       cgd      1181:        }
1.144     dyoung   1182:        if (mp != NULL)
1.140     dyoung   1183:                *mp = NULL;
1.160     ad       1184:
                   1185:        /*
                   1186:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1187:         * protocol processing soft interrupts from interrupting us and
                   1188:         * blocking (expensive).
                   1189:         */
                   1190:        s = splsoftnet();
                   1191:        solock(so);
1.54      lukem    1192:  restart:
1.160     ad       1193:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1194:                sounlock(so);
                   1195:                splx(s);
1.144     dyoung   1196:                return error;
1.160     ad       1197:        }
1.1       cgd      1198:
                   1199:        m = so->so_rcv.sb_mb;
                   1200:        /*
                   1201:         * If we have less data than requested, block awaiting more
                   1202:         * (subject to any timeout) if:
1.15      mycroft  1203:         *   1. the current count is less than the low water mark,
1.1       cgd      1204:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1205:         *      receive operation at once if we block (resid <= hiwat), or
                   1206:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1207:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1208:         * we have to do the receive in sections, and thus risk returning
                   1209:         * a short count if a timeout or signal occurs after we start.
                   1210:         */
1.144     dyoung   1211:        if (m == NULL ||
                   1212:            ((flags & MSG_DONTWAIT) == 0 &&
                   1213:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1214:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1215:              ((flags & MSG_WAITALL) &&
                   1216:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1217:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1218: #ifdef DIAGNOSTIC
1.144     dyoung   1219:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1220:                        panic("receive 1");
                   1221: #endif
                   1222:                if (so->so_error) {
1.144     dyoung   1223:                        if (m != NULL)
1.15      mycroft  1224:                                goto dontblock;
1.1       cgd      1225:                        error = so->so_error;
                   1226:                        if ((flags & MSG_PEEK) == 0)
                   1227:                                so->so_error = 0;
                   1228:                        goto release;
                   1229:                }
                   1230:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1231:                        if (m != NULL)
1.15      mycroft  1232:                                goto dontblock;
1.1       cgd      1233:                        else
                   1234:                                goto release;
                   1235:                }
1.144     dyoung   1236:                for (; m != NULL; m = m->m_next)
1.1       cgd      1237:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1238:                                m = so->so_rcv.sb_mb;
                   1239:                                goto dontblock;
                   1240:                        }
                   1241:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1242:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1243:                        error = ENOTCONN;
                   1244:                        goto release;
                   1245:                }
                   1246:                if (uio->uio_resid == 0)
                   1247:                        goto release;
1.206     christos 1248:                if ((so->so_state & SS_NBIO) ||
                   1249:                    (flags & (MSG_DONTWAIT|MSG_NBIO))) {
1.1       cgd      1250:                        error = EWOULDBLOCK;
                   1251:                        goto release;
                   1252:                }
1.69      thorpej  1253:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1254:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1255:                sbunlock(&so->so_rcv);
1.196     dsl      1256:                if (wakeup_state & SS_RESTARTSYS)
                   1257:                        error = ERESTART;
                   1258:                else
                   1259:                        error = sbwait(&so->so_rcv);
1.160     ad       1260:                if (error != 0) {
                   1261:                        sounlock(so);
                   1262:                        splx(s);
1.144     dyoung   1263:                        return error;
1.160     ad       1264:                }
1.196     dsl      1265:                wakeup_state = so->so_state;
1.1       cgd      1266:                goto restart;
                   1267:        }
1.54      lukem    1268:  dontblock:
1.69      thorpej  1269:        /*
                   1270:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1271:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1272:         * pointer to the next record in the socket buffer.  We must keep the
                   1273:         * various socket buffer pointers and local stack versions of the
                   1274:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1275:         * socket lock, and re-reading them when picking it up.
1.159     ad       1276:         *
                   1277:         * Otherwise, we will race with the network stack appending new data
                   1278:         * or records onto the socket buffer by using inconsistent/stale
                   1279:         * versions of the field, possibly resulting in socket buffer
                   1280:         * corruption.
                   1281:         *
                   1282:         * By holding the high-level sblock(), we prevent simultaneous
                   1283:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1284:         */
1.144     dyoung   1285:        if (l != NULL)
1.157     ad       1286:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1287:        KASSERT(m == so->so_rcv.sb_mb);
                   1288:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1289:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1290:        nextrecord = m->m_nextpkt;
                   1291:        if (pr->pr_flags & PR_ADDR) {
                   1292: #ifdef DIAGNOSTIC
                   1293:                if (m->m_type != MT_SONAME)
                   1294:                        panic("receive 1a");
                   1295: #endif
1.3       andrew   1296:                orig_resid = 0;
1.1       cgd      1297:                if (flags & MSG_PEEK) {
                   1298:                        if (paddr)
                   1299:                                *paddr = m_copy(m, 0, m->m_len);
                   1300:                        m = m->m_next;
                   1301:                } else {
                   1302:                        sbfree(&so->so_rcv, m);
1.67      he       1303:                        mbuf_removed = 1;
1.144     dyoung   1304:                        if (paddr != NULL) {
1.1       cgd      1305:                                *paddr = m;
                   1306:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1307:                                m->m_next = NULL;
1.1       cgd      1308:                                m = so->so_rcv.sb_mb;
                   1309:                        } else {
                   1310:                                MFREE(m, so->so_rcv.sb_mb);
                   1311:                                m = so->so_rcv.sb_mb;
                   1312:                        }
1.159     ad       1313:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1314:                }
                   1315:        }
1.159     ad       1316:
                   1317:        /*
                   1318:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1319:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1320:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1321:         * perform externalization (or freeing if controlp == NULL).
                   1322:         */
                   1323:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1324:                struct mbuf *cm = NULL, *cmn;
                   1325:                struct mbuf **cme = &cm;
                   1326:
                   1327:                do {
                   1328:                        if (flags & MSG_PEEK) {
                   1329:                                if (controlp != NULL) {
                   1330:                                        *controlp = m_copy(m, 0, m->m_len);
                   1331:                                        controlp = &(*controlp)->m_next;
                   1332:                                }
                   1333:                                m = m->m_next;
                   1334:                        } else {
                   1335:                                sbfree(&so->so_rcv, m);
1.1       cgd      1336:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1337:                                m->m_next = NULL;
1.159     ad       1338:                                *cme = m;
                   1339:                                cme = &(*cme)->m_next;
1.1       cgd      1340:                                m = so->so_rcv.sb_mb;
1.159     ad       1341:                        }
                   1342:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1343:                if ((flags & MSG_PEEK) == 0)
                   1344:                        sbsync(&so->so_rcv, nextrecord);
                   1345:                for (; cm != NULL; cm = cmn) {
                   1346:                        cmn = cm->m_next;
                   1347:                        cm->m_next = NULL;
                   1348:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1349:                        if (controlp != NULL) {
                   1350:                                if (dom->dom_externalize != NULL &&
                   1351:                                    type == SCM_RIGHTS) {
1.160     ad       1352:                                        sounlock(so);
1.159     ad       1353:                                        splx(s);
1.204     christos 1354:                                        error = (*dom->dom_externalize)(cm, l,
                   1355:                                            (flags & MSG_CMSG_CLOEXEC) ?
                   1356:                                            O_CLOEXEC : 0);
1.159     ad       1357:                                        s = splsoftnet();
1.160     ad       1358:                                        solock(so);
1.159     ad       1359:                                }
                   1360:                                *controlp = cm;
                   1361:                                while (*controlp != NULL)
                   1362:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1363:                        } else {
1.106     itojun   1364:                                /*
                   1365:                                 * Dispose of any SCM_RIGHTS message that went
                   1366:                                 * through the read path rather than recv.
                   1367:                                 */
1.159     ad       1368:                                if (dom->dom_dispose != NULL &&
                   1369:                                    type == SCM_RIGHTS) {
1.160     ad       1370:                                        sounlock(so);
1.159     ad       1371:                                        (*dom->dom_dispose)(cm);
1.160     ad       1372:                                        solock(so);
1.159     ad       1373:                                }
                   1374:                                m_freem(cm);
1.1       cgd      1375:                        }
                   1376:                }
1.159     ad       1377:                if (m != NULL)
                   1378:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1379:                else
                   1380:                        nextrecord = so->so_rcv.sb_mb;
                   1381:                orig_resid = 0;
1.1       cgd      1382:        }
1.69      thorpej  1383:
1.159     ad       1384:        /* If m is non-NULL, we have some data to read. */
                   1385:        if (__predict_true(m != NULL)) {
1.1       cgd      1386:                type = m->m_type;
                   1387:                if (type == MT_OOBDATA)
                   1388:                        flags |= MSG_OOB;
                   1389:        }
1.69      thorpej  1390:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1391:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1392:
1.1       cgd      1393:        moff = 0;
                   1394:        offset = 0;
1.144     dyoung   1395:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1396:                if (m->m_type == MT_OOBDATA) {
                   1397:                        if (type != MT_OOBDATA)
                   1398:                                break;
                   1399:                } else if (type == MT_OOBDATA)
                   1400:                        break;
                   1401: #ifdef DIAGNOSTIC
                   1402:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1403:                        panic("receive 3");
                   1404: #endif
                   1405:                so->so_state &= ~SS_RCVATMARK;
1.196     dsl      1406:                wakeup_state = 0;
1.1       cgd      1407:                len = uio->uio_resid;
                   1408:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1409:                        len = so->so_oobmark - offset;
                   1410:                if (len > m->m_len - moff)
                   1411:                        len = m->m_len - moff;
                   1412:                /*
                   1413:                 * If mp is set, just pass back the mbufs.
                   1414:                 * Otherwise copy them out via the uio, then free.
                   1415:                 * Sockbuf must be consistent here (points to current mbuf,
                   1416:                 * it points to next record) when we drop priority;
                   1417:                 * we must note any additions to the sockbuf when we
                   1418:                 * block interrupts again.
                   1419:                 */
1.144     dyoung   1420:                if (mp == NULL) {
1.69      thorpej  1421:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1422:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1423:                        sounlock(so);
1.1       cgd      1424:                        splx(s);
1.211     chs      1425:                        error = uiomove(mtod(m, char *) + moff, len, uio);
1.20      mycroft  1426:                        s = splsoftnet();
1.160     ad       1427:                        solock(so);
1.144     dyoung   1428:                        if (error != 0) {
1.67      he       1429:                                /*
                   1430:                                 * If any part of the record has been removed
                   1431:                                 * (such as the MT_SONAME mbuf, which will
                   1432:                                 * happen when PR_ADDR, and thus also
                   1433:                                 * PR_ATOMIC, is set), then drop the entire
                   1434:                                 * record to maintain the atomicity of the
                   1435:                                 * receive operation.
                   1436:                                 *
                   1437:                                 * This avoids a later panic("receive 1a")
                   1438:                                 * when compiled with DIAGNOSTIC.
                   1439:                                 */
1.146     dyoung   1440:                                if (m && mbuf_removed && atomic)
1.67      he       1441:                                        (void) sbdroprecord(&so->so_rcv);
                   1442:
1.57      jdolecek 1443:                                goto release;
1.67      he       1444:                        }
1.1       cgd      1445:                } else
                   1446:                        uio->uio_resid -= len;
                   1447:                if (len == m->m_len - moff) {
                   1448:                        if (m->m_flags & M_EOR)
                   1449:                                flags |= MSG_EOR;
                   1450:                        if (flags & MSG_PEEK) {
                   1451:                                m = m->m_next;
                   1452:                                moff = 0;
                   1453:                        } else {
                   1454:                                nextrecord = m->m_nextpkt;
                   1455:                                sbfree(&so->so_rcv, m);
                   1456:                                if (mp) {
                   1457:                                        *mp = m;
                   1458:                                        mp = &m->m_next;
                   1459:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1460:                                        *mp = NULL;
1.1       cgd      1461:                                } else {
                   1462:                                        MFREE(m, so->so_rcv.sb_mb);
                   1463:                                        m = so->so_rcv.sb_mb;
                   1464:                                }
1.69      thorpej  1465:                                /*
                   1466:                                 * If m != NULL, we also know that
                   1467:                                 * so->so_rcv.sb_mb != NULL.
                   1468:                                 */
                   1469:                                KASSERT(so->so_rcv.sb_mb == m);
                   1470:                                if (m) {
1.1       cgd      1471:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1472:                                        if (nextrecord == NULL)
                   1473:                                                so->so_rcv.sb_lastrecord = m;
                   1474:                                } else {
                   1475:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1476:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1477:                                }
                   1478:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1479:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1480:                        }
1.144     dyoung   1481:                } else if (flags & MSG_PEEK)
                   1482:                        moff += len;
                   1483:                else {
1.160     ad       1484:                        if (mp != NULL) {
                   1485:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1486:                                if (__predict_false(mt == NULL)) {
                   1487:                                        sounlock(so);
                   1488:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1489:                                        solock(so);
                   1490:                                }
                   1491:                                *mp = mt;
                   1492:                        }
1.144     dyoung   1493:                        m->m_data += len;
                   1494:                        m->m_len -= len;
                   1495:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1496:                }
                   1497:                if (so->so_oobmark) {
                   1498:                        if ((flags & MSG_PEEK) == 0) {
                   1499:                                so->so_oobmark -= len;
                   1500:                                if (so->so_oobmark == 0) {
                   1501:                                        so->so_state |= SS_RCVATMARK;
                   1502:                                        break;
                   1503:                                }
1.7       cgd      1504:                        } else {
1.1       cgd      1505:                                offset += len;
1.7       cgd      1506:                                if (offset == so->so_oobmark)
                   1507:                                        break;
                   1508:                        }
1.1       cgd      1509:                }
                   1510:                if (flags & MSG_EOR)
                   1511:                        break;
                   1512:                /*
                   1513:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1514:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1515:                 * termination.  If a signal/timeout occurs, return
                   1516:                 * with a short count but without error.
                   1517:                 * Keep sockbuf locked against other readers.
                   1518:                 */
1.144     dyoung   1519:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1520:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1521:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1522:                                break;
1.68      matt     1523:                        /*
                   1524:                         * If we are peeking and the socket receive buffer is
                   1525:                         * full, stop since we can't get more data to peek at.
                   1526:                         */
                   1527:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1528:                                break;
                   1529:                        /*
                   1530:                         * If we've drained the socket buffer, tell the
                   1531:                         * protocol in case it needs to do something to
                   1532:                         * get it filled again.
                   1533:                         */
                   1534:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1.233   ! rtr      1535:                                (*pr->pr_usrreqs->pr_rcvd)(so, flags, l);
1.69      thorpej  1536:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1537:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.196     dsl      1538:                        if (wakeup_state & SS_RESTARTSYS)
                   1539:                                error = ERESTART;
                   1540:                        else
                   1541:                                error = sbwait(&so->so_rcv);
1.144     dyoung   1542:                        if (error != 0) {
1.1       cgd      1543:                                sbunlock(&so->so_rcv);
1.160     ad       1544:                                sounlock(so);
1.1       cgd      1545:                                splx(s);
1.144     dyoung   1546:                                return 0;
1.1       cgd      1547:                        }
1.21      christos 1548:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1549:                                nextrecord = m->m_nextpkt;
1.196     dsl      1550:                        wakeup_state = so->so_state;
1.1       cgd      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)
1.233   ! rtr      1576:                        (*pr->pr_usrreqs->pr_rcvd)(so, flags, l);
1.1       cgd      1577:        }
1.3       andrew   1578:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1579:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1580:                sbunlock(&so->so_rcv);
                   1581:                goto restart;
                   1582:        }
1.108     perry    1583:
1.144     dyoung   1584:        if (flagsp != NULL)
1.1       cgd      1585:                *flagsp |= flags;
1.54      lukem    1586:  release:
1.1       cgd      1587:        sbunlock(&so->so_rcv);
1.160     ad       1588:        sounlock(so);
1.1       cgd      1589:        splx(s);
1.144     dyoung   1590:        return error;
1.1       cgd      1591: }
                   1592:
1.14      mycroft  1593: int
1.54      lukem    1594: soshutdown(struct socket *so, int how)
1.1       cgd      1595: {
1.99      matt     1596:        const struct protosw    *pr;
1.160     ad       1597:        int     error;
                   1598:
                   1599:        KASSERT(solocked(so));
1.34      kleink   1600:
1.54      lukem    1601:        pr = so->so_proto;
1.34      kleink   1602:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1603:                return (EINVAL);
1.1       cgd      1604:
1.160     ad       1605:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1606:                sorflush(so);
1.160     ad       1607:                error = 0;
                   1608:        }
1.34      kleink   1609:        if (how == SHUT_WR || how == SHUT_RDWR)
1.229     rtr      1610:                error = (*pr->pr_usrreqs->pr_shutdown)(so);
1.160     ad       1611:
                   1612:        return error;
1.1       cgd      1613: }
                   1614:
1.195     dsl      1615: void
1.196     dsl      1616: sorestart(struct socket *so)
1.188     ad       1617: {
1.196     dsl      1618:        /*
                   1619:         * An application has called close() on an fd on which another
                   1620:         * of its threads has called a socket system call.
                   1621:         * Mark this and wake everyone up, and code that would block again
                   1622:         * instead returns ERESTART.
                   1623:         * On system call re-entry the fd is validated and EBADF returned.
                   1624:         * Any other fd will block again on the 2nd syscall.
                   1625:         */
1.188     ad       1626:        solock(so);
1.196     dsl      1627:        so->so_state |= SS_RESTARTSYS;
1.188     ad       1628:        cv_broadcast(&so->so_cv);
1.196     dsl      1629:        cv_broadcast(&so->so_snd.sb_cv);
                   1630:        cv_broadcast(&so->so_rcv.sb_cv);
1.188     ad       1631:        sounlock(so);
                   1632: }
                   1633:
1.14      mycroft  1634: void
1.54      lukem    1635: sorflush(struct socket *so)
1.1       cgd      1636: {
1.54      lukem    1637:        struct sockbuf  *sb, asb;
1.99      matt     1638:        const struct protosw    *pr;
1.160     ad       1639:
                   1640:        KASSERT(solocked(so));
1.1       cgd      1641:
1.54      lukem    1642:        sb = &so->so_rcv;
                   1643:        pr = so->so_proto;
1.160     ad       1644:        socantrcvmore(so);
1.1       cgd      1645:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1646:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1647:        sbunlock(sb);
                   1648:        asb = *sb;
1.86      wrstuden 1649:        /*
                   1650:         * Clear most of the sockbuf structure, but leave some of the
                   1651:         * fields valid.
                   1652:         */
                   1653:        memset(&sb->sb_startzero, 0,
                   1654:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1655:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1656:                sounlock(so);
1.1       cgd      1657:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1658:                solock(so);
                   1659:        }
1.98      christos 1660:        sbrelease(&asb, so);
1.1       cgd      1661: }
                   1662:
1.171     plunky   1663: /*
                   1664:  * internal set SOL_SOCKET options
                   1665:  */
1.142     dyoung   1666: static int
1.171     plunky   1667: sosetopt1(struct socket *so, const struct sockopt *sopt)
1.1       cgd      1668: {
1.219     christos 1669:        int error = EINVAL, opt;
                   1670:        int optval = 0; /* XXX: gcc */
1.171     plunky   1671:        struct linger l;
                   1672:        struct timeval tv;
1.142     dyoung   1673:
1.179     christos 1674:        switch ((opt = sopt->sopt_name)) {
1.142     dyoung   1675:
1.170     tls      1676:        case SO_ACCEPTFILTER:
1.177     ad       1677:                error = accept_filt_setopt(so, sopt);
                   1678:                KASSERT(solocked(so));
1.170     tls      1679:                break;
                   1680:
1.171     plunky   1681:        case SO_LINGER:
                   1682:                error = sockopt_get(sopt, &l, sizeof(l));
1.177     ad       1683:                solock(so);
1.171     plunky   1684:                if (error)
1.177     ad       1685:                        break;
1.171     plunky   1686:                if (l.l_linger < 0 || l.l_linger > USHRT_MAX ||
1.177     ad       1687:                    l.l_linger > (INT_MAX / hz)) {
                   1688:                        error = EDOM;
                   1689:                        break;
                   1690:                }
1.171     plunky   1691:                so->so_linger = l.l_linger;
                   1692:                if (l.l_onoff)
                   1693:                        so->so_options |= SO_LINGER;
                   1694:                else
                   1695:                        so->so_options &= ~SO_LINGER;
1.177     ad       1696:                break;
1.1       cgd      1697:
1.142     dyoung   1698:        case SO_DEBUG:
                   1699:        case SO_KEEPALIVE:
                   1700:        case SO_DONTROUTE:
                   1701:        case SO_USELOOPBACK:
                   1702:        case SO_BROADCAST:
                   1703:        case SO_REUSEADDR:
                   1704:        case SO_REUSEPORT:
                   1705:        case SO_OOBINLINE:
                   1706:        case SO_TIMESTAMP:
1.207     christos 1707:        case SO_NOSIGPIPE:
1.184     christos 1708: #ifdef SO_OTIMESTAMP
                   1709:        case SO_OTIMESTAMP:
                   1710: #endif
1.171     plunky   1711:                error = sockopt_getint(sopt, &optval);
1.177     ad       1712:                solock(so);
1.171     plunky   1713:                if (error)
1.177     ad       1714:                        break;
1.171     plunky   1715:                if (optval)
1.179     christos 1716:                        so->so_options |= opt;
1.142     dyoung   1717:                else
1.179     christos 1718:                        so->so_options &= ~opt;
1.142     dyoung   1719:                break;
                   1720:
                   1721:        case SO_SNDBUF:
                   1722:        case SO_RCVBUF:
                   1723:        case SO_SNDLOWAT:
                   1724:        case SO_RCVLOWAT:
1.171     plunky   1725:                error = sockopt_getint(sopt, &optval);
1.177     ad       1726:                solock(so);
1.171     plunky   1727:                if (error)
1.177     ad       1728:                        break;
1.1       cgd      1729:
1.142     dyoung   1730:                /*
                   1731:                 * Values < 1 make no sense for any of these
                   1732:                 * options, so disallow them.
                   1733:                 */
1.177     ad       1734:                if (optval < 1) {
                   1735:                        error = EINVAL;
                   1736:                        break;
                   1737:                }
1.1       cgd      1738:
1.179     christos 1739:                switch (opt) {
1.171     plunky   1740:                case SO_SNDBUF:
1.177     ad       1741:                        if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) {
                   1742:                                error = ENOBUFS;
                   1743:                                break;
                   1744:                        }
1.171     plunky   1745:                        so->so_snd.sb_flags &= ~SB_AUTOSIZE;
                   1746:                        break;
1.1       cgd      1747:
                   1748:                case SO_RCVBUF:
1.177     ad       1749:                        if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) {
                   1750:                                error = ENOBUFS;
                   1751:                                break;
                   1752:                        }
1.171     plunky   1753:                        so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1.142     dyoung   1754:                        break;
                   1755:
                   1756:                /*
                   1757:                 * Make sure the low-water is never greater than
                   1758:                 * the high-water.
                   1759:                 */
1.1       cgd      1760:                case SO_SNDLOWAT:
1.171     plunky   1761:                        if (optval > so->so_snd.sb_hiwat)
                   1762:                                optval = so->so_snd.sb_hiwat;
                   1763:
                   1764:                        so->so_snd.sb_lowat = optval;
1.142     dyoung   1765:                        break;
1.171     plunky   1766:
1.1       cgd      1767:                case SO_RCVLOWAT:
1.171     plunky   1768:                        if (optval > so->so_rcv.sb_hiwat)
                   1769:                                optval = so->so_rcv.sb_hiwat;
                   1770:
                   1771:                        so->so_rcv.sb_lowat = optval;
1.142     dyoung   1772:                        break;
                   1773:                }
                   1774:                break;
1.28      thorpej  1775:
1.179     christos 1776: #ifdef COMPAT_50
                   1777:        case SO_OSNDTIMEO:
                   1778:        case SO_ORCVTIMEO: {
                   1779:                struct timeval50 otv;
                   1780:                error = sockopt_get(sopt, &otv, sizeof(otv));
1.186     pooka    1781:                if (error) {
                   1782:                        solock(so);
1.183     christos 1783:                        break;
1.186     pooka    1784:                }
1.179     christos 1785:                timeval50_to_timeval(&otv, &tv);
                   1786:                opt = opt == SO_OSNDTIMEO ? SO_SNDTIMEO : SO_RCVTIMEO;
1.182     christos 1787:                error = 0;
1.179     christos 1788:                /*FALLTHROUGH*/
                   1789:        }
                   1790: #endif /* COMPAT_50 */
                   1791:
1.142     dyoung   1792:        case SO_SNDTIMEO:
                   1793:        case SO_RCVTIMEO:
1.182     christos 1794:                if (error)
1.179     christos 1795:                        error = sockopt_get(sopt, &tv, sizeof(tv));
1.177     ad       1796:                solock(so);
1.171     plunky   1797:                if (error)
1.177     ad       1798:                        break;
1.171     plunky   1799:
1.177     ad       1800:                if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) {
                   1801:                        error = EDOM;
                   1802:                        break;
                   1803:                }
1.28      thorpej  1804:
1.171     plunky   1805:                optval = tv.tv_sec * hz + tv.tv_usec / tick;
                   1806:                if (optval == 0 && tv.tv_usec != 0)
                   1807:                        optval = 1;
1.28      thorpej  1808:
1.179     christos 1809:                switch (opt) {
1.142     dyoung   1810:                case SO_SNDTIMEO:
1.171     plunky   1811:                        so->so_snd.sb_timeo = optval;
1.1       cgd      1812:                        break;
                   1813:                case SO_RCVTIMEO:
1.171     plunky   1814:                        so->so_rcv.sb_timeo = optval;
1.142     dyoung   1815:                        break;
                   1816:                }
                   1817:                break;
1.1       cgd      1818:
1.142     dyoung   1819:        default:
1.177     ad       1820:                solock(so);
                   1821:                error = ENOPROTOOPT;
                   1822:                break;
1.142     dyoung   1823:        }
1.177     ad       1824:        KASSERT(solocked(so));
                   1825:        return error;
1.142     dyoung   1826: }
1.1       cgd      1827:
1.142     dyoung   1828: int
1.171     plunky   1829: sosetopt(struct socket *so, struct sockopt *sopt)
1.142     dyoung   1830: {
                   1831:        int error, prerr;
1.1       cgd      1832:
1.177     ad       1833:        if (sopt->sopt_level == SOL_SOCKET) {
1.171     plunky   1834:                error = sosetopt1(so, sopt);
1.177     ad       1835:                KASSERT(solocked(so));
                   1836:        } else {
1.142     dyoung   1837:                error = ENOPROTOOPT;
1.177     ad       1838:                solock(so);
                   1839:        }
1.1       cgd      1840:
1.142     dyoung   1841:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1842:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1843:                /* give the protocol stack a shot */
1.171     plunky   1844:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt);
1.142     dyoung   1845:                if (prerr == 0)
                   1846:                        error = 0;
                   1847:                else if (prerr != ENOPROTOOPT)
                   1848:                        error = prerr;
1.171     plunky   1849:        }
1.160     ad       1850:        sounlock(so);
1.142     dyoung   1851:        return error;
1.1       cgd      1852: }
                   1853:
1.171     plunky   1854: /*
                   1855:  * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt()
                   1856:  */
                   1857: int
                   1858: so_setsockopt(struct lwp *l, struct socket *so, int level, int name,
                   1859:     const void *val, size_t valsize)
                   1860: {
                   1861:        struct sockopt sopt;
                   1862:        int error;
                   1863:
                   1864:        KASSERT(valsize == 0 || val != NULL);
                   1865:
                   1866:        sockopt_init(&sopt, level, name, valsize);
                   1867:        sockopt_set(&sopt, val, valsize);
                   1868:
                   1869:        error = sosetopt(so, &sopt);
                   1870:
                   1871:        sockopt_destroy(&sopt);
                   1872:
                   1873:        return error;
                   1874: }
                   1875:
                   1876: /*
                   1877:  * internal get SOL_SOCKET options
                   1878:  */
                   1879: static int
                   1880: sogetopt1(struct socket *so, struct sockopt *sopt)
                   1881: {
1.179     christos 1882:        int error, optval, opt;
1.171     plunky   1883:        struct linger l;
                   1884:        struct timeval tv;
                   1885:
1.179     christos 1886:        switch ((opt = sopt->sopt_name)) {
1.171     plunky   1887:
                   1888:        case SO_ACCEPTFILTER:
1.177     ad       1889:                error = accept_filt_getopt(so, sopt);
1.171     plunky   1890:                break;
                   1891:
                   1892:        case SO_LINGER:
                   1893:                l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0;
                   1894:                l.l_linger = so->so_linger;
                   1895:
                   1896:                error = sockopt_set(sopt, &l, sizeof(l));
                   1897:                break;
                   1898:
                   1899:        case SO_USELOOPBACK:
                   1900:        case SO_DONTROUTE:
                   1901:        case SO_DEBUG:
                   1902:        case SO_KEEPALIVE:
                   1903:        case SO_REUSEADDR:
                   1904:        case SO_REUSEPORT:
                   1905:        case SO_BROADCAST:
                   1906:        case SO_OOBINLINE:
                   1907:        case SO_TIMESTAMP:
1.207     christos 1908:        case SO_NOSIGPIPE:
1.184     christos 1909: #ifdef SO_OTIMESTAMP
                   1910:        case SO_OTIMESTAMP:
                   1911: #endif
1.218     seanb    1912:        case SO_ACCEPTCONN:
1.179     christos 1913:                error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0);
1.171     plunky   1914:                break;
                   1915:
                   1916:        case SO_TYPE:
                   1917:                error = sockopt_setint(sopt, so->so_type);
                   1918:                break;
                   1919:
                   1920:        case SO_ERROR:
                   1921:                error = sockopt_setint(sopt, so->so_error);
                   1922:                so->so_error = 0;
                   1923:                break;
                   1924:
                   1925:        case SO_SNDBUF:
                   1926:                error = sockopt_setint(sopt, so->so_snd.sb_hiwat);
                   1927:                break;
                   1928:
                   1929:        case SO_RCVBUF:
                   1930:                error = sockopt_setint(sopt, so->so_rcv.sb_hiwat);
                   1931:                break;
                   1932:
                   1933:        case SO_SNDLOWAT:
                   1934:                error = sockopt_setint(sopt, so->so_snd.sb_lowat);
                   1935:                break;
                   1936:
                   1937:        case SO_RCVLOWAT:
                   1938:                error = sockopt_setint(sopt, so->so_rcv.sb_lowat);
                   1939:                break;
                   1940:
1.179     christos 1941: #ifdef COMPAT_50
                   1942:        case SO_OSNDTIMEO:
                   1943:        case SO_ORCVTIMEO: {
                   1944:                struct timeval50 otv;
                   1945:
                   1946:                optval = (opt == SO_OSNDTIMEO ?
                   1947:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1948:
                   1949:                otv.tv_sec = optval / hz;
                   1950:                otv.tv_usec = (optval % hz) * tick;
                   1951:
                   1952:                error = sockopt_set(sopt, &otv, sizeof(otv));
                   1953:                break;
                   1954:        }
                   1955: #endif /* COMPAT_50 */
                   1956:
1.171     plunky   1957:        case SO_SNDTIMEO:
                   1958:        case SO_RCVTIMEO:
1.179     christos 1959:                optval = (opt == SO_SNDTIMEO ?
1.171     plunky   1960:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1961:
                   1962:                tv.tv_sec = optval / hz;
                   1963:                tv.tv_usec = (optval % hz) * tick;
                   1964:
                   1965:                error = sockopt_set(sopt, &tv, sizeof(tv));
                   1966:                break;
                   1967:
                   1968:        case SO_OVERFLOWED:
                   1969:                error = sockopt_setint(sopt, so->so_rcv.sb_overflowed);
                   1970:                break;
                   1971:
                   1972:        default:
                   1973:                error = ENOPROTOOPT;
                   1974:                break;
                   1975:        }
                   1976:
                   1977:        return (error);
                   1978: }
                   1979:
1.14      mycroft  1980: int
1.171     plunky   1981: sogetopt(struct socket *so, struct sockopt *sopt)
1.1       cgd      1982: {
1.160     ad       1983:        int             error;
1.1       cgd      1984:
1.160     ad       1985:        solock(so);
1.171     plunky   1986:        if (sopt->sopt_level != SOL_SOCKET) {
1.1       cgd      1987:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       1988:                        error = ((*so->so_proto->pr_ctloutput)
1.171     plunky   1989:                            (PRCO_GETOPT, so, sopt));
1.1       cgd      1990:                } else
1.160     ad       1991:                        error = (ENOPROTOOPT);
1.1       cgd      1992:        } else {
1.171     plunky   1993:                error = sogetopt1(so, sopt);
                   1994:        }
                   1995:        sounlock(so);
                   1996:        return (error);
                   1997: }
                   1998:
                   1999: /*
                   2000:  * alloc sockopt data buffer buffer
                   2001:  *     - will be released at destroy
                   2002:  */
1.176     plunky   2003: static int
                   2004: sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag)
1.171     plunky   2005: {
                   2006:
                   2007:        KASSERT(sopt->sopt_size == 0);
                   2008:
1.176     plunky   2009:        if (len > sizeof(sopt->sopt_buf)) {
                   2010:                sopt->sopt_data = kmem_zalloc(len, kmflag);
                   2011:                if (sopt->sopt_data == NULL)
                   2012:                        return ENOMEM;
                   2013:        } else
1.171     plunky   2014:                sopt->sopt_data = sopt->sopt_buf;
                   2015:
                   2016:        sopt->sopt_size = len;
1.176     plunky   2017:        return 0;
1.171     plunky   2018: }
                   2019:
                   2020: /*
                   2021:  * initialise sockopt storage
1.176     plunky   2022:  *     - MAY sleep during allocation
1.171     plunky   2023:  */
                   2024: void
                   2025: sockopt_init(struct sockopt *sopt, int level, int name, size_t size)
                   2026: {
1.1       cgd      2027:
1.171     plunky   2028:        memset(sopt, 0, sizeof(*sopt));
1.1       cgd      2029:
1.171     plunky   2030:        sopt->sopt_level = level;
                   2031:        sopt->sopt_name = name;
1.176     plunky   2032:        (void)sockopt_alloc(sopt, size, KM_SLEEP);
1.171     plunky   2033: }
                   2034:
                   2035: /*
                   2036:  * destroy sockopt storage
                   2037:  *     - will release any held memory references
                   2038:  */
                   2039: void
                   2040: sockopt_destroy(struct sockopt *sopt)
                   2041: {
                   2042:
                   2043:        if (sopt->sopt_data != sopt->sopt_buf)
1.173     plunky   2044:                kmem_free(sopt->sopt_data, sopt->sopt_size);
1.171     plunky   2045:
                   2046:        memset(sopt, 0, sizeof(*sopt));
                   2047: }
                   2048:
                   2049: /*
                   2050:  * set sockopt value
                   2051:  *     - value is copied into sockopt
1.176     plunky   2052:  *     - memory is allocated when necessary, will not sleep
1.171     plunky   2053:  */
                   2054: int
                   2055: sockopt_set(struct sockopt *sopt, const void *buf, size_t len)
                   2056: {
1.176     plunky   2057:        int error;
1.171     plunky   2058:
1.176     plunky   2059:        if (sopt->sopt_size == 0) {
                   2060:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2061:                if (error)
                   2062:                        return error;
                   2063:        }
1.171     plunky   2064:
                   2065:        KASSERT(sopt->sopt_size == len);
                   2066:        memcpy(sopt->sopt_data, buf, len);
                   2067:        return 0;
                   2068: }
                   2069:
                   2070: /*
                   2071:  * common case of set sockopt integer value
                   2072:  */
                   2073: int
                   2074: sockopt_setint(struct sockopt *sopt, int val)
                   2075: {
                   2076:
                   2077:        return sockopt_set(sopt, &val, sizeof(int));
                   2078: }
                   2079:
                   2080: /*
                   2081:  * get sockopt value
                   2082:  *     - correct size must be given
                   2083:  */
                   2084: int
                   2085: sockopt_get(const struct sockopt *sopt, void *buf, size_t len)
                   2086: {
1.170     tls      2087:
1.171     plunky   2088:        if (sopt->sopt_size != len)
                   2089:                return EINVAL;
1.1       cgd      2090:
1.171     plunky   2091:        memcpy(buf, sopt->sopt_data, len);
                   2092:        return 0;
                   2093: }
1.1       cgd      2094:
1.171     plunky   2095: /*
                   2096:  * common case of get sockopt integer value
                   2097:  */
                   2098: int
                   2099: sockopt_getint(const struct sockopt *sopt, int *valp)
                   2100: {
1.1       cgd      2101:
1.171     plunky   2102:        return sockopt_get(sopt, valp, sizeof(int));
                   2103: }
1.1       cgd      2104:
1.171     plunky   2105: /*
                   2106:  * set sockopt value from mbuf
                   2107:  *     - ONLY for legacy code
                   2108:  *     - mbuf is released by sockopt
1.176     plunky   2109:  *     - will not sleep
1.171     plunky   2110:  */
                   2111: int
                   2112: sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m)
                   2113: {
                   2114:        size_t len;
1.176     plunky   2115:        int error;
1.1       cgd      2116:
1.171     plunky   2117:        len = m_length(m);
1.1       cgd      2118:
1.176     plunky   2119:        if (sopt->sopt_size == 0) {
                   2120:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2121:                if (error)
                   2122:                        return error;
                   2123:        }
1.1       cgd      2124:
1.171     plunky   2125:        KASSERT(sopt->sopt_size == len);
                   2126:        m_copydata(m, 0, len, sopt->sopt_data);
                   2127:        m_freem(m);
1.1       cgd      2128:
1.171     plunky   2129:        return 0;
                   2130: }
1.1       cgd      2131:
1.171     plunky   2132: /*
                   2133:  * get sockopt value into mbuf
                   2134:  *     - ONLY for legacy code
                   2135:  *     - mbuf to be released by the caller
1.176     plunky   2136:  *     - will not sleep
1.171     plunky   2137:  */
                   2138: struct mbuf *
                   2139: sockopt_getmbuf(const struct sockopt *sopt)
                   2140: {
                   2141:        struct mbuf *m;
1.107     darrenr  2142:
1.176     plunky   2143:        if (sopt->sopt_size > MCLBYTES)
                   2144:                return NULL;
                   2145:
                   2146:        m = m_get(M_DONTWAIT, MT_SOOPTS);
1.171     plunky   2147:        if (m == NULL)
                   2148:                return NULL;
                   2149:
1.176     plunky   2150:        if (sopt->sopt_size > MLEN) {
                   2151:                MCLGET(m, M_DONTWAIT);
                   2152:                if ((m->m_flags & M_EXT) == 0) {
                   2153:                        m_free(m);
                   2154:                        return NULL;
                   2155:                }
1.1       cgd      2156:        }
1.176     plunky   2157:
                   2158:        memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size);
                   2159:        m->m_len = sopt->sopt_size;
1.160     ad       2160:
1.171     plunky   2161:        return m;
1.1       cgd      2162: }
                   2163:
1.14      mycroft  2164: void
1.54      lukem    2165: sohasoutofband(struct socket *so)
1.1       cgd      2166: {
1.153     rmind    2167:
1.90      christos 2168:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.189     ad       2169:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT);
1.1       cgd      2170: }
1.72      jdolecek 2171:
                   2172: static void
                   2173: filt_sordetach(struct knote *kn)
                   2174: {
                   2175:        struct socket   *so;
                   2176:
1.155     ad       2177:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2178:        solock(so);
1.73      christos 2179:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2180:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 2181:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       2182:        sounlock(so);
1.72      jdolecek 2183: }
                   2184:
                   2185: /*ARGSUSED*/
                   2186: static int
1.129     yamt     2187: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 2188: {
                   2189:        struct socket   *so;
1.160     ad       2190:        int rv;
1.72      jdolecek 2191:
1.155     ad       2192:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2193:        if (hint != NOTE_SUBMIT)
                   2194:                solock(so);
1.72      jdolecek 2195:        kn->kn_data = so->so_rcv.sb_cc;
                   2196:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    2197:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2198:                kn->kn_fflags = so->so_error;
1.160     ad       2199:                rv = 1;
                   2200:        } else if (so->so_error)        /* temporary udp error */
                   2201:                rv = 1;
                   2202:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2203:                rv = (kn->kn_data >= kn->kn_sdata);
                   2204:        else
                   2205:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   2206:        if (hint != NOTE_SUBMIT)
                   2207:                sounlock(so);
                   2208:        return rv;
1.72      jdolecek 2209: }
                   2210:
                   2211: static void
                   2212: filt_sowdetach(struct knote *kn)
                   2213: {
                   2214:        struct socket   *so;
                   2215:
1.155     ad       2216:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2217:        solock(so);
1.73      christos 2218:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2219:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 2220:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       2221:        sounlock(so);
1.72      jdolecek 2222: }
                   2223:
                   2224: /*ARGSUSED*/
                   2225: static int
1.129     yamt     2226: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 2227: {
                   2228:        struct socket   *so;
1.160     ad       2229:        int rv;
1.72      jdolecek 2230:
1.155     ad       2231:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2232:        if (hint != NOTE_SUBMIT)
                   2233:                solock(so);
1.72      jdolecek 2234:        kn->kn_data = sbspace(&so->so_snd);
                   2235:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    2236:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2237:                kn->kn_fflags = so->so_error;
1.160     ad       2238:                rv = 1;
                   2239:        } else if (so->so_error)        /* temporary udp error */
                   2240:                rv = 1;
                   2241:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 2242:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       2243:                rv = 0;
                   2244:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2245:                rv = (kn->kn_data >= kn->kn_sdata);
                   2246:        else
                   2247:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   2248:        if (hint != NOTE_SUBMIT)
                   2249:                sounlock(so);
                   2250:        return rv;
1.72      jdolecek 2251: }
                   2252:
                   2253: /*ARGSUSED*/
                   2254: static int
1.129     yamt     2255: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 2256: {
                   2257:        struct socket   *so;
1.160     ad       2258:        int rv;
1.72      jdolecek 2259:
1.155     ad       2260:        so = ((file_t *)kn->kn_obj)->f_data;
1.72      jdolecek 2261:
                   2262:        /*
                   2263:         * Set kn_data to number of incoming connections, not
                   2264:         * counting partial (incomplete) connections.
1.108     perry    2265:         */
1.160     ad       2266:        if (hint != NOTE_SUBMIT)
                   2267:                solock(so);
1.72      jdolecek 2268:        kn->kn_data = so->so_qlen;
1.160     ad       2269:        rv = (kn->kn_data > 0);
                   2270:        if (hint != NOTE_SUBMIT)
                   2271:                sounlock(so);
                   2272:        return rv;
1.72      jdolecek 2273: }
                   2274:
                   2275: static const struct filterops solisten_filtops =
                   2276:        { 1, NULL, filt_sordetach, filt_solisten };
                   2277: static const struct filterops soread_filtops =
                   2278:        { 1, NULL, filt_sordetach, filt_soread };
                   2279: static const struct filterops sowrite_filtops =
                   2280:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   2281:
                   2282: int
1.129     yamt     2283: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 2284: {
                   2285:        struct socket   *so;
                   2286:        struct sockbuf  *sb;
                   2287:
1.155     ad       2288:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2289:        solock(so);
1.72      jdolecek 2290:        switch (kn->kn_filter) {
                   2291:        case EVFILT_READ:
                   2292:                if (so->so_options & SO_ACCEPTCONN)
                   2293:                        kn->kn_fop = &solisten_filtops;
                   2294:                else
                   2295:                        kn->kn_fop = &soread_filtops;
                   2296:                sb = &so->so_rcv;
                   2297:                break;
                   2298:        case EVFILT_WRITE:
                   2299:                kn->kn_fop = &sowrite_filtops;
                   2300:                sb = &so->so_snd;
                   2301:                break;
                   2302:        default:
1.160     ad       2303:                sounlock(so);
1.149     pooka    2304:                return (EINVAL);
1.72      jdolecek 2305:        }
1.73      christos 2306:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 2307:        sb->sb_flags |= SB_KNOTE;
1.160     ad       2308:        sounlock(so);
1.72      jdolecek 2309:        return (0);
                   2310: }
                   2311:
1.154     ad       2312: static int
                   2313: sodopoll(struct socket *so, int events)
                   2314: {
                   2315:        int revents;
                   2316:
                   2317:        revents = 0;
                   2318:
                   2319:        if (events & (POLLIN | POLLRDNORM))
                   2320:                if (soreadable(so))
                   2321:                        revents |= events & (POLLIN | POLLRDNORM);
                   2322:
                   2323:        if (events & (POLLOUT | POLLWRNORM))
                   2324:                if (sowritable(so))
                   2325:                        revents |= events & (POLLOUT | POLLWRNORM);
                   2326:
                   2327:        if (events & (POLLPRI | POLLRDBAND))
                   2328:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   2329:                        revents |= events & (POLLPRI | POLLRDBAND);
                   2330:
                   2331:        return revents;
                   2332: }
                   2333:
                   2334: int
                   2335: sopoll(struct socket *so, int events)
                   2336: {
                   2337:        int revents = 0;
                   2338:
1.160     ad       2339: #ifndef DIAGNOSTIC
                   2340:        /*
                   2341:         * Do a quick, unlocked check in expectation that the socket
                   2342:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   2343:         * as the solocked() assertions will fail.
                   2344:         */
1.154     ad       2345:        if ((revents = sodopoll(so, events)) != 0)
                   2346:                return revents;
1.160     ad       2347: #endif
1.154     ad       2348:
1.160     ad       2349:        solock(so);
1.154     ad       2350:        if ((revents = sodopoll(so, events)) == 0) {
                   2351:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   2352:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       2353:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       2354:                }
                   2355:
                   2356:                if (events & (POLLOUT | POLLWRNORM)) {
                   2357:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2358:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2359:                }
                   2360:        }
1.160     ad       2361:        sounlock(so);
1.154     ad       2362:
                   2363:        return revents;
                   2364: }
                   2365:
                   2366:
1.94      yamt     2367: #include <sys/sysctl.h>
                   2368:
                   2369: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
1.212     pooka    2370: static int sysctl_kern_sbmax(SYSCTLFN_PROTO);
1.94      yamt     2371:
                   2372: /*
                   2373:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2374:  * value is not too small.
                   2375:  * (XXX should we maybe make sure it's not too large as well?)
                   2376:  */
                   2377: static int
                   2378: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2379: {
                   2380:        int error, new_somaxkva;
                   2381:        struct sysctlnode node;
                   2382:
                   2383:        new_somaxkva = somaxkva;
                   2384:        node = *rnode;
                   2385:        node.sysctl_data = &new_somaxkva;
                   2386:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2387:        if (error || newp == NULL)
                   2388:                return (error);
                   2389:
                   2390:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2391:                return (EINVAL);
                   2392:
1.136     ad       2393:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2394:        somaxkva = new_somaxkva;
1.136     ad       2395:        cv_broadcast(&socurkva_cv);
                   2396:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2397:
                   2398:        return (error);
                   2399: }
                   2400:
1.212     pooka    2401: /*
                   2402:  * sysctl helper routine for kern.sbmax. Basically just ensures that
                   2403:  * any new value is not too small.
                   2404:  */
                   2405: static int
                   2406: sysctl_kern_sbmax(SYSCTLFN_ARGS)
                   2407: {
                   2408:        int error, new_sbmax;
                   2409:        struct sysctlnode node;
                   2410:
                   2411:        new_sbmax = sb_max;
                   2412:        node = *rnode;
                   2413:        node.sysctl_data = &new_sbmax;
                   2414:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2415:        if (error || newp == NULL)
                   2416:                return (error);
                   2417:
                   2418:        KERNEL_LOCK(1, NULL);
                   2419:        error = sb_max_set(new_sbmax);
                   2420:        KERNEL_UNLOCK_ONE(NULL);
                   2421:
                   2422:        return (error);
                   2423: }
                   2424:
1.178     pooka    2425: static void
1.212     pooka    2426: sysctl_kern_socket_setup(void)
1.94      yamt     2427: {
                   2428:
1.178     pooka    2429:        KASSERT(socket_sysctllog == NULL);
1.97      atatat   2430:
1.178     pooka    2431:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2432:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2433:                       CTLTYPE_INT, "somaxkva",
                   2434:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2435:                                    "used for socket buffers"),
1.94      yamt     2436:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2437:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
1.212     pooka    2438:
                   2439:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
                   2440:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
                   2441:                       CTLTYPE_INT, "sbmax",
                   2442:                       SYSCTL_DESCR("Maximum socket buffer size"),
                   2443:                       sysctl_kern_sbmax, 0, NULL, 0,
                   2444:                       CTL_KERN, KERN_SBMAX, CTL_EOL);
1.94      yamt     2445: }

CVSweb <webmaster@jp.NetBSD.org>