[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 1.192

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

CVSweb <webmaster@jp.NetBSD.org>