[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 1.172

1.172   ! ad          1: /*     $NetBSD: uipc_socket.c,v 1.171 2008/08/06 15:01:23 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>
1.172   ! ad         66: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.171 2008/08/06 15:01:23 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>
1.9       mycroft    80: #include <sys/malloc.h>
                     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>
1.72      jdolecek   89: #include <sys/event.h>
1.89      christos   90: #include <sys/poll.h>
1.118     elad       91: #include <sys/kauth.h>
1.136     ad         92: #include <sys/mutex.h>
                     93: #include <sys/condvar.h>
1.37      thorpej    94:
1.64      thorpej    95: #include <uvm/uvm.h>
                     96:
1.77      thorpej    97: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
                     98: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37      thorpej    99:
1.142     dyoung    100: extern const struct fileops socketops;
                    101:
1.54      lukem     102: extern int     somaxconn;                      /* patchable (XXX sysctl) */
                    103: int            somaxconn = SOMAXCONN;
1.160     ad        104: kmutex_t       *softnet_lock;
1.49      jonathan  105:
1.64      thorpej   106: #ifdef SOSEND_COUNTERS
                    107: #include <sys/device.h>
                    108:
1.113     thorpej   109: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   110:     NULL, "sosend", "loan big");
1.113     thorpej   111: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   112:     NULL, "sosend", "copy big");
1.113     thorpej   113: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   114:     NULL, "sosend", "copy small");
1.113     thorpej   115: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   116:     NULL, "sosend", "kva limit");
                    117:
                    118: #define        SOSEND_COUNTER_INCR(ev)         (ev)->ev_count++
                    119:
1.101     matt      120: EVCNT_ATTACH_STATIC(sosend_loan_big);
                    121: EVCNT_ATTACH_STATIC(sosend_copy_big);
                    122: EVCNT_ATTACH_STATIC(sosend_copy_small);
                    123: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64      thorpej   124: #else
                    125:
                    126: #define        SOSEND_COUNTER_INCR(ev)         /* nothing */
                    127:
                    128: #endif /* SOSEND_COUNTERS */
                    129:
1.119     yamt      130: static struct callback_entry sokva_reclaimerentry;
1.1       cgd       131:
1.167     ad        132: #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR)
1.121     yamt      133: int sock_loan_thresh = -1;
1.71      thorpej   134: #else
1.121     yamt      135: int sock_loan_thresh = 4096;
1.65      thorpej   136: #endif
1.64      thorpej   137:
1.136     ad        138: static kmutex_t so_pendfree_lock;
1.113     thorpej   139: static struct mbuf *so_pendfree;
1.64      thorpej   140:
1.84      ragge     141: #ifndef SOMAXKVA
                    142: #define        SOMAXKVA (16 * 1024 * 1024)
                    143: #endif
                    144: int somaxkva = SOMAXKVA;
1.113     thorpej   145: static int socurkva;
1.136     ad        146: static kcondvar_t socurkva_cv;
1.64      thorpej   147:
                    148: #define        SOCK_LOAN_CHUNK         65536
                    149:
1.117     yamt      150: static size_t sodopendfree(void);
                    151: static size_t sodopendfreel(void);
1.93      yamt      152:
1.113     thorpej   153: static vsize_t
1.129     yamt      154: sokvareserve(struct socket *so, vsize_t len)
1.80      yamt      155: {
1.98      christos  156:        int error;
1.80      yamt      157:
1.136     ad        158:        mutex_enter(&so_pendfree_lock);
1.80      yamt      159:        while (socurkva + len > somaxkva) {
1.93      yamt      160:                size_t freed;
                    161:
                    162:                /*
                    163:                 * try to do pendfree.
                    164:                 */
                    165:
1.117     yamt      166:                freed = sodopendfreel();
1.93      yamt      167:
                    168:                /*
                    169:                 * if some kva was freed, try again.
                    170:                 */
                    171:
                    172:                if (freed)
1.80      yamt      173:                        continue;
1.93      yamt      174:
1.80      yamt      175:                SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136     ad        176:                error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98      christos  177:                if (error) {
                    178:                        len = 0;
                    179:                        break;
                    180:                }
1.80      yamt      181:        }
1.93      yamt      182:        socurkva += len;
1.136     ad        183:        mutex_exit(&so_pendfree_lock);
1.98      christos  184:        return len;
1.95      yamt      185: }
                    186:
1.113     thorpej   187: static void
1.95      yamt      188: sokvaunreserve(vsize_t len)
                    189: {
                    190:
1.136     ad        191:        mutex_enter(&so_pendfree_lock);
1.95      yamt      192:        socurkva -= len;
1.136     ad        193:        cv_broadcast(&socurkva_cv);
                    194:        mutex_exit(&so_pendfree_lock);
1.95      yamt      195: }
                    196:
                    197: /*
                    198:  * sokvaalloc: allocate kva for loan.
                    199:  */
                    200:
                    201: vaddr_t
                    202: sokvaalloc(vsize_t len, struct socket *so)
                    203: {
                    204:        vaddr_t lva;
                    205:
                    206:        /*
                    207:         * reserve kva.
                    208:         */
                    209:
1.98      christos  210:        if (sokvareserve(so, len) == 0)
                    211:                return 0;
1.93      yamt      212:
                    213:        /*
                    214:         * allocate kva.
                    215:         */
1.80      yamt      216:
1.109     yamt      217:        lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95      yamt      218:        if (lva == 0) {
                    219:                sokvaunreserve(len);
1.80      yamt      220:                return (0);
1.95      yamt      221:        }
1.80      yamt      222:
                    223:        return lva;
                    224: }
                    225:
1.93      yamt      226: /*
                    227:  * sokvafree: free kva for loan.
                    228:  */
                    229:
1.80      yamt      230: void
                    231: sokvafree(vaddr_t sva, vsize_t len)
                    232: {
1.93      yamt      233:
                    234:        /*
                    235:         * free kva.
                    236:         */
1.80      yamt      237:
1.109     yamt      238:        uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93      yamt      239:
                    240:        /*
                    241:         * unreserve kva.
                    242:         */
                    243:
1.95      yamt      244:        sokvaunreserve(len);
1.80      yamt      245: }
                    246:
1.64      thorpej   247: static void
1.134     christos  248: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64      thorpej   249: {
1.156     yamt      250:        vaddr_t sva, eva;
1.64      thorpej   251:        vsize_t len;
1.156     yamt      252:        int npgs;
                    253:
                    254:        KASSERT(pgs != NULL);
1.64      thorpej   255:
                    256:        eva = round_page((vaddr_t) buf + size);
                    257:        sva = trunc_page((vaddr_t) buf);
                    258:        len = eva - sva;
                    259:        npgs = len >> PAGE_SHIFT;
                    260:
                    261:        pmap_kremove(sva, len);
                    262:        pmap_update(pmap_kernel());
                    263:        uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80      yamt      264:        sokvafree(sva, len);
1.64      thorpej   265: }
                    266:
                    267: static size_t
1.152     matt      268: sodopendfree(void)
1.64      thorpej   269: {
1.93      yamt      270:        size_t rv;
1.64      thorpej   271:
1.160     ad        272:        if (__predict_true(so_pendfree == NULL))
                    273:                return 0;
                    274:
1.136     ad        275:        mutex_enter(&so_pendfree_lock);
1.117     yamt      276:        rv = sodopendfreel();
1.136     ad        277:        mutex_exit(&so_pendfree_lock);
1.93      yamt      278:
                    279:        return rv;
                    280: }
                    281:
                    282: /*
                    283:  * sodopendfreel: free mbufs on "pendfree" list.
1.136     ad        284:  * unlock and relock so_pendfree_lock when freeing mbufs.
1.93      yamt      285:  *
1.136     ad        286:  * => called with so_pendfree_lock held.
1.93      yamt      287:  */
                    288:
                    289: static size_t
1.152     matt      290: sodopendfreel(void)
1.93      yamt      291: {
1.137     ad        292:        struct mbuf *m, *next;
1.93      yamt      293:        size_t rv = 0;
                    294:
1.136     ad        295:        KASSERT(mutex_owned(&so_pendfree_lock));
1.64      thorpej   296:
1.137     ad        297:        while (so_pendfree != NULL) {
1.64      thorpej   298:                m = so_pendfree;
1.93      yamt      299:                so_pendfree = NULL;
1.136     ad        300:                mutex_exit(&so_pendfree_lock);
1.93      yamt      301:
                    302:                for (; m != NULL; m = next) {
                    303:                        next = m->m_next;
1.156     yamt      304:                        KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
                    305:                        KASSERT(m->m_ext.ext_refcnt == 0);
1.93      yamt      306:
                    307:                        rv += m->m_ext.ext_size;
1.156     yamt      308:                        sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
1.93      yamt      309:                            m->m_ext.ext_size);
1.145     ad        310:                        pool_cache_put(mb_cache, m);
1.93      yamt      311:                }
1.64      thorpej   312:
1.136     ad        313:                mutex_enter(&so_pendfree_lock);
1.64      thorpej   314:        }
                    315:
                    316:        return (rv);
                    317: }
                    318:
1.80      yamt      319: void
1.134     christos  320: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64      thorpej   321: {
                    322:
1.156     yamt      323:        KASSERT(m != NULL);
1.64      thorpej   324:
1.93      yamt      325:        /*
                    326:         * postpone freeing mbuf.
                    327:         *
                    328:         * we can't do it in interrupt context
                    329:         * because we need to put kva back to kernel_map.
                    330:         */
                    331:
1.136     ad        332:        mutex_enter(&so_pendfree_lock);
1.92      yamt      333:        m->m_next = so_pendfree;
                    334:        so_pendfree = m;
1.136     ad        335:        cv_broadcast(&socurkva_cv);
                    336:        mutex_exit(&so_pendfree_lock);
1.64      thorpej   337: }
                    338:
                    339: static long
                    340: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
                    341: {
                    342:        struct iovec *iov = uio->uio_iov;
                    343:        vaddr_t sva, eva;
                    344:        vsize_t len;
1.156     yamt      345:        vaddr_t lva;
                    346:        int npgs, error;
                    347:        vaddr_t va;
                    348:        int i;
1.64      thorpej   349:
1.116     yamt      350:        if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64      thorpej   351:                return (0);
                    352:
                    353:        if (iov->iov_len < (size_t) space)
                    354:                space = iov->iov_len;
                    355:        if (space > SOCK_LOAN_CHUNK)
                    356:                space = SOCK_LOAN_CHUNK;
                    357:
                    358:        eva = round_page((vaddr_t) iov->iov_base + space);
                    359:        sva = trunc_page((vaddr_t) iov->iov_base);
                    360:        len = eva - sva;
                    361:        npgs = len >> PAGE_SHIFT;
                    362:
1.79      thorpej   363:        KASSERT(npgs <= M_EXT_MAXPAGES);
                    364:
1.80      yamt      365:        lva = sokvaalloc(len, so);
1.64      thorpej   366:        if (lva == 0)
1.80      yamt      367:                return 0;
1.64      thorpej   368:
1.116     yamt      369:        error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79      thorpej   370:            m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64      thorpej   371:        if (error) {
1.80      yamt      372:                sokvafree(lva, len);
1.64      thorpej   373:                return (0);
                    374:        }
                    375:
                    376:        for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79      thorpej   377:                pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
                    378:                    VM_PROT_READ);
1.64      thorpej   379:        pmap_update(pmap_kernel());
                    380:
                    381:        lva += (vaddr_t) iov->iov_base & PAGE_MASK;
                    382:
1.134     christos  383:        MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79      thorpej   384:        m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64      thorpej   385:
                    386:        uio->uio_resid -= space;
                    387:        /* uio_offset not updated, not set/used for write(2) */
1.134     christos  388:        uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64      thorpej   389:        uio->uio_iov->iov_len -= space;
                    390:        if (uio->uio_iov->iov_len == 0) {
                    391:                uio->uio_iov++;
                    392:                uio->uio_iovcnt--;
                    393:        }
                    394:
                    395:        return (space);
                    396: }
                    397:
1.119     yamt      398: static int
1.129     yamt      399: sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg)
1.119     yamt      400: {
                    401:
                    402:        KASSERT(ce == &sokva_reclaimerentry);
                    403:        KASSERT(obj == NULL);
                    404:
                    405:        sodopendfree();
                    406:        if (!vm_map_starved_p(kernel_map)) {
                    407:                return CALLBACK_CHAIN_ABORT;
                    408:        }
                    409:        return CALLBACK_CHAIN_CONTINUE;
                    410: }
                    411:
1.142     dyoung    412: struct mbuf *
1.147     dyoung    413: getsombuf(struct socket *so, int type)
1.142     dyoung    414: {
                    415:        struct mbuf *m;
                    416:
1.147     dyoung    417:        m = m_get(M_WAIT, type);
1.142     dyoung    418:        MCLAIM(m, so->so_mowner);
                    419:        return m;
                    420: }
                    421:
1.119     yamt      422: void
                    423: soinit(void)
                    424: {
                    425:
1.148     ad        426:        mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.160     ad        427:        softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.136     ad        428:        cv_init(&socurkva_cv, "sokva");
1.166     ad        429:        soinit2();
1.136     ad        430:
1.119     yamt      431:        /* Set the initial adjusted socket buffer size. */
                    432:        if (sb_max_set(sb_max))
                    433:                panic("bad initial sb_max value: %lu", sb_max);
                    434:
                    435:        callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback,
                    436:            &sokva_reclaimerentry, NULL, sokva_reclaim_callback);
                    437: }
                    438:
1.1       cgd       439: /*
                    440:  * Socket operation routines.
                    441:  * These routines are called by the routines in
                    442:  * sys_socket.c or from a system process, and
                    443:  * implement the semantics of socket operations by
                    444:  * switching out to the protocol specific routines.
                    445:  */
                    446: /*ARGSUSED*/
1.3       andrew    447: int
1.160     ad        448: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l,
                    449:         struct socket *lockso)
1.1       cgd       450: {
1.99      matt      451:        const struct protosw    *prp;
1.54      lukem     452:        struct socket   *so;
1.115     yamt      453:        uid_t           uid;
1.160     ad        454:        int             error;
                    455:        kmutex_t        *lock;
1.1       cgd       456:
1.132     elad      457:        error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
                    458:            KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
                    459:            KAUTH_ARG(proto));
1.140     dyoung    460:        if (error != 0)
                    461:                return error;
1.127     elad      462:
1.1       cgd       463:        if (proto)
                    464:                prp = pffindproto(dom, proto, type);
                    465:        else
                    466:                prp = pffindtype(dom, type);
1.140     dyoung    467:        if (prp == NULL) {
1.120     ginsbach  468:                /* no support for domain */
                    469:                if (pffinddomain(dom) == 0)
1.140     dyoung    470:                        return EAFNOSUPPORT;
1.120     ginsbach  471:                /* no support for socket type */
                    472:                if (proto == 0 && type != 0)
1.140     dyoung    473:                        return EPROTOTYPE;
                    474:                return EPROTONOSUPPORT;
1.120     ginsbach  475:        }
1.140     dyoung    476:        if (prp->pr_usrreq == NULL)
                    477:                return EPROTONOSUPPORT;
1.1       cgd       478:        if (prp->pr_type != type)
1.140     dyoung    479:                return EPROTOTYPE;
1.160     ad        480:
                    481:        so = soget(true);
1.1       cgd       482:        so->so_type = type;
                    483:        so->so_proto = prp;
1.33      matt      484:        so->so_send = sosend;
                    485:        so->so_receive = soreceive;
1.78      matt      486: #ifdef MBUFTRACE
                    487:        so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
                    488:        so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
                    489:        so->so_mowner = &prp->pr_domain->dom_mowner;
                    490: #endif
1.138     rmind     491:        uid = kauth_cred_geteuid(l->l_cred);
1.115     yamt      492:        so->so_uidinfo = uid_find(uid);
1.168     yamt      493:        so->so_egid = kauth_cred_getegid(l->l_cred);
                    494:        so->so_cpid = l->l_proc->p_pid;
1.160     ad        495:        if (lockso != NULL) {
                    496:                /* Caller wants us to share a lock. */
                    497:                lock = lockso->so_lock;
                    498:                so->so_lock = lock;
                    499:                mutex_obj_hold(lock);
                    500:                mutex_enter(lock);
                    501:        } else {
                    502:                /* Lock assigned and taken during PRU_ATTACH. */
                    503:        }
1.140     dyoung    504:        error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
                    505:            (struct mbuf *)(long)proto, NULL, l);
1.160     ad        506:        KASSERT(solocked(so));
1.140     dyoung    507:        if (error != 0) {
1.1       cgd       508:                so->so_state |= SS_NOFDREF;
                    509:                sofree(so);
1.140     dyoung    510:                return error;
1.1       cgd       511:        }
1.160     ad        512:        sounlock(so);
1.1       cgd       513:        *aso = so;
1.140     dyoung    514:        return 0;
1.1       cgd       515: }
                    516:
1.142     dyoung    517: /* On success, write file descriptor to fdout and return zero.  On
                    518:  * failure, return non-zero; *fdout will be undefined.
                    519:  */
                    520: int
                    521: fsocreate(int domain, struct socket **sop, int type, int protocol,
                    522:     struct lwp *l, int *fdout)
                    523: {
                    524:        struct socket   *so;
                    525:        struct file     *fp;
                    526:        int             fd, error;
                    527:
1.155     ad        528:        if ((error = fd_allocfile(&fp, &fd)) != 0)
1.142     dyoung    529:                return (error);
                    530:        fp->f_flag = FREAD|FWRITE;
                    531:        fp->f_type = DTYPE_SOCKET;
                    532:        fp->f_ops = &socketops;
1.160     ad        533:        error = socreate(domain, &so, type, protocol, l, NULL);
1.142     dyoung    534:        if (error != 0) {
1.155     ad        535:                fd_abort(curproc, fp, fd);
1.142     dyoung    536:        } else {
                    537:                if (sop != NULL)
                    538:                        *sop = so;
                    539:                fp->f_data = so;
1.155     ad        540:                fd_affix(curproc, fp, fd);
1.142     dyoung    541:                *fdout = fd;
                    542:        }
                    543:        return error;
                    544: }
                    545:
1.3       andrew    546: int
1.114     christos  547: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       548: {
1.160     ad        549:        int     error;
1.1       cgd       550:
1.160     ad        551:        solock(so);
1.140     dyoung    552:        error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.160     ad        553:        sounlock(so);
1.140     dyoung    554:        return error;
1.1       cgd       555: }
                    556:
1.3       andrew    557: int
1.150     elad      558: solisten(struct socket *so, int backlog, struct lwp *l)
1.1       cgd       559: {
1.160     ad        560:        int     error;
1.1       cgd       561:
1.160     ad        562:        solock(so);
1.158     ad        563:        if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
1.163     ad        564:            SS_ISDISCONNECTING)) != 0) {
                    565:                sounlock(so);
1.158     ad        566:                return (EOPNOTSUPP);
1.163     ad        567:        }
1.140     dyoung    568:        error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
1.150     elad      569:            NULL, NULL, l);
1.140     dyoung    570:        if (error != 0) {
1.160     ad        571:                sounlock(so);
1.140     dyoung    572:                return error;
1.1       cgd       573:        }
1.63      matt      574:        if (TAILQ_EMPTY(&so->so_q))
1.1       cgd       575:                so->so_options |= SO_ACCEPTCONN;
                    576:        if (backlog < 0)
                    577:                backlog = 0;
1.49      jonathan  578:        so->so_qlimit = min(backlog, somaxconn);
1.160     ad        579:        sounlock(so);
1.140     dyoung    580:        return 0;
1.1       cgd       581: }
                    582:
1.21      christos  583: void
1.54      lukem     584: sofree(struct socket *so)
1.1       cgd       585: {
1.161     ad        586:        u_int refs;
1.1       cgd       587:
1.160     ad        588:        KASSERT(solocked(so));
                    589:
                    590:        if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) {
                    591:                sounlock(so);
1.1       cgd       592:                return;
1.160     ad        593:        }
1.43      mycroft   594:        if (so->so_head) {
                    595:                /*
                    596:                 * We must not decommission a socket that's on the accept(2)
                    597:                 * queue.  If we do, then accept(2) may hang after select(2)
                    598:                 * indicated that the listening socket was ready.
                    599:                 */
1.160     ad        600:                if (!soqremque(so, 0)) {
                    601:                        sounlock(so);
1.43      mycroft   602:                        return;
1.160     ad        603:                }
1.43      mycroft   604:        }
1.98      christos  605:        if (so->so_rcv.sb_hiwat)
1.110     christos  606:                (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98      christos  607:                    RLIM_INFINITY);
                    608:        if (so->so_snd.sb_hiwat)
1.110     christos  609:                (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98      christos  610:                    RLIM_INFINITY);
                    611:        sbrelease(&so->so_snd, so);
1.160     ad        612:        KASSERT(!cv_has_waiters(&so->so_cv));
                    613:        KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv));
                    614:        KASSERT(!cv_has_waiters(&so->so_snd.sb_cv));
1.1       cgd       615:        sorflush(so);
1.161     ad        616:        refs = so->so_aborting; /* XXX */
1.170     tls       617: #ifdef INET
                    618:        /* remove acccept filter if one is present. */
                    619:        if (so->so_accf != NULL)
                    620:                do_setopt_accept_filter(so, NULL);
                    621: #endif
1.160     ad        622:        sounlock(so);
1.161     ad        623:        if (refs == 0)          /* XXX */
                    624:                soput(so);
1.1       cgd       625: }
                    626:
                    627: /*
                    628:  * Close a socket on last file table reference removal.
                    629:  * Initiate disconnect if connected.
                    630:  * Free socket when disconnect complete.
                    631:  */
1.3       andrew    632: int
1.54      lukem     633: soclose(struct socket *so)
1.1       cgd       634: {
1.54      lukem     635:        struct socket   *so2;
1.160     ad        636:        int             error;
                    637:        int             error2;
1.1       cgd       638:
1.54      lukem     639:        error = 0;
1.160     ad        640:        solock(so);
1.1       cgd       641:        if (so->so_options & SO_ACCEPTCONN) {
1.172   ! ad        642:                for (;;) {
        !           643:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.160     ad        644:                                KASSERT(solocked2(so, so2));
                    645:                                (void) soqremque(so2, 0);
                    646:                                /* soabort drops the lock. */
                    647:                                (void) soabort(so2);
                    648:                                solock(so);
1.172   ! ad        649:                                continue;
1.160     ad        650:                        }
1.172   ! ad        651:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.160     ad        652:                                KASSERT(solocked2(so, so2));
                    653:                                (void) soqremque(so2, 1);
                    654:                                /* soabort drops the lock. */
                    655:                                (void) soabort(so2);
                    656:                                solock(so);
1.172   ! ad        657:                                continue;
1.160     ad        658:                        }
1.172   ! ad        659:                        break;
        !           660:                }
1.1       cgd       661:        }
                    662:        if (so->so_pcb == 0)
                    663:                goto discard;
                    664:        if (so->so_state & SS_ISCONNECTED) {
                    665:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    666:                        error = sodisconnect(so);
                    667:                        if (error)
                    668:                                goto drop;
                    669:                }
                    670:                if (so->so_options & SO_LINGER) {
1.151     ad        671:                        if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio)
1.1       cgd       672:                                goto drop;
1.21      christos  673:                        while (so->so_state & SS_ISCONNECTED) {
1.160     ad        674:                                error = sowait(so, so->so_linger * hz);
1.21      christos  675:                                if (error)
1.1       cgd       676:                                        break;
1.21      christos  677:                        }
1.1       cgd       678:                }
                    679:        }
1.54      lukem     680:  drop:
1.1       cgd       681:        if (so->so_pcb) {
1.160     ad        682:                error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140     dyoung    683:                    NULL, NULL, NULL, NULL);
1.1       cgd       684:                if (error == 0)
                    685:                        error = error2;
                    686:        }
1.54      lukem     687:  discard:
1.1       cgd       688:        if (so->so_state & SS_NOFDREF)
                    689:                panic("soclose: NOFDREF");
                    690:        so->so_state |= SS_NOFDREF;
                    691:        sofree(so);
                    692:        return (error);
                    693: }
                    694:
                    695: /*
1.160     ad        696:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       697:  */
1.3       andrew    698: int
1.54      lukem     699: soabort(struct socket *so)
1.1       cgd       700: {
1.161     ad        701:        u_int refs;
1.139     yamt      702:        int error;
1.160     ad        703:
                    704:        KASSERT(solocked(so));
                    705:        KASSERT(so->so_head == NULL);
1.1       cgd       706:
1.161     ad        707:        so->so_aborting++;              /* XXX */
1.140     dyoung    708:        error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
                    709:            NULL, NULL, NULL);
1.161     ad        710:        refs = --so->so_aborting;       /* XXX */
1.164     drochner  711:        if (error || (refs == 0)) {
1.139     yamt      712:                sofree(so);
1.160     ad        713:        } else {
                    714:                sounlock(so);
1.139     yamt      715:        }
                    716:        return error;
1.1       cgd       717: }
                    718:
1.3       andrew    719: int
1.54      lukem     720: soaccept(struct socket *so, struct mbuf *nam)
1.1       cgd       721: {
1.160     ad        722:        int     error;
                    723:
                    724:        KASSERT(solocked(so));
1.1       cgd       725:
1.54      lukem     726:        error = 0;
1.1       cgd       727:        if ((so->so_state & SS_NOFDREF) == 0)
                    728:                panic("soaccept: !NOFDREF");
                    729:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   730:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    731:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41      mycroft   732:                error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140     dyoung    733:                    NULL, nam, NULL, NULL);
1.41      mycroft   734:        else
1.53      itojun    735:                error = ECONNABORTED;
1.52      itojun    736:
1.1       cgd       737:        return (error);
                    738: }
                    739:
1.3       andrew    740: int
1.114     christos  741: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       742: {
1.160     ad        743:        int             error;
                    744:
                    745:        KASSERT(solocked(so));
1.1       cgd       746:
                    747:        if (so->so_options & SO_ACCEPTCONN)
                    748:                return (EOPNOTSUPP);
                    749:        /*
                    750:         * If protocol is connection-based, can only connect once.
                    751:         * Otherwise, if connected, try to disconnect first.
                    752:         * This allows user to disconnect by connecting to, e.g.,
                    753:         * a null address.
                    754:         */
                    755:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    756:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
                    757:            (error = sodisconnect(so))))
                    758:                error = EISCONN;
                    759:        else
                    760:                error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140     dyoung    761:                    NULL, nam, NULL, l);
1.1       cgd       762:        return (error);
                    763: }
                    764:
1.3       andrew    765: int
1.54      lukem     766: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       767: {
1.160     ad        768:        int     error;
                    769:
                    770:        KASSERT(solocked2(so1, so2));
1.1       cgd       771:
1.22      mycroft   772:        error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140     dyoung    773:            NULL, (struct mbuf *)so2, NULL, NULL);
1.1       cgd       774:        return (error);
                    775: }
                    776:
1.3       andrew    777: int
1.54      lukem     778: sodisconnect(struct socket *so)
1.1       cgd       779: {
1.160     ad        780:        int     error;
                    781:
                    782:        KASSERT(solocked(so));
1.1       cgd       783:
                    784:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    785:                error = ENOTCONN;
1.160     ad        786:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       787:                error = EALREADY;
1.160     ad        788:        } else {
                    789:                error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
                    790:                    NULL, NULL, NULL, NULL);
1.1       cgd       791:        }
1.117     yamt      792:        sodopendfree();
1.1       cgd       793:        return (error);
                    794: }
                    795:
1.15      mycroft   796: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       797: /*
                    798:  * Send on a socket.
                    799:  * If send must go all at once and message is larger than
                    800:  * send buffering, then hard error.
                    801:  * Lock against other senders.
                    802:  * If must go all at once and not enough room now, then
                    803:  * inform user that this would block and do nothing.
                    804:  * Otherwise, if nonblocking, send as much as possible.
                    805:  * The data to be sent is described by "uio" if nonzero,
                    806:  * otherwise by the mbuf chain "top" (which must be null
                    807:  * if uio is not).  Data provided in mbuf chain must be small
                    808:  * enough to send all at once.
                    809:  *
                    810:  * Returns nonzero on error, timeout or signal; callers
                    811:  * must check for short counts if EINTR/ERESTART are returned.
                    812:  * Data and control buffers are freed on return.
                    813:  */
1.3       andrew    814: int
1.54      lukem     815: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114     christos  816:        struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       817: {
1.54      lukem     818:        struct mbuf     **mp, *m;
1.114     christos  819:        struct proc     *p;
1.58      jdolecek  820:        long            space, len, resid, clen, mlen;
                    821:        int             error, s, dontroute, atomic;
1.54      lukem     822:
1.114     christos  823:        p = l->l_proc;
1.117     yamt      824:        sodopendfree();
1.160     ad        825:        clen = 0;
1.64      thorpej   826:
1.160     ad        827:        /*
                    828:         * solock() provides atomicity of access.  splsoftnet() prevents
                    829:         * protocol processing soft interrupts from interrupting us and
                    830:         * blocking (expensive).
                    831:         */
                    832:        s = splsoftnet();
                    833:        solock(so);
1.54      lukem     834:        atomic = sosendallatonce(so) || top;
1.1       cgd       835:        if (uio)
                    836:                resid = uio->uio_resid;
                    837:        else
                    838:                resid = top->m_pkthdr.len;
1.7       cgd       839:        /*
                    840:         * In theory resid should be unsigned.
                    841:         * However, space must be signed, as it might be less than 0
                    842:         * if we over-committed, and we must use a signed comparison
                    843:         * of space and resid.  On the other hand, a negative resid
                    844:         * causes us to loop sending 0-length segments to the protocol.
                    845:         */
1.29      mycroft   846:        if (resid < 0) {
                    847:                error = EINVAL;
                    848:                goto out;
                    849:        }
1.1       cgd       850:        dontroute =
                    851:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    852:            (so->so_proto->pr_flags & PR_ATOMIC);
1.165     christos  853:        l->l_ru.ru_msgsnd++;
1.1       cgd       854:        if (control)
                    855:                clen = control->m_len;
1.54      lukem     856:  restart:
1.21      christos  857:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       858:                goto out;
                    859:        do {
1.160     ad        860:                if (so->so_state & SS_CANTSENDMORE) {
                    861:                        error = EPIPE;
                    862:                        goto release;
                    863:                }
1.48      thorpej   864:                if (so->so_error) {
                    865:                        error = so->so_error;
                    866:                        so->so_error = 0;
                    867:                        goto release;
                    868:                }
1.1       cgd       869:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    870:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
                    871:                                if ((so->so_state & SS_ISCONFIRMING) == 0 &&
1.160     ad        872:                                    !(resid == 0 && clen != 0)) {
                    873:                                        error = ENOTCONN;
                    874:                                        goto release;
                    875:                                }
                    876:                        } else if (addr == 0) {
                    877:                                error = EDESTADDRREQ;
                    878:                                goto release;
                    879:                        }
1.1       cgd       880:                }
                    881:                space = sbspace(&so->so_snd);
                    882:                if (flags & MSG_OOB)
                    883:                        space += 1024;
1.21      christos  884:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        885:                    clen > so->so_snd.sb_hiwat) {
                    886:                        error = EMSGSIZE;
                    887:                        goto release;
                    888:                }
1.96      mycroft   889:                if (space < resid + clen &&
1.1       cgd       890:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.160     ad        891:                        if (so->so_nbio) {
                    892:                                error = EWOULDBLOCK;
                    893:                                goto release;
                    894:                        }
1.1       cgd       895:                        sbunlock(&so->so_snd);
                    896:                        error = sbwait(&so->so_snd);
                    897:                        if (error)
                    898:                                goto out;
                    899:                        goto restart;
                    900:                }
                    901:                mp = &top;
                    902:                space -= clen;
                    903:                do {
1.45      tv        904:                        if (uio == NULL) {
                    905:                                /*
                    906:                                 * Data is prepackaged in "top".
                    907:                                 */
                    908:                                resid = 0;
                    909:                                if (flags & MSG_EOR)
                    910:                                        top->m_flags |= M_EOR;
                    911:                        } else do {
1.160     ad        912:                                sounlock(so);
                    913:                                splx(s);
1.144     dyoung    914:                                if (top == NULL) {
1.78      matt      915:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv        916:                                        mlen = MHLEN;
                    917:                                        m->m_pkthdr.len = 0;
1.140     dyoung    918:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv        919:                                } else {
1.78      matt      920:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv        921:                                        mlen = MLEN;
                    922:                                }
1.78      matt      923:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt      924:                                if (sock_loan_thresh >= 0 &&
                    925:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                    926:                                    space >= sock_loan_thresh &&
1.64      thorpej   927:                                    (len = sosend_loan(so, uio, m,
                    928:                                                       space)) != 0) {
                    929:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                    930:                                        space -= len;
                    931:                                        goto have_data;
                    932:                                }
1.45      tv        933:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej   934:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78      matt      935:                                        m_clget(m, M_WAIT);
1.45      tv        936:                                        if ((m->m_flags & M_EXT) == 0)
                    937:                                                goto nopages;
                    938:                                        mlen = MCLBYTES;
                    939:                                        if (atomic && top == 0) {
1.58      jdolecek  940:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem     941:                                                    resid);
1.45      tv        942:                                                m->m_data += max_hdr;
                    943:                                        } else
1.58      jdolecek  944:                                                len = lmin(MCLBYTES, resid);
1.45      tv        945:                                        space -= len;
                    946:                                } else {
1.64      thorpej   947:  nopages:
                    948:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek  949:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv        950:                                        space -= len;
                    951:                                        /*
                    952:                                         * For datagram protocols, leave room
                    953:                                         * for protocol headers in first mbuf.
                    954:                                         */
                    955:                                        if (atomic && top == 0 && len < mlen)
                    956:                                                MH_ALIGN(m, len);
                    957:                                }
1.144     dyoung    958:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej   959:  have_data:
1.45      tv        960:                                resid = uio->uio_resid;
                    961:                                m->m_len = len;
                    962:                                *mp = m;
                    963:                                top->m_pkthdr.len += len;
1.160     ad        964:                                s = splsoftnet();
                    965:                                solock(so);
1.144     dyoung    966:                                if (error != 0)
1.45      tv        967:                                        goto release;
                    968:                                mp = &m->m_next;
                    969:                                if (resid <= 0) {
                    970:                                        if (flags & MSG_EOR)
                    971:                                                top->m_flags |= M_EOR;
                    972:                                        break;
                    973:                                }
                    974:                        } while (space > 0 && atomic);
1.108     perry     975:
1.160     ad        976:                        if (so->so_state & SS_CANTSENDMORE) {
                    977:                                error = EPIPE;
                    978:                                goto release;
                    979:                        }
1.45      tv        980:                        if (dontroute)
                    981:                                so->so_options |= SO_DONTROUTE;
                    982:                        if (resid > 0)
                    983:                                so->so_state |= SS_MORETOCOME;
1.46      sommerfe  984:                        error = (*so->so_proto->pr_usrreq)(so,
                    985:                            (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.160     ad        986:                            top, addr, control, curlwp);
1.45      tv        987:                        if (dontroute)
                    988:                                so->so_options &= ~SO_DONTROUTE;
                    989:                        if (resid > 0)
                    990:                                so->so_state &= ~SS_MORETOCOME;
                    991:                        clen = 0;
1.144     dyoung    992:                        control = NULL;
                    993:                        top = NULL;
1.45      tv        994:                        mp = &top;
1.144     dyoung    995:                        if (error != 0)
1.1       cgd       996:                                goto release;
                    997:                } while (resid && space > 0);
                    998:        } while (resid);
                    999:
1.54      lukem    1000:  release:
1.1       cgd      1001:        sbunlock(&so->so_snd);
1.54      lukem    1002:  out:
1.160     ad       1003:        sounlock(so);
                   1004:        splx(s);
1.1       cgd      1005:        if (top)
                   1006:                m_freem(top);
                   1007:        if (control)
                   1008:                m_freem(control);
                   1009:        return (error);
                   1010: }
                   1011:
                   1012: /*
1.159     ad       1013:  * Following replacement or removal of the first mbuf on the first
                   1014:  * mbuf chain of a socket buffer, push necessary state changes back
                   1015:  * into the socket buffer so that other consumers see the values
                   1016:  * consistently.  'nextrecord' is the callers locally stored value of
                   1017:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1018:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1019:  */
                   1020: static void
                   1021: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1022: {
                   1023:
1.160     ad       1024:        KASSERT(solocked(sb->sb_so));
                   1025:
1.159     ad       1026:        /*
                   1027:         * First, update for the new value of nextrecord.  If necessary,
                   1028:         * make it the first record.
                   1029:         */
                   1030:        if (sb->sb_mb != NULL)
                   1031:                sb->sb_mb->m_nextpkt = nextrecord;
                   1032:        else
                   1033:                sb->sb_mb = nextrecord;
                   1034:
                   1035:         /*
                   1036:          * Now update any dependent socket buffer fields to reflect
                   1037:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1038:          * the addition of a second clause that takes care of the
                   1039:          * case where sb_mb has been updated, but remains the last
                   1040:          * record.
                   1041:          */
                   1042:         if (sb->sb_mb == NULL) {
                   1043:                 sb->sb_mbtail = NULL;
                   1044:                 sb->sb_lastrecord = NULL;
                   1045:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1046:                 sb->sb_lastrecord = sb->sb_mb;
                   1047: }
                   1048:
                   1049: /*
1.1       cgd      1050:  * Implement receive operations on a socket.
                   1051:  * We depend on the way that records are added to the sockbuf
                   1052:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1053:  * must begin with an address if the protocol so specifies,
                   1054:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1055:  * and then zero or more mbufs of data.
                   1056:  * In order to avoid blocking network interrupts for the entire time here,
                   1057:  * we splx() while doing the actual copy to user space.
                   1058:  * Although the sockbuf is locked, new data may still be appended,
                   1059:  * and thus we must maintain consistency of the sockbuf during that time.
                   1060:  *
                   1061:  * The caller may receive the data as a single mbuf chain by supplying
                   1062:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1063:  * only for the count in uio_resid.
                   1064:  */
1.3       andrew   1065: int
1.54      lukem    1066: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1067:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1068: {
1.116     yamt     1069:        struct lwp *l = curlwp;
1.160     ad       1070:        struct mbuf     *m, **mp, *mt;
1.146     dyoung   1071:        int atomic, flags, len, error, s, offset, moff, type, orig_resid;
1.99      matt     1072:        const struct protosw    *pr;
1.54      lukem    1073:        struct mbuf     *nextrecord;
1.67      he       1074:        int             mbuf_removed = 0;
1.146     dyoung   1075:        const struct domain *dom;
1.64      thorpej  1076:
1.54      lukem    1077:        pr = so->so_proto;
1.146     dyoung   1078:        atomic = pr->pr_flags & PR_ATOMIC;
                   1079:        dom = pr->pr_domain;
1.1       cgd      1080:        mp = mp0;
1.54      lukem    1081:        type = 0;
                   1082:        orig_resid = uio->uio_resid;
1.102     jonathan 1083:
1.144     dyoung   1084:        if (paddr != NULL)
                   1085:                *paddr = NULL;
                   1086:        if (controlp != NULL)
                   1087:                *controlp = NULL;
                   1088:        if (flagsp != NULL)
1.1       cgd      1089:                flags = *flagsp &~ MSG_EOR;
                   1090:        else
                   1091:                flags = 0;
1.66      enami    1092:
                   1093:        if ((flags & MSG_DONTWAIT) == 0)
1.117     yamt     1094:                sodopendfree();
1.66      enami    1095:
1.1       cgd      1096:        if (flags & MSG_OOB) {
                   1097:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1098:                solock(so);
1.17      cgd      1099:                error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140     dyoung   1100:                    (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.160     ad       1101:                sounlock(so);
1.1       cgd      1102:                if (error)
                   1103:                        goto bad;
                   1104:                do {
1.134     christos 1105:                        error = uiomove(mtod(m, void *),
1.1       cgd      1106:                            (int) min(uio->uio_resid, m->m_len), uio);
                   1107:                        m = m_free(m);
1.144     dyoung   1108:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1109:  bad:
1.144     dyoung   1110:                if (m != NULL)
1.1       cgd      1111:                        m_freem(m);
1.144     dyoung   1112:                return error;
1.1       cgd      1113:        }
1.144     dyoung   1114:        if (mp != NULL)
1.140     dyoung   1115:                *mp = NULL;
1.160     ad       1116:
                   1117:        /*
                   1118:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1119:         * protocol processing soft interrupts from interrupting us and
                   1120:         * blocking (expensive).
                   1121:         */
                   1122:        s = splsoftnet();
                   1123:        solock(so);
1.1       cgd      1124:        if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140     dyoung   1125:                (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1       cgd      1126:
1.54      lukem    1127:  restart:
1.160     ad       1128:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1129:                sounlock(so);
                   1130:                splx(s);
1.144     dyoung   1131:                return error;
1.160     ad       1132:        }
1.1       cgd      1133:
                   1134:        m = so->so_rcv.sb_mb;
                   1135:        /*
                   1136:         * If we have less data than requested, block awaiting more
                   1137:         * (subject to any timeout) if:
1.15      mycroft  1138:         *   1. the current count is less than the low water mark,
1.1       cgd      1139:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1140:         *      receive operation at once if we block (resid <= hiwat), or
                   1141:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1142:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1143:         * we have to do the receive in sections, and thus risk returning
                   1144:         * a short count if a timeout or signal occurs after we start.
                   1145:         */
1.144     dyoung   1146:        if (m == NULL ||
                   1147:            ((flags & MSG_DONTWAIT) == 0 &&
                   1148:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1149:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1150:              ((flags & MSG_WAITALL) &&
                   1151:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1152:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1153: #ifdef DIAGNOSTIC
1.144     dyoung   1154:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1155:                        panic("receive 1");
                   1156: #endif
                   1157:                if (so->so_error) {
1.144     dyoung   1158:                        if (m != NULL)
1.15      mycroft  1159:                                goto dontblock;
1.1       cgd      1160:                        error = so->so_error;
                   1161:                        if ((flags & MSG_PEEK) == 0)
                   1162:                                so->so_error = 0;
                   1163:                        goto release;
                   1164:                }
                   1165:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1166:                        if (m != NULL)
1.15      mycroft  1167:                                goto dontblock;
1.1       cgd      1168:                        else
                   1169:                                goto release;
                   1170:                }
1.144     dyoung   1171:                for (; m != NULL; m = m->m_next)
1.1       cgd      1172:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1173:                                m = so->so_rcv.sb_mb;
                   1174:                                goto dontblock;
                   1175:                        }
                   1176:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1177:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1178:                        error = ENOTCONN;
                   1179:                        goto release;
                   1180:                }
                   1181:                if (uio->uio_resid == 0)
                   1182:                        goto release;
1.151     ad       1183:                if (so->so_nbio || (flags & MSG_DONTWAIT)) {
1.1       cgd      1184:                        error = EWOULDBLOCK;
                   1185:                        goto release;
                   1186:                }
1.69      thorpej  1187:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1188:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1189:                sbunlock(&so->so_rcv);
                   1190:                error = sbwait(&so->so_rcv);
1.160     ad       1191:                if (error != 0) {
                   1192:                        sounlock(so);
                   1193:                        splx(s);
1.144     dyoung   1194:                        return error;
1.160     ad       1195:                }
1.1       cgd      1196:                goto restart;
                   1197:        }
1.54      lukem    1198:  dontblock:
1.69      thorpej  1199:        /*
                   1200:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1201:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1202:         * pointer to the next record in the socket buffer.  We must keep the
                   1203:         * various socket buffer pointers and local stack versions of the
                   1204:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1205:         * socket lock, and re-reading them when picking it up.
1.159     ad       1206:         *
                   1207:         * Otherwise, we will race with the network stack appending new data
                   1208:         * or records onto the socket buffer by using inconsistent/stale
                   1209:         * versions of the field, possibly resulting in socket buffer
                   1210:         * corruption.
                   1211:         *
                   1212:         * By holding the high-level sblock(), we prevent simultaneous
                   1213:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1214:         */
1.144     dyoung   1215:        if (l != NULL)
1.157     ad       1216:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1217:        KASSERT(m == so->so_rcv.sb_mb);
                   1218:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1219:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1220:        nextrecord = m->m_nextpkt;
                   1221:        if (pr->pr_flags & PR_ADDR) {
                   1222: #ifdef DIAGNOSTIC
                   1223:                if (m->m_type != MT_SONAME)
                   1224:                        panic("receive 1a");
                   1225: #endif
1.3       andrew   1226:                orig_resid = 0;
1.1       cgd      1227:                if (flags & MSG_PEEK) {
                   1228:                        if (paddr)
                   1229:                                *paddr = m_copy(m, 0, m->m_len);
                   1230:                        m = m->m_next;
                   1231:                } else {
                   1232:                        sbfree(&so->so_rcv, m);
1.67      he       1233:                        mbuf_removed = 1;
1.144     dyoung   1234:                        if (paddr != NULL) {
1.1       cgd      1235:                                *paddr = m;
                   1236:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1237:                                m->m_next = NULL;
1.1       cgd      1238:                                m = so->so_rcv.sb_mb;
                   1239:                        } else {
                   1240:                                MFREE(m, so->so_rcv.sb_mb);
                   1241:                                m = so->so_rcv.sb_mb;
                   1242:                        }
1.159     ad       1243:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1244:                }
                   1245:        }
1.159     ad       1246:
                   1247:        /*
                   1248:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1249:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1250:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1251:         * perform externalization (or freeing if controlp == NULL).
                   1252:         */
                   1253:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1254:                struct mbuf *cm = NULL, *cmn;
                   1255:                struct mbuf **cme = &cm;
                   1256:
                   1257:                do {
                   1258:                        if (flags & MSG_PEEK) {
                   1259:                                if (controlp != NULL) {
                   1260:                                        *controlp = m_copy(m, 0, m->m_len);
                   1261:                                        controlp = &(*controlp)->m_next;
                   1262:                                }
                   1263:                                m = m->m_next;
                   1264:                        } else {
                   1265:                                sbfree(&so->so_rcv, m);
1.1       cgd      1266:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1267:                                m->m_next = NULL;
1.159     ad       1268:                                *cme = m;
                   1269:                                cme = &(*cme)->m_next;
1.1       cgd      1270:                                m = so->so_rcv.sb_mb;
1.159     ad       1271:                        }
                   1272:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1273:                if ((flags & MSG_PEEK) == 0)
                   1274:                        sbsync(&so->so_rcv, nextrecord);
                   1275:                for (; cm != NULL; cm = cmn) {
                   1276:                        cmn = cm->m_next;
                   1277:                        cm->m_next = NULL;
                   1278:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1279:                        if (controlp != NULL) {
                   1280:                                if (dom->dom_externalize != NULL &&
                   1281:                                    type == SCM_RIGHTS) {
1.160     ad       1282:                                        sounlock(so);
1.159     ad       1283:                                        splx(s);
                   1284:                                        error = (*dom->dom_externalize)(cm, l);
                   1285:                                        s = splsoftnet();
1.160     ad       1286:                                        solock(so);
1.159     ad       1287:                                }
                   1288:                                *controlp = cm;
                   1289:                                while (*controlp != NULL)
                   1290:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1291:                        } else {
1.106     itojun   1292:                                /*
                   1293:                                 * Dispose of any SCM_RIGHTS message that went
                   1294:                                 * through the read path rather than recv.
                   1295:                                 */
1.159     ad       1296:                                if (dom->dom_dispose != NULL &&
                   1297:                                    type == SCM_RIGHTS) {
1.160     ad       1298:                                        sounlock(so);
1.159     ad       1299:                                        (*dom->dom_dispose)(cm);
1.160     ad       1300:                                        solock(so);
1.159     ad       1301:                                }
                   1302:                                m_freem(cm);
1.1       cgd      1303:                        }
                   1304:                }
1.159     ad       1305:                if (m != NULL)
                   1306:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1307:                else
                   1308:                        nextrecord = so->so_rcv.sb_mb;
                   1309:                orig_resid = 0;
1.1       cgd      1310:        }
1.69      thorpej  1311:
1.159     ad       1312:        /* If m is non-NULL, we have some data to read. */
                   1313:        if (__predict_true(m != NULL)) {
1.1       cgd      1314:                type = m->m_type;
                   1315:                if (type == MT_OOBDATA)
                   1316:                        flags |= MSG_OOB;
                   1317:        }
1.69      thorpej  1318:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1319:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1320:
1.1       cgd      1321:        moff = 0;
                   1322:        offset = 0;
1.144     dyoung   1323:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1324:                if (m->m_type == MT_OOBDATA) {
                   1325:                        if (type != MT_OOBDATA)
                   1326:                                break;
                   1327:                } else if (type == MT_OOBDATA)
                   1328:                        break;
                   1329: #ifdef DIAGNOSTIC
                   1330:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1331:                        panic("receive 3");
                   1332: #endif
                   1333:                so->so_state &= ~SS_RCVATMARK;
                   1334:                len = uio->uio_resid;
                   1335:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1336:                        len = so->so_oobmark - offset;
                   1337:                if (len > m->m_len - moff)
                   1338:                        len = m->m_len - moff;
                   1339:                /*
                   1340:                 * If mp is set, just pass back the mbufs.
                   1341:                 * Otherwise copy them out via the uio, then free.
                   1342:                 * Sockbuf must be consistent here (points to current mbuf,
                   1343:                 * it points to next record) when we drop priority;
                   1344:                 * we must note any additions to the sockbuf when we
                   1345:                 * block interrupts again.
                   1346:                 */
1.144     dyoung   1347:                if (mp == NULL) {
1.69      thorpej  1348:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1349:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1350:                        sounlock(so);
1.1       cgd      1351:                        splx(s);
1.134     christos 1352:                        error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20      mycroft  1353:                        s = splsoftnet();
1.160     ad       1354:                        solock(so);
1.144     dyoung   1355:                        if (error != 0) {
1.67      he       1356:                                /*
                   1357:                                 * If any part of the record has been removed
                   1358:                                 * (such as the MT_SONAME mbuf, which will
                   1359:                                 * happen when PR_ADDR, and thus also
                   1360:                                 * PR_ATOMIC, is set), then drop the entire
                   1361:                                 * record to maintain the atomicity of the
                   1362:                                 * receive operation.
                   1363:                                 *
                   1364:                                 * This avoids a later panic("receive 1a")
                   1365:                                 * when compiled with DIAGNOSTIC.
                   1366:                                 */
1.146     dyoung   1367:                                if (m && mbuf_removed && atomic)
1.67      he       1368:                                        (void) sbdroprecord(&so->so_rcv);
                   1369:
1.57      jdolecek 1370:                                goto release;
1.67      he       1371:                        }
1.1       cgd      1372:                } else
                   1373:                        uio->uio_resid -= len;
                   1374:                if (len == m->m_len - moff) {
                   1375:                        if (m->m_flags & M_EOR)
                   1376:                                flags |= MSG_EOR;
                   1377:                        if (flags & MSG_PEEK) {
                   1378:                                m = m->m_next;
                   1379:                                moff = 0;
                   1380:                        } else {
                   1381:                                nextrecord = m->m_nextpkt;
                   1382:                                sbfree(&so->so_rcv, m);
                   1383:                                if (mp) {
                   1384:                                        *mp = m;
                   1385:                                        mp = &m->m_next;
                   1386:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1387:                                        *mp = NULL;
1.1       cgd      1388:                                } else {
                   1389:                                        MFREE(m, so->so_rcv.sb_mb);
                   1390:                                        m = so->so_rcv.sb_mb;
                   1391:                                }
1.69      thorpej  1392:                                /*
                   1393:                                 * If m != NULL, we also know that
                   1394:                                 * so->so_rcv.sb_mb != NULL.
                   1395:                                 */
                   1396:                                KASSERT(so->so_rcv.sb_mb == m);
                   1397:                                if (m) {
1.1       cgd      1398:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1399:                                        if (nextrecord == NULL)
                   1400:                                                so->so_rcv.sb_lastrecord = m;
                   1401:                                } else {
                   1402:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1403:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1404:                                }
                   1405:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1406:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1407:                        }
1.144     dyoung   1408:                } else if (flags & MSG_PEEK)
                   1409:                        moff += len;
                   1410:                else {
1.160     ad       1411:                        if (mp != NULL) {
                   1412:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1413:                                if (__predict_false(mt == NULL)) {
                   1414:                                        sounlock(so);
                   1415:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1416:                                        solock(so);
                   1417:                                }
                   1418:                                *mp = mt;
                   1419:                        }
1.144     dyoung   1420:                        m->m_data += len;
                   1421:                        m->m_len -= len;
                   1422:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1423:                }
                   1424:                if (so->so_oobmark) {
                   1425:                        if ((flags & MSG_PEEK) == 0) {
                   1426:                                so->so_oobmark -= len;
                   1427:                                if (so->so_oobmark == 0) {
                   1428:                                        so->so_state |= SS_RCVATMARK;
                   1429:                                        break;
                   1430:                                }
1.7       cgd      1431:                        } else {
1.1       cgd      1432:                                offset += len;
1.7       cgd      1433:                                if (offset == so->so_oobmark)
                   1434:                                        break;
                   1435:                        }
1.1       cgd      1436:                }
                   1437:                if (flags & MSG_EOR)
                   1438:                        break;
                   1439:                /*
                   1440:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1441:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1442:                 * termination.  If a signal/timeout occurs, return
                   1443:                 * with a short count but without error.
                   1444:                 * Keep sockbuf locked against other readers.
                   1445:                 */
1.144     dyoung   1446:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1447:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1448:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1449:                                break;
1.68      matt     1450:                        /*
                   1451:                         * If we are peeking and the socket receive buffer is
                   1452:                         * full, stop since we can't get more data to peek at.
                   1453:                         */
                   1454:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1455:                                break;
                   1456:                        /*
                   1457:                         * If we've drained the socket buffer, tell the
                   1458:                         * protocol in case it needs to do something to
                   1459:                         * get it filled again.
                   1460:                         */
                   1461:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
                   1462:                                (*pr->pr_usrreq)(so, PRU_RCVD,
1.140     dyoung   1463:                                    NULL, (struct mbuf *)(long)flags, NULL, l);
1.69      thorpej  1464:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1465:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1       cgd      1466:                        error = sbwait(&so->so_rcv);
1.144     dyoung   1467:                        if (error != 0) {
1.1       cgd      1468:                                sbunlock(&so->so_rcv);
1.160     ad       1469:                                sounlock(so);
1.1       cgd      1470:                                splx(s);
1.144     dyoung   1471:                                return 0;
1.1       cgd      1472:                        }
1.21      christos 1473:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1474:                                nextrecord = m->m_nextpkt;
                   1475:                }
                   1476:        }
1.3       andrew   1477:
1.146     dyoung   1478:        if (m && atomic) {
1.3       andrew   1479:                flags |= MSG_TRUNC;
                   1480:                if ((flags & MSG_PEEK) == 0)
                   1481:                        (void) sbdroprecord(&so->so_rcv);
                   1482:        }
1.1       cgd      1483:        if ((flags & MSG_PEEK) == 0) {
1.144     dyoung   1484:                if (m == NULL) {
1.69      thorpej  1485:                        /*
1.70      thorpej  1486:                         * First part is an inline SB_EMPTY_FIXUP().  Second
1.69      thorpej  1487:                         * part makes sure sb_lastrecord is up-to-date if
                   1488:                         * there is still data in the socket buffer.
                   1489:                         */
1.1       cgd      1490:                        so->so_rcv.sb_mb = nextrecord;
1.69      thorpej  1491:                        if (so->so_rcv.sb_mb == NULL) {
                   1492:                                so->so_rcv.sb_mbtail = NULL;
                   1493:                                so->so_rcv.sb_lastrecord = NULL;
                   1494:                        } else if (nextrecord->m_nextpkt == NULL)
                   1495:                                so->so_rcv.sb_lastrecord = nextrecord;
                   1496:                }
                   1497:                SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
                   1498:                SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1       cgd      1499:                if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140     dyoung   1500:                        (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
                   1501:                            (struct mbuf *)(long)flags, NULL, l);
1.1       cgd      1502:        }
1.3       andrew   1503:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1504:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1505:                sbunlock(&so->so_rcv);
                   1506:                goto restart;
                   1507:        }
1.108     perry    1508:
1.144     dyoung   1509:        if (flagsp != NULL)
1.1       cgd      1510:                *flagsp |= flags;
1.54      lukem    1511:  release:
1.1       cgd      1512:        sbunlock(&so->so_rcv);
1.160     ad       1513:        sounlock(so);
1.1       cgd      1514:        splx(s);
1.144     dyoung   1515:        return error;
1.1       cgd      1516: }
                   1517:
1.14      mycroft  1518: int
1.54      lukem    1519: soshutdown(struct socket *so, int how)
1.1       cgd      1520: {
1.99      matt     1521:        const struct protosw    *pr;
1.160     ad       1522:        int     error;
                   1523:
                   1524:        KASSERT(solocked(so));
1.34      kleink   1525:
1.54      lukem    1526:        pr = so->so_proto;
1.34      kleink   1527:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1528:                return (EINVAL);
1.1       cgd      1529:
1.160     ad       1530:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1531:                sorflush(so);
1.160     ad       1532:                error = 0;
                   1533:        }
1.34      kleink   1534:        if (how == SHUT_WR || how == SHUT_RDWR)
1.160     ad       1535:                error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1.140     dyoung   1536:                    NULL, NULL, NULL);
1.160     ad       1537:
                   1538:        return error;
1.1       cgd      1539: }
                   1540:
1.14      mycroft  1541: void
1.54      lukem    1542: sorflush(struct socket *so)
1.1       cgd      1543: {
1.54      lukem    1544:        struct sockbuf  *sb, asb;
1.99      matt     1545:        const struct protosw    *pr;
1.160     ad       1546:
                   1547:        KASSERT(solocked(so));
1.1       cgd      1548:
1.54      lukem    1549:        sb = &so->so_rcv;
                   1550:        pr = so->so_proto;
1.160     ad       1551:        socantrcvmore(so);
1.1       cgd      1552:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1553:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1554:        sbunlock(sb);
                   1555:        asb = *sb;
1.86      wrstuden 1556:        /*
                   1557:         * Clear most of the sockbuf structure, but leave some of the
                   1558:         * fields valid.
                   1559:         */
                   1560:        memset(&sb->sb_startzero, 0,
                   1561:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1562:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1563:                sounlock(so);
1.1       cgd      1564:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1565:                solock(so);
                   1566:        }
1.98      christos 1567:        sbrelease(&asb, so);
1.1       cgd      1568: }
                   1569:
1.171     plunky   1570: /*
                   1571:  * internal set SOL_SOCKET options
                   1572:  */
1.142     dyoung   1573: static int
1.171     plunky   1574: sosetopt1(struct socket *so, const struct sockopt *sopt)
1.1       cgd      1575: {
1.171     plunky   1576:        int error, optval;
                   1577:        struct linger l;
                   1578:        struct timeval tv;
1.142     dyoung   1579:
1.171     plunky   1580:        switch (sopt->sopt_name) {
1.142     dyoung   1581:
1.170     tls      1582: #ifdef INET
                   1583:        case SO_ACCEPTFILTER:
1.171     plunky   1584:                error = do_setopt_accept_filter(so, sopt);
1.170     tls      1585:                if (error)
                   1586:                        return error;
                   1587:                break;
                   1588: #endif
                   1589:
1.171     plunky   1590:        case SO_LINGER:
                   1591:                error = sockopt_get(sopt, &l, sizeof(l));
                   1592:                if (error)
                   1593:                        return (error);
                   1594:
                   1595:                if (l.l_linger < 0 || l.l_linger > USHRT_MAX ||
                   1596:                    l.l_linger > (INT_MAX / hz))
1.142     dyoung   1597:                        return EDOM;
1.171     plunky   1598:                so->so_linger = l.l_linger;
                   1599:                if (l.l_onoff)
                   1600:                        so->so_options |= SO_LINGER;
                   1601:                else
                   1602:                        so->so_options &= ~SO_LINGER;
                   1603:
                   1604:                break;
1.1       cgd      1605:
1.142     dyoung   1606:        case SO_DEBUG:
                   1607:        case SO_KEEPALIVE:
                   1608:        case SO_DONTROUTE:
                   1609:        case SO_USELOOPBACK:
                   1610:        case SO_BROADCAST:
                   1611:        case SO_REUSEADDR:
                   1612:        case SO_REUSEPORT:
                   1613:        case SO_OOBINLINE:
                   1614:        case SO_TIMESTAMP:
1.171     plunky   1615:                error = sockopt_getint(sopt, &optval);
                   1616:                if (error)
                   1617:                        return (error);
                   1618:
                   1619:                if (optval)
                   1620:                        so->so_options |= sopt->sopt_name;
1.142     dyoung   1621:                else
1.171     plunky   1622:                        so->so_options &= ~sopt->sopt_name;
1.142     dyoung   1623:                break;
                   1624:
                   1625:        case SO_SNDBUF:
                   1626:        case SO_RCVBUF:
                   1627:        case SO_SNDLOWAT:
                   1628:        case SO_RCVLOWAT:
1.171     plunky   1629:                error = sockopt_getint(sopt, &optval);
                   1630:                if (error)
                   1631:                        return (error);
1.1       cgd      1632:
1.142     dyoung   1633:                /*
                   1634:                 * Values < 1 make no sense for any of these
                   1635:                 * options, so disallow them.
                   1636:                 */
                   1637:                if (optval < 1)
                   1638:                        return EINVAL;
1.1       cgd      1639:
1.171     plunky   1640:                switch (sopt->sopt_name) {
                   1641:                case SO_SNDBUF:
                   1642:                        if (sbreserve(&so->so_snd, (u_long)optval, so) == 0)
                   1643:                                return ENOBUFS;
                   1644:
                   1645:                        so->so_snd.sb_flags &= ~SB_AUTOSIZE;
                   1646:                        break;
1.1       cgd      1647:
                   1648:                case SO_RCVBUF:
1.171     plunky   1649:                        if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0)
1.142     dyoung   1650:                                return ENOBUFS;
1.171     plunky   1651:
                   1652:                        so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1.142     dyoung   1653:                        break;
                   1654:
                   1655:                /*
                   1656:                 * Make sure the low-water is never greater than
                   1657:                 * the high-water.
                   1658:                 */
1.1       cgd      1659:                case SO_SNDLOWAT:
1.171     plunky   1660:                        if (optval > so->so_snd.sb_hiwat)
                   1661:                                optval = so->so_snd.sb_hiwat;
                   1662:
                   1663:                        so->so_snd.sb_lowat = optval;
1.142     dyoung   1664:                        break;
1.171     plunky   1665:
1.1       cgd      1666:                case SO_RCVLOWAT:
1.171     plunky   1667:                        if (optval > so->so_rcv.sb_hiwat)
                   1668:                                optval = so->so_rcv.sb_hiwat;
                   1669:
                   1670:                        so->so_rcv.sb_lowat = optval;
1.142     dyoung   1671:                        break;
                   1672:                }
                   1673:                break;
1.28      thorpej  1674:
1.142     dyoung   1675:        case SO_SNDTIMEO:
                   1676:        case SO_RCVTIMEO:
1.171     plunky   1677:                error = sockopt_get(sopt, &tv, sizeof(tv));
                   1678:                if (error)
                   1679:                        return (error);
                   1680:
                   1681:                if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz)
1.142     dyoung   1682:                        return EDOM;
1.28      thorpej  1683:
1.171     plunky   1684:                optval = tv.tv_sec * hz + tv.tv_usec / tick;
                   1685:                if (optval == 0 && tv.tv_usec != 0)
                   1686:                        optval = 1;
1.28      thorpej  1687:
1.171     plunky   1688:                switch (sopt->sopt_name) {
1.142     dyoung   1689:                case SO_SNDTIMEO:
1.171     plunky   1690:                        so->so_snd.sb_timeo = optval;
1.1       cgd      1691:                        break;
                   1692:                case SO_RCVTIMEO:
1.171     plunky   1693:                        so->so_rcv.sb_timeo = optval;
1.142     dyoung   1694:                        break;
                   1695:                }
                   1696:                break;
1.1       cgd      1697:
1.142     dyoung   1698:        default:
                   1699:                return ENOPROTOOPT;
                   1700:        }
                   1701:        return 0;
                   1702: }
1.1       cgd      1703:
1.142     dyoung   1704: int
1.171     plunky   1705: sosetopt(struct socket *so, struct sockopt *sopt)
1.142     dyoung   1706: {
                   1707:        int error, prerr;
1.1       cgd      1708:
1.160     ad       1709:        solock(so);
1.171     plunky   1710:        if (sopt->sopt_level == SOL_SOCKET)
                   1711:                error = sosetopt1(so, sopt);
1.142     dyoung   1712:        else
                   1713:                error = ENOPROTOOPT;
1.1       cgd      1714:
1.142     dyoung   1715:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1716:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1717:                /* give the protocol stack a shot */
1.171     plunky   1718:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt);
1.142     dyoung   1719:                if (prerr == 0)
                   1720:                        error = 0;
                   1721:                else if (prerr != ENOPROTOOPT)
                   1722:                        error = prerr;
1.171     plunky   1723:        }
1.160     ad       1724:        sounlock(so);
1.142     dyoung   1725:        return error;
1.1       cgd      1726: }
                   1727:
1.171     plunky   1728: /*
                   1729:  * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt()
                   1730:  */
                   1731: int
                   1732: so_setsockopt(struct lwp *l, struct socket *so, int level, int name,
                   1733:     const void *val, size_t valsize)
                   1734: {
                   1735:        struct sockopt sopt;
                   1736:        int error;
                   1737:
                   1738:        KASSERT(valsize == 0 || val != NULL);
                   1739:
                   1740:        sockopt_init(&sopt, level, name, valsize);
                   1741:        sockopt_set(&sopt, val, valsize);
                   1742:
                   1743:        error = sosetopt(so, &sopt);
                   1744:
                   1745:        sockopt_destroy(&sopt);
                   1746:
                   1747:        return error;
                   1748: }
                   1749:
                   1750: /*
                   1751:  * internal get SOL_SOCKET options
                   1752:  */
                   1753: static int
                   1754: sogetopt1(struct socket *so, struct sockopt *sopt)
                   1755: {
                   1756:        int error, optval;
                   1757:        struct linger l;
                   1758:        struct timeval tv;
                   1759:
                   1760:        switch (sopt->sopt_name) {
                   1761:
                   1762: #ifdef INET
                   1763:        case SO_ACCEPTFILTER:
                   1764:                error = do_getopt_accept_filter(so, sopt);
                   1765:                break;
                   1766: #endif
                   1767:
                   1768:        case SO_LINGER:
                   1769:                l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0;
                   1770:                l.l_linger = so->so_linger;
                   1771:
                   1772:                error = sockopt_set(sopt, &l, sizeof(l));
                   1773:                break;
                   1774:
                   1775:        case SO_USELOOPBACK:
                   1776:        case SO_DONTROUTE:
                   1777:        case SO_DEBUG:
                   1778:        case SO_KEEPALIVE:
                   1779:        case SO_REUSEADDR:
                   1780:        case SO_REUSEPORT:
                   1781:        case SO_BROADCAST:
                   1782:        case SO_OOBINLINE:
                   1783:        case SO_TIMESTAMP:
                   1784:                error = sockopt_setint(sopt,
                   1785:                    (so->so_options & sopt->sopt_name) ? 1 : 0);
                   1786:                break;
                   1787:
                   1788:        case SO_TYPE:
                   1789:                error = sockopt_setint(sopt, so->so_type);
                   1790:                break;
                   1791:
                   1792:        case SO_ERROR:
                   1793:                error = sockopt_setint(sopt, so->so_error);
                   1794:                so->so_error = 0;
                   1795:                break;
                   1796:
                   1797:        case SO_SNDBUF:
                   1798:                error = sockopt_setint(sopt, so->so_snd.sb_hiwat);
                   1799:                break;
                   1800:
                   1801:        case SO_RCVBUF:
                   1802:                error = sockopt_setint(sopt, so->so_rcv.sb_hiwat);
                   1803:                break;
                   1804:
                   1805:        case SO_SNDLOWAT:
                   1806:                error = sockopt_setint(sopt, so->so_snd.sb_lowat);
                   1807:                break;
                   1808:
                   1809:        case SO_RCVLOWAT:
                   1810:                error = sockopt_setint(sopt, so->so_rcv.sb_lowat);
                   1811:                break;
                   1812:
                   1813:        case SO_SNDTIMEO:
                   1814:        case SO_RCVTIMEO:
                   1815:                optval = (sopt->sopt_name == SO_SNDTIMEO ?
                   1816:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1817:
                   1818:                tv.tv_sec = optval / hz;
                   1819:                tv.tv_usec = (optval % hz) * tick;
                   1820:
                   1821:                error = sockopt_set(sopt, &tv, sizeof(tv));
                   1822:                break;
                   1823:
                   1824:        case SO_OVERFLOWED:
                   1825:                error = sockopt_setint(sopt, so->so_rcv.sb_overflowed);
                   1826:                break;
                   1827:
                   1828:        default:
                   1829:                error = ENOPROTOOPT;
                   1830:                break;
                   1831:        }
                   1832:
                   1833:        return (error);
                   1834: }
                   1835:
1.14      mycroft  1836: int
1.171     plunky   1837: sogetopt(struct socket *so, struct sockopt *sopt)
1.1       cgd      1838: {
1.160     ad       1839:        int             error;
1.1       cgd      1840:
1.160     ad       1841:        solock(so);
1.171     plunky   1842:        if (sopt->sopt_level != SOL_SOCKET) {
1.1       cgd      1843:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       1844:                        error = ((*so->so_proto->pr_ctloutput)
1.171     plunky   1845:                            (PRCO_GETOPT, so, sopt));
1.1       cgd      1846:                } else
1.160     ad       1847:                        error = (ENOPROTOOPT);
1.1       cgd      1848:        } else {
1.171     plunky   1849:                error = sogetopt1(so, sopt);
                   1850:        }
                   1851:        sounlock(so);
                   1852:        return (error);
                   1853: }
                   1854:
                   1855: /*
                   1856:  * alloc sockopt data buffer buffer
                   1857:  *     - will be released at destroy
                   1858:  */
                   1859: static void
                   1860: sockopt_alloc(struct sockopt *sopt, size_t len)
                   1861: {
                   1862:
                   1863:        KASSERT(sopt->sopt_size == 0);
                   1864:
                   1865:        if (len > sizeof(sopt->sopt_buf))
                   1866:                sopt->sopt_data = malloc(len, M_SOOPTS, M_WAITOK | M_ZERO);
                   1867:        else
                   1868:                sopt->sopt_data = sopt->sopt_buf;
                   1869:
                   1870:        sopt->sopt_size = len;
                   1871: }
                   1872:
                   1873: /*
                   1874:  * initialise sockopt storage
                   1875:  */
                   1876: void
                   1877: sockopt_init(struct sockopt *sopt, int level, int name, size_t size)
                   1878: {
1.1       cgd      1879:
1.171     plunky   1880:        memset(sopt, 0, sizeof(*sopt));
1.1       cgd      1881:
1.171     plunky   1882:        sopt->sopt_level = level;
                   1883:        sopt->sopt_name = name;
                   1884:        sockopt_alloc(sopt, size);
                   1885: }
                   1886:
                   1887: /*
                   1888:  * destroy sockopt storage
                   1889:  *     - will release any held memory references
                   1890:  */
                   1891: void
                   1892: sockopt_destroy(struct sockopt *sopt)
                   1893: {
                   1894:
                   1895:        if (sopt->sopt_data != sopt->sopt_buf)
                   1896:                free(sopt->sopt_data, M_SOOPTS);
                   1897:
                   1898:        memset(sopt, 0, sizeof(*sopt));
                   1899: }
                   1900:
                   1901: /*
                   1902:  * set sockopt value
                   1903:  *     - value is copied into sockopt
                   1904:  *     - memory is allocated when necessary
                   1905:  */
                   1906: int
                   1907: sockopt_set(struct sockopt *sopt, const void *buf, size_t len)
                   1908: {
                   1909:
                   1910:        if (sopt->sopt_size == 0)
                   1911:                sockopt_alloc(sopt, len);
                   1912:
                   1913:        KASSERT(sopt->sopt_size == len);
                   1914:        memcpy(sopt->sopt_data, buf, len);
                   1915:        return 0;
                   1916: }
                   1917:
                   1918: /*
                   1919:  * common case of set sockopt integer value
                   1920:  */
                   1921: int
                   1922: sockopt_setint(struct sockopt *sopt, int val)
                   1923: {
                   1924:
                   1925:        return sockopt_set(sopt, &val, sizeof(int));
                   1926: }
                   1927:
                   1928: /*
                   1929:  * get sockopt value
                   1930:  *     - correct size must be given
                   1931:  */
                   1932: int
                   1933: sockopt_get(const struct sockopt *sopt, void *buf, size_t len)
                   1934: {
1.170     tls      1935:
1.171     plunky   1936:        if (sopt->sopt_size != len)
                   1937:                return EINVAL;
1.1       cgd      1938:
1.171     plunky   1939:        memcpy(buf, sopt->sopt_data, len);
                   1940:        return 0;
                   1941: }
1.1       cgd      1942:
1.171     plunky   1943: /*
                   1944:  * common case of get sockopt integer value
                   1945:  */
                   1946: int
                   1947: sockopt_getint(const struct sockopt *sopt, int *valp)
                   1948: {
1.1       cgd      1949:
1.171     plunky   1950:        return sockopt_get(sopt, valp, sizeof(int));
                   1951: }
1.1       cgd      1952:
1.171     plunky   1953: /*
                   1954:  * set sockopt value from mbuf
                   1955:  *     - ONLY for legacy code
                   1956:  *     - mbuf is released by sockopt
                   1957:  */
                   1958: int
                   1959: sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m)
                   1960: {
                   1961:        size_t len;
1.1       cgd      1962:
1.171     plunky   1963:        len = m_length(m);
1.1       cgd      1964:
1.171     plunky   1965:        if (sopt->sopt_size == 0)
                   1966:                sockopt_alloc(sopt, len);
1.1       cgd      1967:
1.171     plunky   1968:        KASSERT(sopt->sopt_size == len);
                   1969:        m_copydata(m, 0, len, sopt->sopt_data);
                   1970:        m_freem(m);
1.1       cgd      1971:
1.171     plunky   1972:        return 0;
                   1973: }
1.1       cgd      1974:
1.171     plunky   1975: /*
                   1976:  * get sockopt value into mbuf
                   1977:  *     - ONLY for legacy code
                   1978:  *     - mbuf to be released by the caller
                   1979:  */
                   1980: struct mbuf *
                   1981: sockopt_getmbuf(const struct sockopt *sopt)
                   1982: {
                   1983:        struct mbuf *m;
1.107     darrenr  1984:
1.171     plunky   1985:        m = m_get(M_WAIT, MT_SOOPTS);
                   1986:        if (m == NULL)
                   1987:                return NULL;
                   1988:
                   1989:        m->m_len = MLEN;
                   1990:        m_copyback(m, 0, sopt->sopt_size, sopt->sopt_data);
                   1991:        if (m_length(m) != max(sopt->sopt_size, MLEN)) {
                   1992:                m_freem(m);
                   1993:                return NULL;
1.1       cgd      1994:        }
1.171     plunky   1995:        m->m_len = min(sopt->sopt_size, MLEN);
1.160     ad       1996:
1.171     plunky   1997:        return m;
1.1       cgd      1998: }
                   1999:
1.14      mycroft  2000: void
1.54      lukem    2001: sohasoutofband(struct socket *so)
1.1       cgd      2002: {
1.153     rmind    2003:
1.90      christos 2004:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.153     rmind    2005:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, 0);
1.1       cgd      2006: }
1.72      jdolecek 2007:
                   2008: static void
                   2009: filt_sordetach(struct knote *kn)
                   2010: {
                   2011:        struct socket   *so;
                   2012:
1.155     ad       2013:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2014:        solock(so);
1.73      christos 2015:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2016:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 2017:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       2018:        sounlock(so);
1.72      jdolecek 2019: }
                   2020:
                   2021: /*ARGSUSED*/
                   2022: static int
1.129     yamt     2023: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 2024: {
                   2025:        struct socket   *so;
1.160     ad       2026:        int rv;
1.72      jdolecek 2027:
1.155     ad       2028:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2029:        if (hint != NOTE_SUBMIT)
                   2030:                solock(so);
1.72      jdolecek 2031:        kn->kn_data = so->so_rcv.sb_cc;
                   2032:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    2033:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2034:                kn->kn_fflags = so->so_error;
1.160     ad       2035:                rv = 1;
                   2036:        } else if (so->so_error)        /* temporary udp error */
                   2037:                rv = 1;
                   2038:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2039:                rv = (kn->kn_data >= kn->kn_sdata);
                   2040:        else
                   2041:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   2042:        if (hint != NOTE_SUBMIT)
                   2043:                sounlock(so);
                   2044:        return rv;
1.72      jdolecek 2045: }
                   2046:
                   2047: static void
                   2048: filt_sowdetach(struct knote *kn)
                   2049: {
                   2050:        struct socket   *so;
                   2051:
1.155     ad       2052:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2053:        solock(so);
1.73      christos 2054:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2055:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 2056:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       2057:        sounlock(so);
1.72      jdolecek 2058: }
                   2059:
                   2060: /*ARGSUSED*/
                   2061: static int
1.129     yamt     2062: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 2063: {
                   2064:        struct socket   *so;
1.160     ad       2065:        int rv;
1.72      jdolecek 2066:
1.155     ad       2067:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2068:        if (hint != NOTE_SUBMIT)
                   2069:                solock(so);
1.72      jdolecek 2070:        kn->kn_data = sbspace(&so->so_snd);
                   2071:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    2072:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2073:                kn->kn_fflags = so->so_error;
1.160     ad       2074:                rv = 1;
                   2075:        } else if (so->so_error)        /* temporary udp error */
                   2076:                rv = 1;
                   2077:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 2078:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       2079:                rv = 0;
                   2080:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2081:                rv = (kn->kn_data >= kn->kn_sdata);
                   2082:        else
                   2083:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   2084:        if (hint != NOTE_SUBMIT)
                   2085:                sounlock(so);
                   2086:        return rv;
1.72      jdolecek 2087: }
                   2088:
                   2089: /*ARGSUSED*/
                   2090: static int
1.129     yamt     2091: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 2092: {
                   2093:        struct socket   *so;
1.160     ad       2094:        int rv;
1.72      jdolecek 2095:
1.155     ad       2096:        so = ((file_t *)kn->kn_obj)->f_data;
1.72      jdolecek 2097:
                   2098:        /*
                   2099:         * Set kn_data to number of incoming connections, not
                   2100:         * counting partial (incomplete) connections.
1.108     perry    2101:         */
1.160     ad       2102:        if (hint != NOTE_SUBMIT)
                   2103:                solock(so);
1.72      jdolecek 2104:        kn->kn_data = so->so_qlen;
1.160     ad       2105:        rv = (kn->kn_data > 0);
                   2106:        if (hint != NOTE_SUBMIT)
                   2107:                sounlock(so);
                   2108:        return rv;
1.72      jdolecek 2109: }
                   2110:
                   2111: static const struct filterops solisten_filtops =
                   2112:        { 1, NULL, filt_sordetach, filt_solisten };
                   2113: static const struct filterops soread_filtops =
                   2114:        { 1, NULL, filt_sordetach, filt_soread };
                   2115: static const struct filterops sowrite_filtops =
                   2116:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   2117:
                   2118: int
1.129     yamt     2119: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 2120: {
                   2121:        struct socket   *so;
                   2122:        struct sockbuf  *sb;
                   2123:
1.155     ad       2124:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2125:        solock(so);
1.72      jdolecek 2126:        switch (kn->kn_filter) {
                   2127:        case EVFILT_READ:
                   2128:                if (so->so_options & SO_ACCEPTCONN)
                   2129:                        kn->kn_fop = &solisten_filtops;
                   2130:                else
                   2131:                        kn->kn_fop = &soread_filtops;
                   2132:                sb = &so->so_rcv;
                   2133:                break;
                   2134:        case EVFILT_WRITE:
                   2135:                kn->kn_fop = &sowrite_filtops;
                   2136:                sb = &so->so_snd;
                   2137:                break;
                   2138:        default:
1.160     ad       2139:                sounlock(so);
1.149     pooka    2140:                return (EINVAL);
1.72      jdolecek 2141:        }
1.73      christos 2142:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 2143:        sb->sb_flags |= SB_KNOTE;
1.160     ad       2144:        sounlock(so);
1.72      jdolecek 2145:        return (0);
                   2146: }
                   2147:
1.154     ad       2148: static int
                   2149: sodopoll(struct socket *so, int events)
                   2150: {
                   2151:        int revents;
                   2152:
                   2153:        revents = 0;
                   2154:
                   2155:        if (events & (POLLIN | POLLRDNORM))
                   2156:                if (soreadable(so))
                   2157:                        revents |= events & (POLLIN | POLLRDNORM);
                   2158:
                   2159:        if (events & (POLLOUT | POLLWRNORM))
                   2160:                if (sowritable(so))
                   2161:                        revents |= events & (POLLOUT | POLLWRNORM);
                   2162:
                   2163:        if (events & (POLLPRI | POLLRDBAND))
                   2164:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   2165:                        revents |= events & (POLLPRI | POLLRDBAND);
                   2166:
                   2167:        return revents;
                   2168: }
                   2169:
                   2170: int
                   2171: sopoll(struct socket *so, int events)
                   2172: {
                   2173:        int revents = 0;
                   2174:
1.160     ad       2175: #ifndef DIAGNOSTIC
                   2176:        /*
                   2177:         * Do a quick, unlocked check in expectation that the socket
                   2178:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   2179:         * as the solocked() assertions will fail.
                   2180:         */
1.154     ad       2181:        if ((revents = sodopoll(so, events)) != 0)
                   2182:                return revents;
1.160     ad       2183: #endif
1.154     ad       2184:
1.160     ad       2185:        solock(so);
1.154     ad       2186:        if ((revents = sodopoll(so, events)) == 0) {
                   2187:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   2188:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       2189:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       2190:                }
                   2191:
                   2192:                if (events & (POLLOUT | POLLWRNORM)) {
                   2193:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2194:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2195:                }
                   2196:        }
1.160     ad       2197:        sounlock(so);
1.154     ad       2198:
                   2199:        return revents;
                   2200: }
                   2201:
                   2202:
1.94      yamt     2203: #include <sys/sysctl.h>
                   2204:
                   2205: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
                   2206:
                   2207: /*
                   2208:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2209:  * value is not too small.
                   2210:  * (XXX should we maybe make sure it's not too large as well?)
                   2211:  */
                   2212: static int
                   2213: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2214: {
                   2215:        int error, new_somaxkva;
                   2216:        struct sysctlnode node;
                   2217:
                   2218:        new_somaxkva = somaxkva;
                   2219:        node = *rnode;
                   2220:        node.sysctl_data = &new_somaxkva;
                   2221:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2222:        if (error || newp == NULL)
                   2223:                return (error);
                   2224:
                   2225:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2226:                return (EINVAL);
                   2227:
1.136     ad       2228:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2229:        somaxkva = new_somaxkva;
1.136     ad       2230:        cv_broadcast(&socurkva_cv);
                   2231:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2232:
                   2233:        return (error);
                   2234: }
                   2235:
                   2236: SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup")
                   2237: {
                   2238:
1.97      atatat   2239:        sysctl_createv(clog, 0, NULL, NULL,
                   2240:                       CTLFLAG_PERMANENT,
                   2241:                       CTLTYPE_NODE, "kern", NULL,
                   2242:                       NULL, 0, NULL, 0,
                   2243:                       CTL_KERN, CTL_EOL);
                   2244:
                   2245:        sysctl_createv(clog, 0, NULL, NULL,
                   2246:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2247:                       CTLTYPE_INT, "somaxkva",
                   2248:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2249:                                    "used for socket buffers"),
1.94      yamt     2250:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2251:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
                   2252: }

CVSweb <webmaster@jp.NetBSD.org>