[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 1.174

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

CVSweb <webmaster@jp.NetBSD.org>