[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 1.177.4.3

1.177.4.3! bouyer      1: /*     $NetBSD$        */
1.64      thorpej     2:
                      3: /*-
1.177.4.2  snj         4:  * Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc.
1.64      thorpej     5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
1.177.4.2  snj         8:  * by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran.
1.64      thorpej     9:  *
                     10:  * Redistribution and use in source and binary forms, with or without
                     11:  * modification, are permitted provided that the following conditions
                     12:  * are met:
                     13:  * 1. Redistributions of source code must retain the above copyright
                     14:  *    notice, this list of conditions and the following disclaimer.
                     15:  * 2. Redistributions in binary form must reproduce the above copyright
                     16:  *    notice, this list of conditions and the following disclaimer in the
                     17:  *    documentation and/or other materials provided with the distribution.
                     18:  *
                     19:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     20:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     21:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     22:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     23:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     24:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     25:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     26:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     27:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     28:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     29:  * POSSIBILITY OF SUCH DAMAGE.
                     30:  */
1.16      cgd        31:
1.1       cgd        32: /*
1.159     ad         33:  * Copyright (c) 2004 The FreeBSD Foundation
                     34:  * Copyright (c) 2004 Robert Watson
1.15      mycroft    35:  * Copyright (c) 1982, 1986, 1988, 1990, 1993
                     36:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        37:  *
                     38:  * Redistribution and use in source and binary forms, with or without
                     39:  * modification, are permitted provided that the following conditions
                     40:  * are met:
                     41:  * 1. Redistributions of source code must retain the above copyright
                     42:  *    notice, this list of conditions and the following disclaimer.
                     43:  * 2. Redistributions in binary form must reproduce the above copyright
                     44:  *    notice, this list of conditions and the following disclaimer in the
                     45:  *    documentation and/or other materials provided with the distribution.
1.85      agc        46:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        47:  *    may be used to endorse or promote products derived from this software
                     48:  *    without specific prior written permission.
                     49:  *
                     50:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     51:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     52:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     53:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     54:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     55:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     56:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     57:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     58:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     59:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     60:  * SUCH DAMAGE.
                     61:  *
1.32      fvdl       62:  *     @(#)uipc_socket.c       8.6 (Berkeley) 5/2/95
1.1       cgd        63:  */
1.59      lukem      64:
                     65: #include <sys/cdefs.h>
1.177.4.3! bouyer     66: __KERNEL_RCSID(0, "$NetBSD$");
1.64      thorpej    67:
                     68: #include "opt_sock_counters.h"
                     69: #include "opt_sosend_loan.h"
1.81      martin     70: #include "opt_mbuftrace.h"
1.84      ragge      71: #include "opt_somaxkva.h"
1.167     ad         72: #include "opt_multiprocessor.h"        /* XXX */
1.1       cgd        73:
1.9       mycroft    74: #include <sys/param.h>
                     75: #include <sys/systm.h>
                     76: #include <sys/proc.h>
                     77: #include <sys/file.h>
1.142     dyoung     78: #include <sys/filedesc.h>
1.173     plunky     79: #include <sys/kmem.h>
1.9       mycroft    80: #include <sys/mbuf.h>
                     81: #include <sys/domain.h>
                     82: #include <sys/kernel.h>
                     83: #include <sys/protosw.h>
                     84: #include <sys/socket.h>
                     85: #include <sys/socketvar.h>
1.21      christos   86: #include <sys/signalvar.h>
1.9       mycroft    87: #include <sys/resourcevar.h>
1.174     pooka      88: #include <sys/uidinfo.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 */
1.177     ad        617:        /* Remove acccept filter if one is present. */
1.170     tls       618:        if (so->so_accf != NULL)
1.177     ad        619:                (void)accept_filt_clear(so);
1.160     ad        620:        sounlock(so);
1.161     ad        621:        if (refs == 0)          /* XXX */
                    622:                soput(so);
1.1       cgd       623: }
                    624:
                    625: /*
                    626:  * Close a socket on last file table reference removal.
                    627:  * Initiate disconnect if connected.
                    628:  * Free socket when disconnect complete.
                    629:  */
1.3       andrew    630: int
1.54      lukem     631: soclose(struct socket *so)
1.1       cgd       632: {
1.54      lukem     633:        struct socket   *so2;
1.160     ad        634:        int             error;
                    635:        int             error2;
1.1       cgd       636:
1.54      lukem     637:        error = 0;
1.160     ad        638:        solock(so);
1.1       cgd       639:        if (so->so_options & SO_ACCEPTCONN) {
1.172     ad        640:                for (;;) {
                    641:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.160     ad        642:                                KASSERT(solocked2(so, so2));
                    643:                                (void) soqremque(so2, 0);
                    644:                                /* soabort drops the lock. */
                    645:                                (void) soabort(so2);
                    646:                                solock(so);
1.172     ad        647:                                continue;
1.160     ad        648:                        }
1.172     ad        649:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.160     ad        650:                                KASSERT(solocked2(so, so2));
                    651:                                (void) soqremque(so2, 1);
                    652:                                /* soabort drops the lock. */
                    653:                                (void) soabort(so2);
                    654:                                solock(so);
1.172     ad        655:                                continue;
1.160     ad        656:                        }
1.172     ad        657:                        break;
                    658:                }
1.1       cgd       659:        }
                    660:        if (so->so_pcb == 0)
                    661:                goto discard;
                    662:        if (so->so_state & SS_ISCONNECTED) {
                    663:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    664:                        error = sodisconnect(so);
                    665:                        if (error)
                    666:                                goto drop;
                    667:                }
                    668:                if (so->so_options & SO_LINGER) {
1.151     ad        669:                        if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio)
1.1       cgd       670:                                goto drop;
1.21      christos  671:                        while (so->so_state & SS_ISCONNECTED) {
1.177.4.1  snj       672:                                error = sowait(so, true, so->so_linger * hz);
1.21      christos  673:                                if (error)
1.1       cgd       674:                                        break;
1.21      christos  675:                        }
1.1       cgd       676:                }
                    677:        }
1.54      lukem     678:  drop:
1.1       cgd       679:        if (so->so_pcb) {
1.160     ad        680:                error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140     dyoung    681:                    NULL, NULL, NULL, NULL);
1.1       cgd       682:                if (error == 0)
                    683:                        error = error2;
                    684:        }
1.54      lukem     685:  discard:
1.1       cgd       686:        if (so->so_state & SS_NOFDREF)
                    687:                panic("soclose: NOFDREF");
                    688:        so->so_state |= SS_NOFDREF;
                    689:        sofree(so);
                    690:        return (error);
                    691: }
                    692:
                    693: /*
1.160     ad        694:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       695:  */
1.3       andrew    696: int
1.54      lukem     697: soabort(struct socket *so)
1.1       cgd       698: {
1.161     ad        699:        u_int refs;
1.139     yamt      700:        int error;
1.160     ad        701:
                    702:        KASSERT(solocked(so));
                    703:        KASSERT(so->so_head == NULL);
1.1       cgd       704:
1.161     ad        705:        so->so_aborting++;              /* XXX */
1.140     dyoung    706:        error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
                    707:            NULL, NULL, NULL);
1.161     ad        708:        refs = --so->so_aborting;       /* XXX */
1.164     drochner  709:        if (error || (refs == 0)) {
1.139     yamt      710:                sofree(so);
1.160     ad        711:        } else {
                    712:                sounlock(so);
1.139     yamt      713:        }
                    714:        return error;
1.1       cgd       715: }
                    716:
1.3       andrew    717: int
1.54      lukem     718: soaccept(struct socket *so, struct mbuf *nam)
1.1       cgd       719: {
1.160     ad        720:        int     error;
                    721:
                    722:        KASSERT(solocked(so));
1.1       cgd       723:
1.54      lukem     724:        error = 0;
1.1       cgd       725:        if ((so->so_state & SS_NOFDREF) == 0)
                    726:                panic("soaccept: !NOFDREF");
                    727:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   728:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    729:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41      mycroft   730:                error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140     dyoung    731:                    NULL, nam, NULL, NULL);
1.41      mycroft   732:        else
1.53      itojun    733:                error = ECONNABORTED;
1.52      itojun    734:
1.1       cgd       735:        return (error);
                    736: }
                    737:
1.3       andrew    738: int
1.114     christos  739: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       740: {
1.160     ad        741:        int             error;
                    742:
                    743:        KASSERT(solocked(so));
1.1       cgd       744:
                    745:        if (so->so_options & SO_ACCEPTCONN)
                    746:                return (EOPNOTSUPP);
                    747:        /*
                    748:         * If protocol is connection-based, can only connect once.
                    749:         * Otherwise, if connected, try to disconnect first.
                    750:         * This allows user to disconnect by connecting to, e.g.,
                    751:         * a null address.
                    752:         */
                    753:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    754:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
                    755:            (error = sodisconnect(so))))
                    756:                error = EISCONN;
                    757:        else
                    758:                error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140     dyoung    759:                    NULL, nam, NULL, l);
1.1       cgd       760:        return (error);
                    761: }
                    762:
1.3       andrew    763: int
1.54      lukem     764: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       765: {
1.160     ad        766:        int     error;
                    767:
                    768:        KASSERT(solocked2(so1, so2));
1.1       cgd       769:
1.22      mycroft   770:        error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140     dyoung    771:            NULL, (struct mbuf *)so2, NULL, NULL);
1.1       cgd       772:        return (error);
                    773: }
                    774:
1.3       andrew    775: int
1.54      lukem     776: sodisconnect(struct socket *so)
1.1       cgd       777: {
1.160     ad        778:        int     error;
                    779:
                    780:        KASSERT(solocked(so));
1.1       cgd       781:
                    782:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    783:                error = ENOTCONN;
1.160     ad        784:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       785:                error = EALREADY;
1.160     ad        786:        } else {
                    787:                error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
                    788:                    NULL, NULL, NULL, NULL);
1.1       cgd       789:        }
1.117     yamt      790:        sodopendfree();
1.1       cgd       791:        return (error);
                    792: }
                    793:
1.15      mycroft   794: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       795: /*
                    796:  * Send on a socket.
                    797:  * If send must go all at once and message is larger than
                    798:  * send buffering, then hard error.
                    799:  * Lock against other senders.
                    800:  * If must go all at once and not enough room now, then
                    801:  * inform user that this would block and do nothing.
                    802:  * Otherwise, if nonblocking, send as much as possible.
                    803:  * The data to be sent is described by "uio" if nonzero,
                    804:  * otherwise by the mbuf chain "top" (which must be null
                    805:  * if uio is not).  Data provided in mbuf chain must be small
                    806:  * enough to send all at once.
                    807:  *
                    808:  * Returns nonzero on error, timeout or signal; callers
                    809:  * must check for short counts if EINTR/ERESTART are returned.
                    810:  * Data and control buffers are freed on return.
                    811:  */
1.3       andrew    812: int
1.54      lukem     813: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114     christos  814:        struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       815: {
1.54      lukem     816:        struct mbuf     **mp, *m;
1.114     christos  817:        struct proc     *p;
1.58      jdolecek  818:        long            space, len, resid, clen, mlen;
                    819:        int             error, s, dontroute, atomic;
1.54      lukem     820:
1.114     christos  821:        p = l->l_proc;
1.117     yamt      822:        sodopendfree();
1.160     ad        823:        clen = 0;
1.64      thorpej   824:
1.160     ad        825:        /*
                    826:         * solock() provides atomicity of access.  splsoftnet() prevents
                    827:         * protocol processing soft interrupts from interrupting us and
                    828:         * blocking (expensive).
                    829:         */
                    830:        s = splsoftnet();
                    831:        solock(so);
1.54      lukem     832:        atomic = sosendallatonce(so) || top;
1.1       cgd       833:        if (uio)
                    834:                resid = uio->uio_resid;
                    835:        else
                    836:                resid = top->m_pkthdr.len;
1.7       cgd       837:        /*
                    838:         * In theory resid should be unsigned.
                    839:         * However, space must be signed, as it might be less than 0
                    840:         * if we over-committed, and we must use a signed comparison
                    841:         * of space and resid.  On the other hand, a negative resid
                    842:         * causes us to loop sending 0-length segments to the protocol.
                    843:         */
1.29      mycroft   844:        if (resid < 0) {
                    845:                error = EINVAL;
                    846:                goto out;
                    847:        }
1.1       cgd       848:        dontroute =
                    849:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    850:            (so->so_proto->pr_flags & PR_ATOMIC);
1.165     christos  851:        l->l_ru.ru_msgsnd++;
1.1       cgd       852:        if (control)
                    853:                clen = control->m_len;
1.54      lukem     854:  restart:
1.21      christos  855:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       856:                goto out;
                    857:        do {
1.160     ad        858:                if (so->so_state & SS_CANTSENDMORE) {
                    859:                        error = EPIPE;
                    860:                        goto release;
                    861:                }
1.48      thorpej   862:                if (so->so_error) {
                    863:                        error = so->so_error;
                    864:                        so->so_error = 0;
                    865:                        goto release;
                    866:                }
1.1       cgd       867:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    868:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
                    869:                                if ((so->so_state & SS_ISCONFIRMING) == 0 &&
1.160     ad        870:                                    !(resid == 0 && clen != 0)) {
                    871:                                        error = ENOTCONN;
                    872:                                        goto release;
                    873:                                }
                    874:                        } else if (addr == 0) {
                    875:                                error = EDESTADDRREQ;
                    876:                                goto release;
                    877:                        }
1.1       cgd       878:                }
                    879:                space = sbspace(&so->so_snd);
                    880:                if (flags & MSG_OOB)
                    881:                        space += 1024;
1.21      christos  882:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        883:                    clen > so->so_snd.sb_hiwat) {
                    884:                        error = EMSGSIZE;
                    885:                        goto release;
                    886:                }
1.96      mycroft   887:                if (space < resid + clen &&
1.1       cgd       888:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.160     ad        889:                        if (so->so_nbio) {
                    890:                                error = EWOULDBLOCK;
                    891:                                goto release;
                    892:                        }
1.1       cgd       893:                        sbunlock(&so->so_snd);
                    894:                        error = sbwait(&so->so_snd);
                    895:                        if (error)
                    896:                                goto out;
                    897:                        goto restart;
                    898:                }
                    899:                mp = &top;
                    900:                space -= clen;
                    901:                do {
1.45      tv        902:                        if (uio == NULL) {
                    903:                                /*
                    904:                                 * Data is prepackaged in "top".
                    905:                                 */
                    906:                                resid = 0;
                    907:                                if (flags & MSG_EOR)
                    908:                                        top->m_flags |= M_EOR;
                    909:                        } else do {
1.160     ad        910:                                sounlock(so);
                    911:                                splx(s);
1.144     dyoung    912:                                if (top == NULL) {
1.78      matt      913:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv        914:                                        mlen = MHLEN;
                    915:                                        m->m_pkthdr.len = 0;
1.140     dyoung    916:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv        917:                                } else {
1.78      matt      918:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv        919:                                        mlen = MLEN;
                    920:                                }
1.78      matt      921:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt      922:                                if (sock_loan_thresh >= 0 &&
                    923:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                    924:                                    space >= sock_loan_thresh &&
1.64      thorpej   925:                                    (len = sosend_loan(so, uio, m,
                    926:                                                       space)) != 0) {
                    927:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                    928:                                        space -= len;
                    929:                                        goto have_data;
                    930:                                }
1.45      tv        931:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej   932:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78      matt      933:                                        m_clget(m, M_WAIT);
1.45      tv        934:                                        if ((m->m_flags & M_EXT) == 0)
                    935:                                                goto nopages;
                    936:                                        mlen = MCLBYTES;
                    937:                                        if (atomic && top == 0) {
1.58      jdolecek  938:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem     939:                                                    resid);
1.45      tv        940:                                                m->m_data += max_hdr;
                    941:                                        } else
1.58      jdolecek  942:                                                len = lmin(MCLBYTES, resid);
1.45      tv        943:                                        space -= len;
                    944:                                } else {
1.64      thorpej   945:  nopages:
                    946:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek  947:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv        948:                                        space -= len;
                    949:                                        /*
                    950:                                         * For datagram protocols, leave room
                    951:                                         * for protocol headers in first mbuf.
                    952:                                         */
                    953:                                        if (atomic && top == 0 && len < mlen)
                    954:                                                MH_ALIGN(m, len);
                    955:                                }
1.144     dyoung    956:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej   957:  have_data:
1.45      tv        958:                                resid = uio->uio_resid;
                    959:                                m->m_len = len;
                    960:                                *mp = m;
                    961:                                top->m_pkthdr.len += len;
1.160     ad        962:                                s = splsoftnet();
                    963:                                solock(so);
1.144     dyoung    964:                                if (error != 0)
1.45      tv        965:                                        goto release;
                    966:                                mp = &m->m_next;
                    967:                                if (resid <= 0) {
                    968:                                        if (flags & MSG_EOR)
                    969:                                                top->m_flags |= M_EOR;
                    970:                                        break;
                    971:                                }
                    972:                        } while (space > 0 && atomic);
1.108     perry     973:
1.160     ad        974:                        if (so->so_state & SS_CANTSENDMORE) {
                    975:                                error = EPIPE;
                    976:                                goto release;
                    977:                        }
1.45      tv        978:                        if (dontroute)
                    979:                                so->so_options |= SO_DONTROUTE;
                    980:                        if (resid > 0)
                    981:                                so->so_state |= SS_MORETOCOME;
1.46      sommerfe  982:                        error = (*so->so_proto->pr_usrreq)(so,
                    983:                            (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.160     ad        984:                            top, addr, control, curlwp);
1.45      tv        985:                        if (dontroute)
                    986:                                so->so_options &= ~SO_DONTROUTE;
                    987:                        if (resid > 0)
                    988:                                so->so_state &= ~SS_MORETOCOME;
                    989:                        clen = 0;
1.144     dyoung    990:                        control = NULL;
                    991:                        top = NULL;
1.45      tv        992:                        mp = &top;
1.144     dyoung    993:                        if (error != 0)
1.1       cgd       994:                                goto release;
                    995:                } while (resid && space > 0);
                    996:        } while (resid);
                    997:
1.54      lukem     998:  release:
1.1       cgd       999:        sbunlock(&so->so_snd);
1.54      lukem    1000:  out:
1.160     ad       1001:        sounlock(so);
                   1002:        splx(s);
1.1       cgd      1003:        if (top)
                   1004:                m_freem(top);
                   1005:        if (control)
                   1006:                m_freem(control);
                   1007:        return (error);
                   1008: }
                   1009:
                   1010: /*
1.159     ad       1011:  * Following replacement or removal of the first mbuf on the first
                   1012:  * mbuf chain of a socket buffer, push necessary state changes back
                   1013:  * into the socket buffer so that other consumers see the values
                   1014:  * consistently.  'nextrecord' is the callers locally stored value of
                   1015:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1016:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1017:  */
                   1018: static void
                   1019: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1020: {
                   1021:
1.160     ad       1022:        KASSERT(solocked(sb->sb_so));
                   1023:
1.159     ad       1024:        /*
                   1025:         * First, update for the new value of nextrecord.  If necessary,
                   1026:         * make it the first record.
                   1027:         */
                   1028:        if (sb->sb_mb != NULL)
                   1029:                sb->sb_mb->m_nextpkt = nextrecord;
                   1030:        else
                   1031:                sb->sb_mb = nextrecord;
                   1032:
                   1033:         /*
                   1034:          * Now update any dependent socket buffer fields to reflect
                   1035:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1036:          * the addition of a second clause that takes care of the
                   1037:          * case where sb_mb has been updated, but remains the last
                   1038:          * record.
                   1039:          */
                   1040:         if (sb->sb_mb == NULL) {
                   1041:                 sb->sb_mbtail = NULL;
                   1042:                 sb->sb_lastrecord = NULL;
                   1043:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1044:                 sb->sb_lastrecord = sb->sb_mb;
                   1045: }
                   1046:
                   1047: /*
1.1       cgd      1048:  * Implement receive operations on a socket.
                   1049:  * We depend on the way that records are added to the sockbuf
                   1050:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1051:  * must begin with an address if the protocol so specifies,
                   1052:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1053:  * and then zero or more mbufs of data.
                   1054:  * In order to avoid blocking network interrupts for the entire time here,
                   1055:  * we splx() while doing the actual copy to user space.
                   1056:  * Although the sockbuf is locked, new data may still be appended,
                   1057:  * and thus we must maintain consistency of the sockbuf during that time.
                   1058:  *
                   1059:  * The caller may receive the data as a single mbuf chain by supplying
                   1060:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1061:  * only for the count in uio_resid.
                   1062:  */
1.3       andrew   1063: int
1.54      lukem    1064: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1065:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1066: {
1.116     yamt     1067:        struct lwp *l = curlwp;
1.160     ad       1068:        struct mbuf     *m, **mp, *mt;
1.146     dyoung   1069:        int atomic, flags, len, error, s, offset, moff, type, orig_resid;
1.99      matt     1070:        const struct protosw    *pr;
1.54      lukem    1071:        struct mbuf     *nextrecord;
1.67      he       1072:        int             mbuf_removed = 0;
1.146     dyoung   1073:        const struct domain *dom;
1.64      thorpej  1074:
1.54      lukem    1075:        pr = so->so_proto;
1.146     dyoung   1076:        atomic = pr->pr_flags & PR_ATOMIC;
                   1077:        dom = pr->pr_domain;
1.1       cgd      1078:        mp = mp0;
1.54      lukem    1079:        type = 0;
                   1080:        orig_resid = uio->uio_resid;
1.102     jonathan 1081:
1.144     dyoung   1082:        if (paddr != NULL)
                   1083:                *paddr = NULL;
                   1084:        if (controlp != NULL)
                   1085:                *controlp = NULL;
                   1086:        if (flagsp != NULL)
1.1       cgd      1087:                flags = *flagsp &~ MSG_EOR;
                   1088:        else
                   1089:                flags = 0;
1.66      enami    1090:
                   1091:        if ((flags & MSG_DONTWAIT) == 0)
1.117     yamt     1092:                sodopendfree();
1.66      enami    1093:
1.1       cgd      1094:        if (flags & MSG_OOB) {
                   1095:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1096:                solock(so);
1.17      cgd      1097:                error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140     dyoung   1098:                    (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.160     ad       1099:                sounlock(so);
1.1       cgd      1100:                if (error)
                   1101:                        goto bad;
                   1102:                do {
1.134     christos 1103:                        error = uiomove(mtod(m, void *),
1.1       cgd      1104:                            (int) min(uio->uio_resid, m->m_len), uio);
                   1105:                        m = m_free(m);
1.144     dyoung   1106:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1107:  bad:
1.144     dyoung   1108:                if (m != NULL)
1.1       cgd      1109:                        m_freem(m);
1.144     dyoung   1110:                return error;
1.1       cgd      1111:        }
1.144     dyoung   1112:        if (mp != NULL)
1.140     dyoung   1113:                *mp = NULL;
1.160     ad       1114:
                   1115:        /*
                   1116:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1117:         * protocol processing soft interrupts from interrupting us and
                   1118:         * blocking (expensive).
                   1119:         */
                   1120:        s = splsoftnet();
                   1121:        solock(so);
1.1       cgd      1122:        if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140     dyoung   1123:                (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1       cgd      1124:
1.54      lukem    1125:  restart:
1.160     ad       1126:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1127:                sounlock(so);
                   1128:                splx(s);
1.144     dyoung   1129:                return error;
1.160     ad       1130:        }
1.1       cgd      1131:
                   1132:        m = so->so_rcv.sb_mb;
                   1133:        /*
                   1134:         * If we have less data than requested, block awaiting more
                   1135:         * (subject to any timeout) if:
1.15      mycroft  1136:         *   1. the current count is less than the low water mark,
1.1       cgd      1137:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1138:         *      receive operation at once if we block (resid <= hiwat), or
                   1139:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1140:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1141:         * we have to do the receive in sections, and thus risk returning
                   1142:         * a short count if a timeout or signal occurs after we start.
                   1143:         */
1.144     dyoung   1144:        if (m == NULL ||
                   1145:            ((flags & MSG_DONTWAIT) == 0 &&
                   1146:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1147:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1148:              ((flags & MSG_WAITALL) &&
                   1149:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1150:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1151: #ifdef DIAGNOSTIC
1.144     dyoung   1152:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1153:                        panic("receive 1");
                   1154: #endif
                   1155:                if (so->so_error) {
1.144     dyoung   1156:                        if (m != NULL)
1.15      mycroft  1157:                                goto dontblock;
1.1       cgd      1158:                        error = so->so_error;
                   1159:                        if ((flags & MSG_PEEK) == 0)
                   1160:                                so->so_error = 0;
                   1161:                        goto release;
                   1162:                }
                   1163:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1164:                        if (m != NULL)
1.15      mycroft  1165:                                goto dontblock;
1.1       cgd      1166:                        else
                   1167:                                goto release;
                   1168:                }
1.144     dyoung   1169:                for (; m != NULL; m = m->m_next)
1.1       cgd      1170:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1171:                                m = so->so_rcv.sb_mb;
                   1172:                                goto dontblock;
                   1173:                        }
                   1174:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1175:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1176:                        error = ENOTCONN;
                   1177:                        goto release;
                   1178:                }
                   1179:                if (uio->uio_resid == 0)
                   1180:                        goto release;
1.151     ad       1181:                if (so->so_nbio || (flags & MSG_DONTWAIT)) {
1.1       cgd      1182:                        error = EWOULDBLOCK;
                   1183:                        goto release;
                   1184:                }
1.69      thorpej  1185:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1186:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1187:                sbunlock(&so->so_rcv);
                   1188:                error = sbwait(&so->so_rcv);
1.160     ad       1189:                if (error != 0) {
                   1190:                        sounlock(so);
                   1191:                        splx(s);
1.144     dyoung   1192:                        return error;
1.160     ad       1193:                }
1.1       cgd      1194:                goto restart;
                   1195:        }
1.54      lukem    1196:  dontblock:
1.69      thorpej  1197:        /*
                   1198:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1199:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1200:         * pointer to the next record in the socket buffer.  We must keep the
                   1201:         * various socket buffer pointers and local stack versions of the
                   1202:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1203:         * socket lock, and re-reading them when picking it up.
1.159     ad       1204:         *
                   1205:         * Otherwise, we will race with the network stack appending new data
                   1206:         * or records onto the socket buffer by using inconsistent/stale
                   1207:         * versions of the field, possibly resulting in socket buffer
                   1208:         * corruption.
                   1209:         *
                   1210:         * By holding the high-level sblock(), we prevent simultaneous
                   1211:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1212:         */
1.144     dyoung   1213:        if (l != NULL)
1.157     ad       1214:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1215:        KASSERT(m == so->so_rcv.sb_mb);
                   1216:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1217:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1218:        nextrecord = m->m_nextpkt;
                   1219:        if (pr->pr_flags & PR_ADDR) {
                   1220: #ifdef DIAGNOSTIC
                   1221:                if (m->m_type != MT_SONAME)
                   1222:                        panic("receive 1a");
                   1223: #endif
1.3       andrew   1224:                orig_resid = 0;
1.1       cgd      1225:                if (flags & MSG_PEEK) {
                   1226:                        if (paddr)
                   1227:                                *paddr = m_copy(m, 0, m->m_len);
                   1228:                        m = m->m_next;
                   1229:                } else {
                   1230:                        sbfree(&so->so_rcv, m);
1.67      he       1231:                        mbuf_removed = 1;
1.144     dyoung   1232:                        if (paddr != NULL) {
1.1       cgd      1233:                                *paddr = m;
                   1234:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1235:                                m->m_next = NULL;
1.1       cgd      1236:                                m = so->so_rcv.sb_mb;
                   1237:                        } else {
                   1238:                                MFREE(m, so->so_rcv.sb_mb);
                   1239:                                m = so->so_rcv.sb_mb;
                   1240:                        }
1.159     ad       1241:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1242:                }
                   1243:        }
1.159     ad       1244:
                   1245:        /*
                   1246:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1247:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1248:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1249:         * perform externalization (or freeing if controlp == NULL).
                   1250:         */
                   1251:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1252:                struct mbuf *cm = NULL, *cmn;
                   1253:                struct mbuf **cme = &cm;
                   1254:
                   1255:                do {
                   1256:                        if (flags & MSG_PEEK) {
                   1257:                                if (controlp != NULL) {
                   1258:                                        *controlp = m_copy(m, 0, m->m_len);
                   1259:                                        controlp = &(*controlp)->m_next;
                   1260:                                }
                   1261:                                m = m->m_next;
                   1262:                        } else {
                   1263:                                sbfree(&so->so_rcv, m);
1.1       cgd      1264:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1265:                                m->m_next = NULL;
1.159     ad       1266:                                *cme = m;
                   1267:                                cme = &(*cme)->m_next;
1.1       cgd      1268:                                m = so->so_rcv.sb_mb;
1.159     ad       1269:                        }
                   1270:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1271:                if ((flags & MSG_PEEK) == 0)
                   1272:                        sbsync(&so->so_rcv, nextrecord);
                   1273:                for (; cm != NULL; cm = cmn) {
                   1274:                        cmn = cm->m_next;
                   1275:                        cm->m_next = NULL;
                   1276:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1277:                        if (controlp != NULL) {
                   1278:                                if (dom->dom_externalize != NULL &&
                   1279:                                    type == SCM_RIGHTS) {
1.160     ad       1280:                                        sounlock(so);
1.159     ad       1281:                                        splx(s);
                   1282:                                        error = (*dom->dom_externalize)(cm, l);
                   1283:                                        s = splsoftnet();
1.160     ad       1284:                                        solock(so);
1.159     ad       1285:                                }
                   1286:                                *controlp = cm;
                   1287:                                while (*controlp != NULL)
                   1288:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1289:                        } else {
1.106     itojun   1290:                                /*
                   1291:                                 * Dispose of any SCM_RIGHTS message that went
                   1292:                                 * through the read path rather than recv.
                   1293:                                 */
1.159     ad       1294:                                if (dom->dom_dispose != NULL &&
                   1295:                                    type == SCM_RIGHTS) {
1.160     ad       1296:                                        sounlock(so);
1.159     ad       1297:                                        (*dom->dom_dispose)(cm);
1.160     ad       1298:                                        solock(so);
1.159     ad       1299:                                }
                   1300:                                m_freem(cm);
1.1       cgd      1301:                        }
                   1302:                }
1.159     ad       1303:                if (m != NULL)
                   1304:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1305:                else
                   1306:                        nextrecord = so->so_rcv.sb_mb;
                   1307:                orig_resid = 0;
1.1       cgd      1308:        }
1.69      thorpej  1309:
1.159     ad       1310:        /* If m is non-NULL, we have some data to read. */
                   1311:        if (__predict_true(m != NULL)) {
1.1       cgd      1312:                type = m->m_type;
                   1313:                if (type == MT_OOBDATA)
                   1314:                        flags |= MSG_OOB;
                   1315:        }
1.69      thorpej  1316:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1317:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1318:
1.1       cgd      1319:        moff = 0;
                   1320:        offset = 0;
1.144     dyoung   1321:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1322:                if (m->m_type == MT_OOBDATA) {
                   1323:                        if (type != MT_OOBDATA)
                   1324:                                break;
                   1325:                } else if (type == MT_OOBDATA)
                   1326:                        break;
                   1327: #ifdef DIAGNOSTIC
                   1328:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1329:                        panic("receive 3");
                   1330: #endif
                   1331:                so->so_state &= ~SS_RCVATMARK;
                   1332:                len = uio->uio_resid;
                   1333:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1334:                        len = so->so_oobmark - offset;
                   1335:                if (len > m->m_len - moff)
                   1336:                        len = m->m_len - moff;
                   1337:                /*
                   1338:                 * If mp is set, just pass back the mbufs.
                   1339:                 * Otherwise copy them out via the uio, then free.
                   1340:                 * Sockbuf must be consistent here (points to current mbuf,
                   1341:                 * it points to next record) when we drop priority;
                   1342:                 * we must note any additions to the sockbuf when we
                   1343:                 * block interrupts again.
                   1344:                 */
1.144     dyoung   1345:                if (mp == NULL) {
1.69      thorpej  1346:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1347:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1348:                        sounlock(so);
1.1       cgd      1349:                        splx(s);
1.134     christos 1350:                        error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20      mycroft  1351:                        s = splsoftnet();
1.160     ad       1352:                        solock(so);
1.144     dyoung   1353:                        if (error != 0) {
1.67      he       1354:                                /*
                   1355:                                 * If any part of the record has been removed
                   1356:                                 * (such as the MT_SONAME mbuf, which will
                   1357:                                 * happen when PR_ADDR, and thus also
                   1358:                                 * PR_ATOMIC, is set), then drop the entire
                   1359:                                 * record to maintain the atomicity of the
                   1360:                                 * receive operation.
                   1361:                                 *
                   1362:                                 * This avoids a later panic("receive 1a")
                   1363:                                 * when compiled with DIAGNOSTIC.
                   1364:                                 */
1.146     dyoung   1365:                                if (m && mbuf_removed && atomic)
1.67      he       1366:                                        (void) sbdroprecord(&so->so_rcv);
                   1367:
1.57      jdolecek 1368:                                goto release;
1.67      he       1369:                        }
1.1       cgd      1370:                } else
                   1371:                        uio->uio_resid -= len;
                   1372:                if (len == m->m_len - moff) {
                   1373:                        if (m->m_flags & M_EOR)
                   1374:                                flags |= MSG_EOR;
                   1375:                        if (flags & MSG_PEEK) {
                   1376:                                m = m->m_next;
                   1377:                                moff = 0;
                   1378:                        } else {
                   1379:                                nextrecord = m->m_nextpkt;
                   1380:                                sbfree(&so->so_rcv, m);
                   1381:                                if (mp) {
                   1382:                                        *mp = m;
                   1383:                                        mp = &m->m_next;
                   1384:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1385:                                        *mp = NULL;
1.1       cgd      1386:                                } else {
                   1387:                                        MFREE(m, so->so_rcv.sb_mb);
                   1388:                                        m = so->so_rcv.sb_mb;
                   1389:                                }
1.69      thorpej  1390:                                /*
                   1391:                                 * If m != NULL, we also know that
                   1392:                                 * so->so_rcv.sb_mb != NULL.
                   1393:                                 */
                   1394:                                KASSERT(so->so_rcv.sb_mb == m);
                   1395:                                if (m) {
1.1       cgd      1396:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1397:                                        if (nextrecord == NULL)
                   1398:                                                so->so_rcv.sb_lastrecord = m;
                   1399:                                } else {
                   1400:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1401:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1402:                                }
                   1403:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1404:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1405:                        }
1.144     dyoung   1406:                } else if (flags & MSG_PEEK)
                   1407:                        moff += len;
                   1408:                else {
1.160     ad       1409:                        if (mp != NULL) {
                   1410:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1411:                                if (__predict_false(mt == NULL)) {
                   1412:                                        sounlock(so);
                   1413:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1414:                                        solock(so);
                   1415:                                }
                   1416:                                *mp = mt;
                   1417:                        }
1.144     dyoung   1418:                        m->m_data += len;
                   1419:                        m->m_len -= len;
                   1420:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1421:                }
                   1422:                if (so->so_oobmark) {
                   1423:                        if ((flags & MSG_PEEK) == 0) {
                   1424:                                so->so_oobmark -= len;
                   1425:                                if (so->so_oobmark == 0) {
                   1426:                                        so->so_state |= SS_RCVATMARK;
                   1427:                                        break;
                   1428:                                }
1.7       cgd      1429:                        } else {
1.1       cgd      1430:                                offset += len;
1.7       cgd      1431:                                if (offset == so->so_oobmark)
                   1432:                                        break;
                   1433:                        }
1.1       cgd      1434:                }
                   1435:                if (flags & MSG_EOR)
                   1436:                        break;
                   1437:                /*
                   1438:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1439:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1440:                 * termination.  If a signal/timeout occurs, return
                   1441:                 * with a short count but without error.
                   1442:                 * Keep sockbuf locked against other readers.
                   1443:                 */
1.144     dyoung   1444:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1445:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1446:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1447:                                break;
1.68      matt     1448:                        /*
                   1449:                         * If we are peeking and the socket receive buffer is
                   1450:                         * full, stop since we can't get more data to peek at.
                   1451:                         */
                   1452:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1453:                                break;
                   1454:                        /*
                   1455:                         * If we've drained the socket buffer, tell the
                   1456:                         * protocol in case it needs to do something to
                   1457:                         * get it filled again.
                   1458:                         */
                   1459:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
                   1460:                                (*pr->pr_usrreq)(so, PRU_RCVD,
1.140     dyoung   1461:                                    NULL, (struct mbuf *)(long)flags, NULL, l);
1.69      thorpej  1462:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1463:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1       cgd      1464:                        error = sbwait(&so->so_rcv);
1.144     dyoung   1465:                        if (error != 0) {
1.1       cgd      1466:                                sbunlock(&so->so_rcv);
1.160     ad       1467:                                sounlock(so);
1.1       cgd      1468:                                splx(s);
1.144     dyoung   1469:                                return 0;
1.1       cgd      1470:                        }
1.21      christos 1471:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1472:                                nextrecord = m->m_nextpkt;
                   1473:                }
                   1474:        }
1.3       andrew   1475:
1.146     dyoung   1476:        if (m && atomic) {
1.3       andrew   1477:                flags |= MSG_TRUNC;
                   1478:                if ((flags & MSG_PEEK) == 0)
                   1479:                        (void) sbdroprecord(&so->so_rcv);
                   1480:        }
1.1       cgd      1481:        if ((flags & MSG_PEEK) == 0) {
1.144     dyoung   1482:                if (m == NULL) {
1.69      thorpej  1483:                        /*
1.70      thorpej  1484:                         * First part is an inline SB_EMPTY_FIXUP().  Second
1.69      thorpej  1485:                         * part makes sure sb_lastrecord is up-to-date if
                   1486:                         * there is still data in the socket buffer.
                   1487:                         */
1.1       cgd      1488:                        so->so_rcv.sb_mb = nextrecord;
1.69      thorpej  1489:                        if (so->so_rcv.sb_mb == NULL) {
                   1490:                                so->so_rcv.sb_mbtail = NULL;
                   1491:                                so->so_rcv.sb_lastrecord = NULL;
                   1492:                        } else if (nextrecord->m_nextpkt == NULL)
                   1493:                                so->so_rcv.sb_lastrecord = nextrecord;
                   1494:                }
                   1495:                SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
                   1496:                SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1       cgd      1497:                if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140     dyoung   1498:                        (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
                   1499:                            (struct mbuf *)(long)flags, NULL, l);
1.1       cgd      1500:        }
1.3       andrew   1501:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1502:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1503:                sbunlock(&so->so_rcv);
                   1504:                goto restart;
                   1505:        }
1.108     perry    1506:
1.144     dyoung   1507:        if (flagsp != NULL)
1.1       cgd      1508:                *flagsp |= flags;
1.54      lukem    1509:  release:
1.1       cgd      1510:        sbunlock(&so->so_rcv);
1.160     ad       1511:        sounlock(so);
1.1       cgd      1512:        splx(s);
1.144     dyoung   1513:        return error;
1.1       cgd      1514: }
                   1515:
1.14      mycroft  1516: int
1.54      lukem    1517: soshutdown(struct socket *so, int how)
1.1       cgd      1518: {
1.99      matt     1519:        const struct protosw    *pr;
1.160     ad       1520:        int     error;
                   1521:
                   1522:        KASSERT(solocked(so));
1.34      kleink   1523:
1.54      lukem    1524:        pr = so->so_proto;
1.34      kleink   1525:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1526:                return (EINVAL);
1.1       cgd      1527:
1.160     ad       1528:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1529:                sorflush(so);
1.160     ad       1530:                error = 0;
                   1531:        }
1.34      kleink   1532:        if (how == SHUT_WR || how == SHUT_RDWR)
1.160     ad       1533:                error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1.140     dyoung   1534:                    NULL, NULL, NULL);
1.160     ad       1535:
                   1536:        return error;
1.1       cgd      1537: }
                   1538:
1.177.4.2  snj      1539: int
                   1540: sodrain(struct socket *so)
                   1541: {
                   1542:        int error;
                   1543:
                   1544:        solock(so);
                   1545:        so->so_state |= SS_ISDRAINING;
                   1546:        cv_broadcast(&so->so_cv);
                   1547:        error = soshutdown(so, SHUT_RDWR);
                   1548:        sounlock(so);
                   1549:
                   1550:        return error;
                   1551: }
                   1552:
1.14      mycroft  1553: void
1.54      lukem    1554: sorflush(struct socket *so)
1.1       cgd      1555: {
1.54      lukem    1556:        struct sockbuf  *sb, asb;
1.99      matt     1557:        const struct protosw    *pr;
1.160     ad       1558:
                   1559:        KASSERT(solocked(so));
1.1       cgd      1560:
1.54      lukem    1561:        sb = &so->so_rcv;
                   1562:        pr = so->so_proto;
1.160     ad       1563:        socantrcvmore(so);
1.1       cgd      1564:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1565:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1566:        sbunlock(sb);
                   1567:        asb = *sb;
1.86      wrstuden 1568:        /*
                   1569:         * Clear most of the sockbuf structure, but leave some of the
                   1570:         * fields valid.
                   1571:         */
                   1572:        memset(&sb->sb_startzero, 0,
                   1573:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1574:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1575:                sounlock(so);
1.1       cgd      1576:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1577:                solock(so);
                   1578:        }
1.98      christos 1579:        sbrelease(&asb, so);
1.1       cgd      1580: }
                   1581:
1.171     plunky   1582: /*
                   1583:  * internal set SOL_SOCKET options
                   1584:  */
1.142     dyoung   1585: static int
1.171     plunky   1586: sosetopt1(struct socket *so, const struct sockopt *sopt)
1.1       cgd      1587: {
1.171     plunky   1588:        int error, optval;
                   1589:        struct linger l;
                   1590:        struct timeval tv;
1.142     dyoung   1591:
1.171     plunky   1592:        switch (sopt->sopt_name) {
1.142     dyoung   1593:
1.170     tls      1594:        case SO_ACCEPTFILTER:
1.177     ad       1595:                error = accept_filt_setopt(so, sopt);
                   1596:                KASSERT(solocked(so));
1.170     tls      1597:                break;
                   1598:
1.171     plunky   1599:        case SO_LINGER:
                   1600:                error = sockopt_get(sopt, &l, sizeof(l));
1.177     ad       1601:                solock(so);
1.171     plunky   1602:                if (error)
1.177     ad       1603:                        break;
1.171     plunky   1604:                if (l.l_linger < 0 || l.l_linger > USHRT_MAX ||
1.177     ad       1605:                    l.l_linger > (INT_MAX / hz)) {
                   1606:                        error = EDOM;
                   1607:                        break;
                   1608:                }
1.171     plunky   1609:                so->so_linger = l.l_linger;
                   1610:                if (l.l_onoff)
                   1611:                        so->so_options |= SO_LINGER;
                   1612:                else
                   1613:                        so->so_options &= ~SO_LINGER;
1.177     ad       1614:                break;
1.1       cgd      1615:
1.142     dyoung   1616:        case SO_DEBUG:
                   1617:        case SO_KEEPALIVE:
                   1618:        case SO_DONTROUTE:
                   1619:        case SO_USELOOPBACK:
                   1620:        case SO_BROADCAST:
                   1621:        case SO_REUSEADDR:
                   1622:        case SO_REUSEPORT:
                   1623:        case SO_OOBINLINE:
                   1624:        case SO_TIMESTAMP:
1.171     plunky   1625:                error = sockopt_getint(sopt, &optval);
1.177     ad       1626:                solock(so);
1.171     plunky   1627:                if (error)
1.177     ad       1628:                        break;
1.171     plunky   1629:                if (optval)
                   1630:                        so->so_options |= sopt->sopt_name;
1.142     dyoung   1631:                else
1.171     plunky   1632:                        so->so_options &= ~sopt->sopt_name;
1.142     dyoung   1633:                break;
                   1634:
                   1635:        case SO_SNDBUF:
                   1636:        case SO_RCVBUF:
                   1637:        case SO_SNDLOWAT:
                   1638:        case SO_RCVLOWAT:
1.171     plunky   1639:                error = sockopt_getint(sopt, &optval);
1.177     ad       1640:                solock(so);
1.171     plunky   1641:                if (error)
1.177     ad       1642:                        break;
1.1       cgd      1643:
1.142     dyoung   1644:                /*
                   1645:                 * Values < 1 make no sense for any of these
                   1646:                 * options, so disallow them.
                   1647:                 */
1.177     ad       1648:                if (optval < 1) {
                   1649:                        error = EINVAL;
                   1650:                        break;
                   1651:                }
1.1       cgd      1652:
1.171     plunky   1653:                switch (sopt->sopt_name) {
                   1654:                case SO_SNDBUF:
1.177     ad       1655:                        if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) {
                   1656:                                error = ENOBUFS;
                   1657:                                break;
                   1658:                        }
1.171     plunky   1659:                        so->so_snd.sb_flags &= ~SB_AUTOSIZE;
                   1660:                        break;
1.1       cgd      1661:
                   1662:                case SO_RCVBUF:
1.177     ad       1663:                        if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) {
                   1664:                                error = ENOBUFS;
                   1665:                                break;
                   1666:                        }
1.171     plunky   1667:                        so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1.142     dyoung   1668:                        break;
                   1669:
                   1670:                /*
                   1671:                 * Make sure the low-water is never greater than
                   1672:                 * the high-water.
                   1673:                 */
1.1       cgd      1674:                case SO_SNDLOWAT:
1.171     plunky   1675:                        if (optval > so->so_snd.sb_hiwat)
                   1676:                                optval = so->so_snd.sb_hiwat;
                   1677:
                   1678:                        so->so_snd.sb_lowat = optval;
1.142     dyoung   1679:                        break;
1.171     plunky   1680:
1.1       cgd      1681:                case SO_RCVLOWAT:
1.171     plunky   1682:                        if (optval > so->so_rcv.sb_hiwat)
                   1683:                                optval = so->so_rcv.sb_hiwat;
                   1684:
                   1685:                        so->so_rcv.sb_lowat = optval;
1.142     dyoung   1686:                        break;
                   1687:                }
                   1688:                break;
1.28      thorpej  1689:
1.142     dyoung   1690:        case SO_SNDTIMEO:
                   1691:        case SO_RCVTIMEO:
1.171     plunky   1692:                error = sockopt_get(sopt, &tv, sizeof(tv));
1.177     ad       1693:                solock(so);
1.171     plunky   1694:                if (error)
1.177     ad       1695:                        break;
1.171     plunky   1696:
1.177     ad       1697:                if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) {
                   1698:                        error = EDOM;
                   1699:                        break;
                   1700:                }
1.28      thorpej  1701:
1.171     plunky   1702:                optval = tv.tv_sec * hz + tv.tv_usec / tick;
                   1703:                if (optval == 0 && tv.tv_usec != 0)
                   1704:                        optval = 1;
1.28      thorpej  1705:
1.171     plunky   1706:                switch (sopt->sopt_name) {
1.142     dyoung   1707:                case SO_SNDTIMEO:
1.171     plunky   1708:                        so->so_snd.sb_timeo = optval;
1.1       cgd      1709:                        break;
                   1710:                case SO_RCVTIMEO:
1.171     plunky   1711:                        so->so_rcv.sb_timeo = optval;
1.142     dyoung   1712:                        break;
                   1713:                }
                   1714:                break;
1.1       cgd      1715:
1.142     dyoung   1716:        default:
1.177     ad       1717:                solock(so);
                   1718:                error = ENOPROTOOPT;
                   1719:                break;
1.142     dyoung   1720:        }
1.177     ad       1721:        KASSERT(solocked(so));
                   1722:        return error;
1.142     dyoung   1723: }
1.1       cgd      1724:
1.142     dyoung   1725: int
1.171     plunky   1726: sosetopt(struct socket *so, struct sockopt *sopt)
1.142     dyoung   1727: {
                   1728:        int error, prerr;
1.1       cgd      1729:
1.177     ad       1730:        if (sopt->sopt_level == SOL_SOCKET) {
1.171     plunky   1731:                error = sosetopt1(so, sopt);
1.177     ad       1732:                KASSERT(solocked(so));
                   1733:        } else {
1.142     dyoung   1734:                error = ENOPROTOOPT;
1.177     ad       1735:                solock(so);
                   1736:        }
1.1       cgd      1737:
1.142     dyoung   1738:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1739:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1740:                /* give the protocol stack a shot */
1.171     plunky   1741:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt);
1.142     dyoung   1742:                if (prerr == 0)
                   1743:                        error = 0;
                   1744:                else if (prerr != ENOPROTOOPT)
                   1745:                        error = prerr;
1.171     plunky   1746:        }
1.160     ad       1747:        sounlock(so);
1.142     dyoung   1748:        return error;
1.1       cgd      1749: }
                   1750:
1.171     plunky   1751: /*
                   1752:  * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt()
                   1753:  */
                   1754: int
                   1755: so_setsockopt(struct lwp *l, struct socket *so, int level, int name,
                   1756:     const void *val, size_t valsize)
                   1757: {
                   1758:        struct sockopt sopt;
                   1759:        int error;
                   1760:
                   1761:        KASSERT(valsize == 0 || val != NULL);
                   1762:
                   1763:        sockopt_init(&sopt, level, name, valsize);
                   1764:        sockopt_set(&sopt, val, valsize);
                   1765:
                   1766:        error = sosetopt(so, &sopt);
                   1767:
                   1768:        sockopt_destroy(&sopt);
                   1769:
                   1770:        return error;
                   1771: }
                   1772:
                   1773: /*
                   1774:  * internal get SOL_SOCKET options
                   1775:  */
                   1776: static int
                   1777: sogetopt1(struct socket *so, struct sockopt *sopt)
                   1778: {
                   1779:        int error, optval;
                   1780:        struct linger l;
                   1781:        struct timeval tv;
                   1782:
                   1783:        switch (sopt->sopt_name) {
                   1784:
                   1785:        case SO_ACCEPTFILTER:
1.177     ad       1786:                error = accept_filt_getopt(so, sopt);
1.171     plunky   1787:                break;
                   1788:
                   1789:        case SO_LINGER:
                   1790:                l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0;
                   1791:                l.l_linger = so->so_linger;
                   1792:
                   1793:                error = sockopt_set(sopt, &l, sizeof(l));
                   1794:                break;
                   1795:
                   1796:        case SO_USELOOPBACK:
                   1797:        case SO_DONTROUTE:
                   1798:        case SO_DEBUG:
                   1799:        case SO_KEEPALIVE:
                   1800:        case SO_REUSEADDR:
                   1801:        case SO_REUSEPORT:
                   1802:        case SO_BROADCAST:
                   1803:        case SO_OOBINLINE:
                   1804:        case SO_TIMESTAMP:
                   1805:                error = sockopt_setint(sopt,
                   1806:                    (so->so_options & sopt->sopt_name) ? 1 : 0);
                   1807:                break;
                   1808:
                   1809:        case SO_TYPE:
                   1810:                error = sockopt_setint(sopt, so->so_type);
                   1811:                break;
                   1812:
                   1813:        case SO_ERROR:
                   1814:                error = sockopt_setint(sopt, so->so_error);
                   1815:                so->so_error = 0;
                   1816:                break;
                   1817:
                   1818:        case SO_SNDBUF:
                   1819:                error = sockopt_setint(sopt, so->so_snd.sb_hiwat);
                   1820:                break;
                   1821:
                   1822:        case SO_RCVBUF:
                   1823:                error = sockopt_setint(sopt, so->so_rcv.sb_hiwat);
                   1824:                break;
                   1825:
                   1826:        case SO_SNDLOWAT:
                   1827:                error = sockopt_setint(sopt, so->so_snd.sb_lowat);
                   1828:                break;
                   1829:
                   1830:        case SO_RCVLOWAT:
                   1831:                error = sockopt_setint(sopt, so->so_rcv.sb_lowat);
                   1832:                break;
                   1833:
                   1834:        case SO_SNDTIMEO:
                   1835:        case SO_RCVTIMEO:
                   1836:                optval = (sopt->sopt_name == SO_SNDTIMEO ?
                   1837:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1838:
                   1839:                tv.tv_sec = optval / hz;
                   1840:                tv.tv_usec = (optval % hz) * tick;
                   1841:
                   1842:                error = sockopt_set(sopt, &tv, sizeof(tv));
                   1843:                break;
                   1844:
                   1845:        case SO_OVERFLOWED:
                   1846:                error = sockopt_setint(sopt, so->so_rcv.sb_overflowed);
                   1847:                break;
                   1848:
                   1849:        default:
                   1850:                error = ENOPROTOOPT;
                   1851:                break;
                   1852:        }
                   1853:
                   1854:        return (error);
                   1855: }
                   1856:
1.14      mycroft  1857: int
1.171     plunky   1858: sogetopt(struct socket *so, struct sockopt *sopt)
1.1       cgd      1859: {
1.160     ad       1860:        int             error;
1.1       cgd      1861:
1.160     ad       1862:        solock(so);
1.171     plunky   1863:        if (sopt->sopt_level != SOL_SOCKET) {
1.1       cgd      1864:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       1865:                        error = ((*so->so_proto->pr_ctloutput)
1.171     plunky   1866:                            (PRCO_GETOPT, so, sopt));
1.1       cgd      1867:                } else
1.160     ad       1868:                        error = (ENOPROTOOPT);
1.1       cgd      1869:        } else {
1.171     plunky   1870:                error = sogetopt1(so, sopt);
                   1871:        }
                   1872:        sounlock(so);
                   1873:        return (error);
                   1874: }
                   1875:
                   1876: /*
                   1877:  * alloc sockopt data buffer buffer
                   1878:  *     - will be released at destroy
                   1879:  */
1.176     plunky   1880: static int
                   1881: sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag)
1.171     plunky   1882: {
                   1883:
                   1884:        KASSERT(sopt->sopt_size == 0);
                   1885:
1.176     plunky   1886:        if (len > sizeof(sopt->sopt_buf)) {
                   1887:                sopt->sopt_data = kmem_zalloc(len, kmflag);
                   1888:                if (sopt->sopt_data == NULL)
                   1889:                        return ENOMEM;
                   1890:        } else
1.171     plunky   1891:                sopt->sopt_data = sopt->sopt_buf;
                   1892:
                   1893:        sopt->sopt_size = len;
1.176     plunky   1894:        return 0;
1.171     plunky   1895: }
                   1896:
                   1897: /*
                   1898:  * initialise sockopt storage
1.176     plunky   1899:  *     - MAY sleep during allocation
1.171     plunky   1900:  */
                   1901: void
                   1902: sockopt_init(struct sockopt *sopt, int level, int name, size_t size)
                   1903: {
1.1       cgd      1904:
1.171     plunky   1905:        memset(sopt, 0, sizeof(*sopt));
1.1       cgd      1906:
1.171     plunky   1907:        sopt->sopt_level = level;
                   1908:        sopt->sopt_name = name;
1.176     plunky   1909:        (void)sockopt_alloc(sopt, size, KM_SLEEP);
1.171     plunky   1910: }
                   1911:
                   1912: /*
                   1913:  * destroy sockopt storage
                   1914:  *     - will release any held memory references
                   1915:  */
                   1916: void
                   1917: sockopt_destroy(struct sockopt *sopt)
                   1918: {
                   1919:
                   1920:        if (sopt->sopt_data != sopt->sopt_buf)
1.173     plunky   1921:                kmem_free(sopt->sopt_data, sopt->sopt_size);
1.171     plunky   1922:
                   1923:        memset(sopt, 0, sizeof(*sopt));
                   1924: }
                   1925:
                   1926: /*
                   1927:  * set sockopt value
                   1928:  *     - value is copied into sockopt
1.176     plunky   1929:  *     - memory is allocated when necessary, will not sleep
1.171     plunky   1930:  */
                   1931: int
                   1932: sockopt_set(struct sockopt *sopt, const void *buf, size_t len)
                   1933: {
1.176     plunky   1934:        int error;
1.171     plunky   1935:
1.176     plunky   1936:        if (sopt->sopt_size == 0) {
                   1937:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   1938:                if (error)
                   1939:                        return error;
                   1940:        }
1.171     plunky   1941:
                   1942:        KASSERT(sopt->sopt_size == len);
                   1943:        memcpy(sopt->sopt_data, buf, len);
                   1944:        return 0;
                   1945: }
                   1946:
                   1947: /*
                   1948:  * common case of set sockopt integer value
                   1949:  */
                   1950: int
                   1951: sockopt_setint(struct sockopt *sopt, int val)
                   1952: {
                   1953:
                   1954:        return sockopt_set(sopt, &val, sizeof(int));
                   1955: }
                   1956:
                   1957: /*
                   1958:  * get sockopt value
                   1959:  *     - correct size must be given
                   1960:  */
                   1961: int
                   1962: sockopt_get(const struct sockopt *sopt, void *buf, size_t len)
                   1963: {
1.170     tls      1964:
1.171     plunky   1965:        if (sopt->sopt_size != len)
                   1966:                return EINVAL;
1.1       cgd      1967:
1.171     plunky   1968:        memcpy(buf, sopt->sopt_data, len);
                   1969:        return 0;
                   1970: }
1.1       cgd      1971:
1.171     plunky   1972: /*
                   1973:  * common case of get sockopt integer value
                   1974:  */
                   1975: int
                   1976: sockopt_getint(const struct sockopt *sopt, int *valp)
                   1977: {
1.1       cgd      1978:
1.171     plunky   1979:        return sockopt_get(sopt, valp, sizeof(int));
                   1980: }
1.1       cgd      1981:
1.171     plunky   1982: /*
                   1983:  * set sockopt value from mbuf
                   1984:  *     - ONLY for legacy code
                   1985:  *     - mbuf is released by sockopt
1.176     plunky   1986:  *     - will not sleep
1.171     plunky   1987:  */
                   1988: int
                   1989: sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m)
                   1990: {
                   1991:        size_t len;
1.176     plunky   1992:        int error;
1.1       cgd      1993:
1.171     plunky   1994:        len = m_length(m);
1.1       cgd      1995:
1.176     plunky   1996:        if (sopt->sopt_size == 0) {
                   1997:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   1998:                if (error)
                   1999:                        return error;
                   2000:        }
1.1       cgd      2001:
1.171     plunky   2002:        KASSERT(sopt->sopt_size == len);
                   2003:        m_copydata(m, 0, len, sopt->sopt_data);
                   2004:        m_freem(m);
1.1       cgd      2005:
1.171     plunky   2006:        return 0;
                   2007: }
1.1       cgd      2008:
1.171     plunky   2009: /*
                   2010:  * get sockopt value into mbuf
                   2011:  *     - ONLY for legacy code
                   2012:  *     - mbuf to be released by the caller
1.176     plunky   2013:  *     - will not sleep
1.171     plunky   2014:  */
                   2015: struct mbuf *
                   2016: sockopt_getmbuf(const struct sockopt *sopt)
                   2017: {
                   2018:        struct mbuf *m;
1.107     darrenr  2019:
1.176     plunky   2020:        if (sopt->sopt_size > MCLBYTES)
                   2021:                return NULL;
                   2022:
                   2023:        m = m_get(M_DONTWAIT, MT_SOOPTS);
1.171     plunky   2024:        if (m == NULL)
                   2025:                return NULL;
                   2026:
1.176     plunky   2027:        if (sopt->sopt_size > MLEN) {
                   2028:                MCLGET(m, M_DONTWAIT);
                   2029:                if ((m->m_flags & M_EXT) == 0) {
                   2030:                        m_free(m);
                   2031:                        return NULL;
                   2032:                }
1.1       cgd      2033:        }
1.176     plunky   2034:
                   2035:        memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size);
                   2036:        m->m_len = sopt->sopt_size;
1.160     ad       2037:
1.171     plunky   2038:        return m;
1.1       cgd      2039: }
                   2040:
1.14      mycroft  2041: void
1.54      lukem    2042: sohasoutofband(struct socket *so)
1.1       cgd      2043: {
1.153     rmind    2044:
1.90      christos 2045:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.177.4.3! bouyer   2046:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT);
1.1       cgd      2047: }
1.72      jdolecek 2048:
                   2049: static void
                   2050: filt_sordetach(struct knote *kn)
                   2051: {
                   2052:        struct socket   *so;
                   2053:
1.155     ad       2054:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2055:        solock(so);
1.73      christos 2056:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2057:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 2058:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       2059:        sounlock(so);
1.72      jdolecek 2060: }
                   2061:
                   2062: /*ARGSUSED*/
                   2063: static int
1.129     yamt     2064: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 2065: {
                   2066:        struct socket   *so;
1.160     ad       2067:        int rv;
1.72      jdolecek 2068:
1.155     ad       2069:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2070:        if (hint != NOTE_SUBMIT)
                   2071:                solock(so);
1.72      jdolecek 2072:        kn->kn_data = so->so_rcv.sb_cc;
                   2073:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    2074:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2075:                kn->kn_fflags = so->so_error;
1.160     ad       2076:                rv = 1;
                   2077:        } else if (so->so_error)        /* temporary udp error */
                   2078:                rv = 1;
                   2079:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2080:                rv = (kn->kn_data >= kn->kn_sdata);
                   2081:        else
                   2082:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   2083:        if (hint != NOTE_SUBMIT)
                   2084:                sounlock(so);
                   2085:        return rv;
1.72      jdolecek 2086: }
                   2087:
                   2088: static void
                   2089: filt_sowdetach(struct knote *kn)
                   2090: {
                   2091:        struct socket   *so;
                   2092:
1.155     ad       2093:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2094:        solock(so);
1.73      christos 2095:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2096:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 2097:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       2098:        sounlock(so);
1.72      jdolecek 2099: }
                   2100:
                   2101: /*ARGSUSED*/
                   2102: static int
1.129     yamt     2103: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 2104: {
                   2105:        struct socket   *so;
1.160     ad       2106:        int rv;
1.72      jdolecek 2107:
1.155     ad       2108:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2109:        if (hint != NOTE_SUBMIT)
                   2110:                solock(so);
1.72      jdolecek 2111:        kn->kn_data = sbspace(&so->so_snd);
                   2112:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    2113:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2114:                kn->kn_fflags = so->so_error;
1.160     ad       2115:                rv = 1;
                   2116:        } else if (so->so_error)        /* temporary udp error */
                   2117:                rv = 1;
                   2118:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 2119:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       2120:                rv = 0;
                   2121:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2122:                rv = (kn->kn_data >= kn->kn_sdata);
                   2123:        else
                   2124:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   2125:        if (hint != NOTE_SUBMIT)
                   2126:                sounlock(so);
                   2127:        return rv;
1.72      jdolecek 2128: }
                   2129:
                   2130: /*ARGSUSED*/
                   2131: static int
1.129     yamt     2132: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 2133: {
                   2134:        struct socket   *so;
1.160     ad       2135:        int rv;
1.72      jdolecek 2136:
1.155     ad       2137:        so = ((file_t *)kn->kn_obj)->f_data;
1.72      jdolecek 2138:
                   2139:        /*
                   2140:         * Set kn_data to number of incoming connections, not
                   2141:         * counting partial (incomplete) connections.
1.108     perry    2142:         */
1.160     ad       2143:        if (hint != NOTE_SUBMIT)
                   2144:                solock(so);
1.72      jdolecek 2145:        kn->kn_data = so->so_qlen;
1.160     ad       2146:        rv = (kn->kn_data > 0);
                   2147:        if (hint != NOTE_SUBMIT)
                   2148:                sounlock(so);
                   2149:        return rv;
1.72      jdolecek 2150: }
                   2151:
                   2152: static const struct filterops solisten_filtops =
                   2153:        { 1, NULL, filt_sordetach, filt_solisten };
                   2154: static const struct filterops soread_filtops =
                   2155:        { 1, NULL, filt_sordetach, filt_soread };
                   2156: static const struct filterops sowrite_filtops =
                   2157:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   2158:
                   2159: int
1.129     yamt     2160: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 2161: {
                   2162:        struct socket   *so;
                   2163:        struct sockbuf  *sb;
                   2164:
1.155     ad       2165:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2166:        solock(so);
1.72      jdolecek 2167:        switch (kn->kn_filter) {
                   2168:        case EVFILT_READ:
                   2169:                if (so->so_options & SO_ACCEPTCONN)
                   2170:                        kn->kn_fop = &solisten_filtops;
                   2171:                else
                   2172:                        kn->kn_fop = &soread_filtops;
                   2173:                sb = &so->so_rcv;
                   2174:                break;
                   2175:        case EVFILT_WRITE:
                   2176:                kn->kn_fop = &sowrite_filtops;
                   2177:                sb = &so->so_snd;
                   2178:                break;
                   2179:        default:
1.160     ad       2180:                sounlock(so);
1.149     pooka    2181:                return (EINVAL);
1.72      jdolecek 2182:        }
1.73      christos 2183:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 2184:        sb->sb_flags |= SB_KNOTE;
1.160     ad       2185:        sounlock(so);
1.72      jdolecek 2186:        return (0);
                   2187: }
                   2188:
1.154     ad       2189: static int
                   2190: sodopoll(struct socket *so, int events)
                   2191: {
                   2192:        int revents;
                   2193:
                   2194:        revents = 0;
                   2195:
                   2196:        if (events & (POLLIN | POLLRDNORM))
                   2197:                if (soreadable(so))
                   2198:                        revents |= events & (POLLIN | POLLRDNORM);
                   2199:
                   2200:        if (events & (POLLOUT | POLLWRNORM))
                   2201:                if (sowritable(so))
                   2202:                        revents |= events & (POLLOUT | POLLWRNORM);
                   2203:
                   2204:        if (events & (POLLPRI | POLLRDBAND))
                   2205:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   2206:                        revents |= events & (POLLPRI | POLLRDBAND);
                   2207:
                   2208:        return revents;
                   2209: }
                   2210:
                   2211: int
                   2212: sopoll(struct socket *so, int events)
                   2213: {
                   2214:        int revents = 0;
                   2215:
1.160     ad       2216: #ifndef DIAGNOSTIC
                   2217:        /*
                   2218:         * Do a quick, unlocked check in expectation that the socket
                   2219:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   2220:         * as the solocked() assertions will fail.
                   2221:         */
1.154     ad       2222:        if ((revents = sodopoll(so, events)) != 0)
                   2223:                return revents;
1.160     ad       2224: #endif
1.154     ad       2225:
1.160     ad       2226:        solock(so);
1.154     ad       2227:        if ((revents = sodopoll(so, events)) == 0) {
                   2228:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   2229:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       2230:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       2231:                }
                   2232:
                   2233:                if (events & (POLLOUT | POLLWRNORM)) {
                   2234:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2235:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2236:                }
                   2237:        }
1.160     ad       2238:        sounlock(so);
1.154     ad       2239:
                   2240:        return revents;
                   2241: }
                   2242:
                   2243:
1.94      yamt     2244: #include <sys/sysctl.h>
                   2245:
                   2246: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
                   2247:
                   2248: /*
                   2249:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2250:  * value is not too small.
                   2251:  * (XXX should we maybe make sure it's not too large as well?)
                   2252:  */
                   2253: static int
                   2254: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2255: {
                   2256:        int error, new_somaxkva;
                   2257:        struct sysctlnode node;
                   2258:
                   2259:        new_somaxkva = somaxkva;
                   2260:        node = *rnode;
                   2261:        node.sysctl_data = &new_somaxkva;
                   2262:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2263:        if (error || newp == NULL)
                   2264:                return (error);
                   2265:
                   2266:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2267:                return (EINVAL);
                   2268:
1.136     ad       2269:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2270:        somaxkva = new_somaxkva;
1.136     ad       2271:        cv_broadcast(&socurkva_cv);
                   2272:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2273:
                   2274:        return (error);
                   2275: }
                   2276:
                   2277: SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup")
                   2278: {
                   2279:
1.97      atatat   2280:        sysctl_createv(clog, 0, NULL, NULL,
                   2281:                       CTLFLAG_PERMANENT,
                   2282:                       CTLTYPE_NODE, "kern", NULL,
                   2283:                       NULL, 0, NULL, 0,
                   2284:                       CTL_KERN, CTL_EOL);
                   2285:
                   2286:        sysctl_createv(clog, 0, NULL, NULL,
                   2287:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2288:                       CTLTYPE_INT, "somaxkva",
                   2289:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2290:                                    "used for socket buffers"),
1.94      yamt     2291:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2292:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
                   2293: }

CVSweb <webmaster@jp.NetBSD.org>