[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 1.209.2.2.2.1

1.209.2.2.2.1! (martin     1:: /*    $NetBSD: uipc_socket.c,v 1.209.2.2 2013/02/14 22:13:59 jdc Exp $        */
1.64      thorpej     2:
                      3: /*-
1.188     ad          4:  * Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc.
1.64      thorpej     5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
1.188     ad          8:  * by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran.
1.64      thorpej     9:  *
                     10:  * Redistribution and use in source and binary forms, with or without
                     11:  * modification, are permitted provided that the following conditions
                     12:  * are met:
                     13:  * 1. Redistributions of source code must retain the above copyright
                     14:  *    notice, this list of conditions and the following disclaimer.
                     15:  * 2. Redistributions in binary form must reproduce the above copyright
                     16:  *    notice, this list of conditions and the following disclaimer in the
                     17:  *    documentation and/or other materials provided with the distribution.
                     18:  *
                     19:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     20:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     21:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     22:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     23:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     24:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     25:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     26:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     27:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     28:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     29:  * POSSIBILITY OF SUCH DAMAGE.
                     30:  */
1.16      cgd        31:
1.1       cgd        32: /*
1.159     ad         33:  * Copyright (c) 2004 The FreeBSD Foundation
                     34:  * Copyright (c) 2004 Robert Watson
1.15      mycroft    35:  * Copyright (c) 1982, 1986, 1988, 1990, 1993
                     36:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        37:  *
                     38:  * Redistribution and use in source and binary forms, with or without
                     39:  * modification, are permitted provided that the following conditions
                     40:  * are met:
                     41:  * 1. Redistributions of source code must retain the above copyright
                     42:  *    notice, this list of conditions and the following disclaimer.
                     43:  * 2. Redistributions in binary form must reproduce the above copyright
                     44:  *    notice, this list of conditions and the following disclaimer in the
                     45:  *    documentation and/or other materials provided with the distribution.
1.85      agc        46:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        47:  *    may be used to endorse or promote products derived from this software
                     48:  *    without specific prior written permission.
                     49:  *
                     50:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     51:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     52:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     53:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     54:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     55:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     56:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     57:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     58:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     59:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     60:  * SUCH DAMAGE.
                     61:  *
1.32      fvdl       62:  *     @(#)uipc_socket.c       8.6 (Berkeley) 5/2/95
1.1       cgd        63:  */
1.59      lukem      64:
                     65: #include <sys/cdefs.h>
1.209.2.2.2.1! (martin    66:: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.209.2.2 2013/02/14 22:13:59 jdc Exp $");
1.64      thorpej    67:
1.179     christos   68: #include "opt_compat_netbsd.h"
1.64      thorpej    69: #include "opt_sock_counters.h"
                     70: #include "opt_sosend_loan.h"
1.81      martin     71: #include "opt_mbuftrace.h"
1.84      ragge      72: #include "opt_somaxkva.h"
1.167     ad         73: #include "opt_multiprocessor.h"        /* XXX */
1.1       cgd        74:
1.9       mycroft    75: #include <sys/param.h>
                     76: #include <sys/systm.h>
                     77: #include <sys/proc.h>
                     78: #include <sys/file.h>
1.142     dyoung     79: #include <sys/filedesc.h>
1.173     plunky     80: #include <sys/kmem.h>
1.9       mycroft    81: #include <sys/mbuf.h>
                     82: #include <sys/domain.h>
                     83: #include <sys/kernel.h>
                     84: #include <sys/protosw.h>
                     85: #include <sys/socket.h>
                     86: #include <sys/socketvar.h>
1.21      christos   87: #include <sys/signalvar.h>
1.9       mycroft    88: #include <sys/resourcevar.h>
1.174     pooka      89: #include <sys/uidinfo.h>
1.72      jdolecek   90: #include <sys/event.h>
1.89      christos   91: #include <sys/poll.h>
1.118     elad       92: #include <sys/kauth.h>
1.136     ad         93: #include <sys/mutex.h>
                     94: #include <sys/condvar.h>
1.205     bouyer     95: #include <sys/kthread.h>
1.37      thorpej    96:
1.179     christos   97: #ifdef COMPAT_50
                     98: #include <compat/sys/time.h>
1.184     christos   99: #include <compat/sys/socket.h>
1.179     christos  100: #endif
                    101:
1.202     uebayasi  102: #include <uvm/uvm_extern.h>
                    103: #include <uvm/uvm_loan.h>
                    104: #include <uvm/uvm_page.h>
1.64      thorpej   105:
1.77      thorpej   106: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
                    107: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37      thorpej   108:
1.142     dyoung    109: extern const struct fileops socketops;
                    110:
1.54      lukem     111: extern int     somaxconn;                      /* patchable (XXX sysctl) */
                    112: int            somaxconn = SOMAXCONN;
1.160     ad        113: kmutex_t       *softnet_lock;
1.49      jonathan  114:
1.64      thorpej   115: #ifdef SOSEND_COUNTERS
                    116: #include <sys/device.h>
                    117:
1.113     thorpej   118: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   119:     NULL, "sosend", "loan big");
1.113     thorpej   120: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   121:     NULL, "sosend", "copy big");
1.113     thorpej   122: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   123:     NULL, "sosend", "copy small");
1.113     thorpej   124: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64      thorpej   125:     NULL, "sosend", "kva limit");
                    126:
                    127: #define        SOSEND_COUNTER_INCR(ev)         (ev)->ev_count++
                    128:
1.101     matt      129: EVCNT_ATTACH_STATIC(sosend_loan_big);
                    130: EVCNT_ATTACH_STATIC(sosend_copy_big);
                    131: EVCNT_ATTACH_STATIC(sosend_copy_small);
                    132: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64      thorpej   133: #else
                    134:
                    135: #define        SOSEND_COUNTER_INCR(ev)         /* nothing */
                    136:
                    137: #endif /* SOSEND_COUNTERS */
                    138:
1.167     ad        139: #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR)
1.121     yamt      140: int sock_loan_thresh = -1;
1.71      thorpej   141: #else
1.121     yamt      142: int sock_loan_thresh = 4096;
1.65      thorpej   143: #endif
1.64      thorpej   144:
1.136     ad        145: static kmutex_t so_pendfree_lock;
1.205     bouyer    146: static struct mbuf *so_pendfree = NULL;
1.64      thorpej   147:
1.84      ragge     148: #ifndef SOMAXKVA
                    149: #define        SOMAXKVA (16 * 1024 * 1024)
                    150: #endif
                    151: int somaxkva = SOMAXKVA;
1.113     thorpej   152: static int socurkva;
1.136     ad        153: static kcondvar_t socurkva_cv;
1.64      thorpej   154:
1.191     elad      155: static kauth_listener_t socket_listener;
                    156:
1.64      thorpej   157: #define        SOCK_LOAN_CHUNK         65536
                    158:
1.205     bouyer    159: static void sopendfree_thread(void *);
                    160: static kcondvar_t pendfree_thread_cv;
                    161: static lwp_t *sopendfree_lwp;
1.93      yamt      162:
1.178     pooka     163: static void sysctl_kern_somaxkva_setup(void);
                    164: static struct sysctllog *socket_sysctllog;
                    165:
1.113     thorpej   166: static vsize_t
1.129     yamt      167: sokvareserve(struct socket *so, vsize_t len)
1.80      yamt      168: {
1.98      christos  169:        int error;
1.80      yamt      170:
1.136     ad        171:        mutex_enter(&so_pendfree_lock);
1.80      yamt      172:        while (socurkva + len > somaxkva) {
                    173:                SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136     ad        174:                error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98      christos  175:                if (error) {
                    176:                        len = 0;
                    177:                        break;
                    178:                }
1.80      yamt      179:        }
1.93      yamt      180:        socurkva += len;
1.136     ad        181:        mutex_exit(&so_pendfree_lock);
1.98      christos  182:        return len;
1.95      yamt      183: }
                    184:
1.113     thorpej   185: static void
1.95      yamt      186: sokvaunreserve(vsize_t len)
                    187: {
                    188:
1.136     ad        189:        mutex_enter(&so_pendfree_lock);
1.95      yamt      190:        socurkva -= len;
1.136     ad        191:        cv_broadcast(&socurkva_cv);
                    192:        mutex_exit(&so_pendfree_lock);
1.95      yamt      193: }
                    194:
                    195: /*
                    196:  * sokvaalloc: allocate kva for loan.
                    197:  */
                    198:
                    199: vaddr_t
1.209     matt      200: sokvaalloc(vaddr_t sva, vsize_t len, struct socket *so)
1.95      yamt      201: {
                    202:        vaddr_t lva;
                    203:
                    204:        /*
                    205:         * reserve kva.
                    206:         */
                    207:
1.98      christos  208:        if (sokvareserve(so, len) == 0)
                    209:                return 0;
1.93      yamt      210:
                    211:        /*
                    212:         * allocate kva.
                    213:         */
1.80      yamt      214:
1.209     matt      215:        lva = uvm_km_alloc(kernel_map, len, atop(sva) & uvmexp.colormask,
                    216:            UVM_KMF_COLORMATCH | UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95      yamt      217:        if (lva == 0) {
                    218:                sokvaunreserve(len);
1.80      yamt      219:                return (0);
1.95      yamt      220:        }
1.80      yamt      221:
                    222:        return lva;
                    223: }
                    224:
1.93      yamt      225: /*
                    226:  * sokvafree: free kva for loan.
                    227:  */
                    228:
1.80      yamt      229: void
                    230: sokvafree(vaddr_t sva, vsize_t len)
                    231: {
1.93      yamt      232:
                    233:        /*
                    234:         * free kva.
                    235:         */
1.80      yamt      236:
1.109     yamt      237:        uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93      yamt      238:
                    239:        /*
                    240:         * unreserve kva.
                    241:         */
                    242:
1.95      yamt      243:        sokvaunreserve(len);
1.80      yamt      244: }
                    245:
1.64      thorpej   246: static void
1.134     christos  247: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64      thorpej   248: {
1.156     yamt      249:        vaddr_t sva, eva;
1.64      thorpej   250:        vsize_t len;
1.156     yamt      251:        int npgs;
                    252:
                    253:        KASSERT(pgs != NULL);
1.64      thorpej   254:
                    255:        eva = round_page((vaddr_t) buf + size);
                    256:        sva = trunc_page((vaddr_t) buf);
                    257:        len = eva - sva;
                    258:        npgs = len >> PAGE_SHIFT;
                    259:
                    260:        pmap_kremove(sva, len);
                    261:        pmap_update(pmap_kernel());
                    262:        uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80      yamt      263:        sokvafree(sva, len);
1.64      thorpej   264: }
                    265:
1.93      yamt      266: /*
1.205     bouyer    267:  * sopendfree_thread: free mbufs on "pendfree" list.
1.136     ad        268:  * unlock and relock so_pendfree_lock when freeing mbufs.
1.93      yamt      269:  */
                    270:
1.205     bouyer    271: static void
                    272: sopendfree_thread(void *v)
1.93      yamt      273: {
1.137     ad        274:        struct mbuf *m, *next;
1.205     bouyer    275:        size_t rv;
1.93      yamt      276:
1.205     bouyer    277:        mutex_enter(&so_pendfree_lock);
1.64      thorpej   278:
1.205     bouyer    279:        for (;;) {
                    280:                rv = 0;
                    281:                while (so_pendfree != NULL) {
                    282:                        m = so_pendfree;
                    283:                        so_pendfree = NULL;
                    284:                        mutex_exit(&so_pendfree_lock);
                    285:
                    286:                        for (; m != NULL; m = next) {
                    287:                                next = m->m_next;
                    288:                                KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
                    289:                                KASSERT(m->m_ext.ext_refcnt == 0);
                    290:
                    291:                                rv += m->m_ext.ext_size;
                    292:                                sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
                    293:                                    m->m_ext.ext_size);
                    294:                                pool_cache_put(mb_cache, m);
                    295:                        }
1.93      yamt      296:
1.205     bouyer    297:                        mutex_enter(&so_pendfree_lock);
1.93      yamt      298:                }
1.205     bouyer    299:                if (rv)
                    300:                        cv_broadcast(&socurkva_cv);
                    301:                cv_wait(&pendfree_thread_cv, &so_pendfree_lock);
1.64      thorpej   302:        }
1.205     bouyer    303:        panic("sopendfree_thread");
                    304:        /* NOTREACHED */
1.64      thorpej   305: }
                    306:
1.80      yamt      307: void
1.134     christos  308: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64      thorpej   309: {
                    310:
1.156     yamt      311:        KASSERT(m != NULL);
1.64      thorpej   312:
1.93      yamt      313:        /*
                    314:         * postpone freeing mbuf.
                    315:         *
                    316:         * we can't do it in interrupt context
                    317:         * because we need to put kva back to kernel_map.
                    318:         */
                    319:
1.136     ad        320:        mutex_enter(&so_pendfree_lock);
1.92      yamt      321:        m->m_next = so_pendfree;
                    322:        so_pendfree = m;
1.205     bouyer    323:        cv_signal(&pendfree_thread_cv);
1.136     ad        324:        mutex_exit(&so_pendfree_lock);
1.64      thorpej   325: }
                    326:
                    327: static long
                    328: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
                    329: {
                    330:        struct iovec *iov = uio->uio_iov;
                    331:        vaddr_t sva, eva;
                    332:        vsize_t len;
1.156     yamt      333:        vaddr_t lva;
                    334:        int npgs, error;
                    335:        vaddr_t va;
                    336:        int i;
1.64      thorpej   337:
1.116     yamt      338:        if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64      thorpej   339:                return (0);
                    340:
                    341:        if (iov->iov_len < (size_t) space)
                    342:                space = iov->iov_len;
                    343:        if (space > SOCK_LOAN_CHUNK)
                    344:                space = SOCK_LOAN_CHUNK;
                    345:
                    346:        eva = round_page((vaddr_t) iov->iov_base + space);
                    347:        sva = trunc_page((vaddr_t) iov->iov_base);
                    348:        len = eva - sva;
                    349:        npgs = len >> PAGE_SHIFT;
                    350:
1.79      thorpej   351:        KASSERT(npgs <= M_EXT_MAXPAGES);
                    352:
1.209     matt      353:        lva = sokvaalloc(sva, len, so);
1.64      thorpej   354:        if (lva == 0)
1.80      yamt      355:                return 0;
1.64      thorpej   356:
1.116     yamt      357:        error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79      thorpej   358:            m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64      thorpej   359:        if (error) {
1.80      yamt      360:                sokvafree(lva, len);
1.64      thorpej   361:                return (0);
                    362:        }
                    363:
                    364:        for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79      thorpej   365:                pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
1.194     cegger    366:                    VM_PROT_READ, 0);
1.64      thorpej   367:        pmap_update(pmap_kernel());
                    368:
                    369:        lva += (vaddr_t) iov->iov_base & PAGE_MASK;
                    370:
1.134     christos  371:        MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79      thorpej   372:        m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64      thorpej   373:
                    374:        uio->uio_resid -= space;
                    375:        /* uio_offset not updated, not set/used for write(2) */
1.134     christos  376:        uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64      thorpej   377:        uio->uio_iov->iov_len -= space;
                    378:        if (uio->uio_iov->iov_len == 0) {
                    379:                uio->uio_iov++;
                    380:                uio->uio_iovcnt--;
                    381:        }
                    382:
                    383:        return (space);
                    384: }
                    385:
1.142     dyoung    386: struct mbuf *
1.147     dyoung    387: getsombuf(struct socket *so, int type)
1.142     dyoung    388: {
                    389:        struct mbuf *m;
                    390:
1.147     dyoung    391:        m = m_get(M_WAIT, type);
1.142     dyoung    392:        MCLAIM(m, so->so_mowner);
                    393:        return m;
                    394: }
                    395:
1.191     elad      396: static int
                    397: socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
                    398:     void *arg0, void *arg1, void *arg2, void *arg3)
                    399: {
                    400:        int result;
                    401:        enum kauth_network_req req;
                    402:
                    403:        result = KAUTH_RESULT_DEFER;
                    404:        req = (enum kauth_network_req)arg0;
                    405:
1.193     elad      406:        if ((action != KAUTH_NETWORK_SOCKET) &&
                    407:            (action != KAUTH_NETWORK_BIND))
1.191     elad      408:                return result;
                    409:
                    410:        switch (req) {
1.193     elad      411:        case KAUTH_REQ_NETWORK_BIND_PORT:
                    412:                result = KAUTH_RESULT_ALLOW;
                    413:                break;
                    414:
1.191     elad      415:        case KAUTH_REQ_NETWORK_SOCKET_DROP: {
                    416:                /* Normal users can only drop their own connections. */
                    417:                struct socket *so = (struct socket *)arg1;
                    418:
1.209.2.2.2.1! (martin   419::               if (proc_uidmatch(cred, so->so_cred) == 0)
1.191     elad      420:                        result = KAUTH_RESULT_ALLOW;
                    421:
                    422:                break;
                    423:                }
                    424:
                    425:        case KAUTH_REQ_NETWORK_SOCKET_OPEN:
                    426:                /* We allow "raw" routing/bluetooth sockets to anyone. */
1.203     matt      427:                if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_OROUTE
                    428:                    || (u_long)arg1 == PF_BLUETOOTH) {
1.191     elad      429:                        result = KAUTH_RESULT_ALLOW;
1.203     matt      430:                } else {
1.191     elad      431:                        /* Privileged, let secmodel handle this. */
                    432:                        if ((u_long)arg2 == SOCK_RAW)
                    433:                                break;
                    434:                }
                    435:
                    436:                result = KAUTH_RESULT_ALLOW;
                    437:
                    438:                break;
                    439:
1.192     elad      440:        case KAUTH_REQ_NETWORK_SOCKET_CANSEE:
                    441:                result = KAUTH_RESULT_ALLOW;
                    442:
                    443:                break;
                    444:
1.191     elad      445:        default:
                    446:                break;
                    447:        }
                    448:
                    449:        return result;
                    450: }
                    451:
1.119     yamt      452: void
                    453: soinit(void)
                    454: {
                    455:
1.178     pooka     456:        sysctl_kern_somaxkva_setup();
                    457:
1.148     ad        458:        mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.160     ad        459:        softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.136     ad        460:        cv_init(&socurkva_cv, "sokva");
1.205     bouyer    461:        cv_init(&pendfree_thread_cv, "sopendfr");
1.166     ad        462:        soinit2();
1.136     ad        463:
1.119     yamt      464:        /* Set the initial adjusted socket buffer size. */
                    465:        if (sb_max_set(sb_max))
                    466:                panic("bad initial sb_max value: %lu", sb_max);
                    467:
1.191     elad      468:        socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
                    469:            socket_listener_cb, NULL);
1.119     yamt      470: }
                    471:
1.205     bouyer    472: void
                    473: soinit1(void)
                    474: {
                    475:        int error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL,
                    476:            sopendfree_thread, NULL, &sopendfree_lwp, "sopendfree");
                    477:        if (error)
                    478:                panic("soinit1 %d", error);
                    479: }
                    480:
1.1       cgd       481: /*
                    482:  * Socket operation routines.
                    483:  * These routines are called by the routines in
                    484:  * sys_socket.c or from a system process, and
                    485:  * implement the semantics of socket operations by
                    486:  * switching out to the protocol specific routines.
                    487:  */
                    488: /*ARGSUSED*/
1.3       andrew    489: int
1.160     ad        490: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l,
                    491:         struct socket *lockso)
1.1       cgd       492: {
1.99      matt      493:        const struct protosw    *prp;
1.54      lukem     494:        struct socket   *so;
1.115     yamt      495:        uid_t           uid;
1.160     ad        496:        int             error;
                    497:        kmutex_t        *lock;
1.1       cgd       498:
1.132     elad      499:        error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
                    500:            KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
                    501:            KAUTH_ARG(proto));
1.140     dyoung    502:        if (error != 0)
                    503:                return error;
1.127     elad      504:
1.1       cgd       505:        if (proto)
                    506:                prp = pffindproto(dom, proto, type);
                    507:        else
                    508:                prp = pffindtype(dom, type);
1.140     dyoung    509:        if (prp == NULL) {
1.120     ginsbach  510:                /* no support for domain */
                    511:                if (pffinddomain(dom) == 0)
1.140     dyoung    512:                        return EAFNOSUPPORT;
1.120     ginsbach  513:                /* no support for socket type */
                    514:                if (proto == 0 && type != 0)
1.140     dyoung    515:                        return EPROTOTYPE;
                    516:                return EPROTONOSUPPORT;
1.120     ginsbach  517:        }
1.140     dyoung    518:        if (prp->pr_usrreq == NULL)
                    519:                return EPROTONOSUPPORT;
1.1       cgd       520:        if (prp->pr_type != type)
1.140     dyoung    521:                return EPROTOTYPE;
1.160     ad        522:
                    523:        so = soget(true);
1.1       cgd       524:        so->so_type = type;
                    525:        so->so_proto = prp;
1.33      matt      526:        so->so_send = sosend;
                    527:        so->so_receive = soreceive;
1.78      matt      528: #ifdef MBUFTRACE
                    529:        so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
                    530:        so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
                    531:        so->so_mowner = &prp->pr_domain->dom_mowner;
                    532: #endif
1.138     rmind     533:        uid = kauth_cred_geteuid(l->l_cred);
1.115     yamt      534:        so->so_uidinfo = uid_find(uid);
1.168     yamt      535:        so->so_cpid = l->l_proc->p_pid;
1.160     ad        536:        if (lockso != NULL) {
                    537:                /* Caller wants us to share a lock. */
                    538:                lock = lockso->so_lock;
                    539:                so->so_lock = lock;
                    540:                mutex_obj_hold(lock);
                    541:                mutex_enter(lock);
                    542:        } else {
                    543:                /* Lock assigned and taken during PRU_ATTACH. */
                    544:        }
1.140     dyoung    545:        error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
                    546:            (struct mbuf *)(long)proto, NULL, l);
1.160     ad        547:        KASSERT(solocked(so));
1.140     dyoung    548:        if (error != 0) {
1.1       cgd       549:                so->so_state |= SS_NOFDREF;
                    550:                sofree(so);
1.140     dyoung    551:                return error;
1.1       cgd       552:        }
1.198     elad      553:        so->so_cred = kauth_cred_dup(l->l_cred);
1.160     ad        554:        sounlock(so);
1.1       cgd       555:        *aso = so;
1.140     dyoung    556:        return 0;
1.1       cgd       557: }
                    558:
1.142     dyoung    559: /* On success, write file descriptor to fdout and return zero.  On
                    560:  * failure, return non-zero; *fdout will be undefined.
                    561:  */
                    562: int
                    563: fsocreate(int domain, struct socket **sop, int type, int protocol,
                    564:     struct lwp *l, int *fdout)
                    565: {
                    566:        struct socket   *so;
                    567:        struct file     *fp;
                    568:        int             fd, error;
1.204     christos  569:        int             flags = type & SOCK_FLAGS_MASK;
1.142     dyoung    570:
1.204     christos  571:        type &= ~SOCK_FLAGS_MASK;
1.155     ad        572:        if ((error = fd_allocfile(&fp, &fd)) != 0)
1.204     christos  573:                return error;
                    574:        fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0);
1.207     christos  575:        fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)|
                    576:            ((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0);
1.142     dyoung    577:        fp->f_type = DTYPE_SOCKET;
                    578:        fp->f_ops = &socketops;
1.160     ad        579:        error = socreate(domain, &so, type, protocol, l, NULL);
1.142     dyoung    580:        if (error != 0) {
1.155     ad        581:                fd_abort(curproc, fp, fd);
1.142     dyoung    582:        } else {
                    583:                if (sop != NULL)
                    584:                        *sop = so;
                    585:                fp->f_data = so;
1.155     ad        586:                fd_affix(curproc, fp, fd);
1.142     dyoung    587:                *fdout = fd;
1.209.2.2  jdc       588:                if (flags & SOCK_NONBLOCK)
                    589:                        so->so_state |= SS_NBIO;
1.142     dyoung    590:        }
                    591:        return error;
                    592: }
                    593:
1.3       andrew    594: int
1.190     dyoung    595: sofamily(const struct socket *so)
                    596: {
                    597:        const struct protosw *pr;
                    598:        const struct domain *dom;
                    599:
                    600:        if ((pr = so->so_proto) == NULL)
                    601:                return AF_UNSPEC;
                    602:        if ((dom = pr->pr_domain) == NULL)
                    603:                return AF_UNSPEC;
                    604:        return dom->dom_family;
                    605: }
                    606:
                    607: int
1.114     christos  608: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       609: {
1.160     ad        610:        int     error;
1.1       cgd       611:
1.160     ad        612:        solock(so);
1.140     dyoung    613:        error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.160     ad        614:        sounlock(so);
1.140     dyoung    615:        return error;
1.1       cgd       616: }
                    617:
1.3       andrew    618: int
1.150     elad      619: solisten(struct socket *so, int backlog, struct lwp *l)
1.1       cgd       620: {
1.160     ad        621:        int     error;
1.1       cgd       622:
1.160     ad        623:        solock(so);
1.158     ad        624:        if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
1.163     ad        625:            SS_ISDISCONNECTING)) != 0) {
                    626:                sounlock(so);
1.158     ad        627:                return (EOPNOTSUPP);
1.163     ad        628:        }
1.140     dyoung    629:        error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
1.150     elad      630:            NULL, NULL, l);
1.140     dyoung    631:        if (error != 0) {
1.160     ad        632:                sounlock(so);
1.140     dyoung    633:                return error;
1.1       cgd       634:        }
1.63      matt      635:        if (TAILQ_EMPTY(&so->so_q))
1.1       cgd       636:                so->so_options |= SO_ACCEPTCONN;
                    637:        if (backlog < 0)
                    638:                backlog = 0;
1.49      jonathan  639:        so->so_qlimit = min(backlog, somaxconn);
1.160     ad        640:        sounlock(so);
1.140     dyoung    641:        return 0;
1.1       cgd       642: }
                    643:
1.21      christos  644: void
1.54      lukem     645: sofree(struct socket *so)
1.1       cgd       646: {
1.161     ad        647:        u_int refs;
1.1       cgd       648:
1.160     ad        649:        KASSERT(solocked(so));
                    650:
                    651:        if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) {
                    652:                sounlock(so);
1.1       cgd       653:                return;
1.160     ad        654:        }
1.43      mycroft   655:        if (so->so_head) {
                    656:                /*
                    657:                 * We must not decommission a socket that's on the accept(2)
                    658:                 * queue.  If we do, then accept(2) may hang after select(2)
                    659:                 * indicated that the listening socket was ready.
                    660:                 */
1.160     ad        661:                if (!soqremque(so, 0)) {
                    662:                        sounlock(so);
1.43      mycroft   663:                        return;
1.160     ad        664:                }
1.43      mycroft   665:        }
1.98      christos  666:        if (so->so_rcv.sb_hiwat)
1.110     christos  667:                (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98      christos  668:                    RLIM_INFINITY);
                    669:        if (so->so_snd.sb_hiwat)
1.110     christos  670:                (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98      christos  671:                    RLIM_INFINITY);
                    672:        sbrelease(&so->so_snd, so);
1.160     ad        673:        KASSERT(!cv_has_waiters(&so->so_cv));
                    674:        KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv));
                    675:        KASSERT(!cv_has_waiters(&so->so_snd.sb_cv));
1.1       cgd       676:        sorflush(so);
1.161     ad        677:        refs = so->so_aborting; /* XXX */
1.177     ad        678:        /* Remove acccept filter if one is present. */
1.170     tls       679:        if (so->so_accf != NULL)
1.177     ad        680:                (void)accept_filt_clear(so);
1.160     ad        681:        sounlock(so);
1.161     ad        682:        if (refs == 0)          /* XXX */
                    683:                soput(so);
1.1       cgd       684: }
                    685:
                    686: /*
                    687:  * Close a socket on last file table reference removal.
                    688:  * Initiate disconnect if connected.
                    689:  * Free socket when disconnect complete.
                    690:  */
1.3       andrew    691: int
1.54      lukem     692: soclose(struct socket *so)
1.1       cgd       693: {
1.54      lukem     694:        struct socket   *so2;
1.160     ad        695:        int             error;
                    696:        int             error2;
1.1       cgd       697:
1.54      lukem     698:        error = 0;
1.160     ad        699:        solock(so);
1.1       cgd       700:        if (so->so_options & SO_ACCEPTCONN) {
1.172     ad        701:                for (;;) {
                    702:                        if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.160     ad        703:                                KASSERT(solocked2(so, so2));
                    704:                                (void) soqremque(so2, 0);
                    705:                                /* soabort drops the lock. */
                    706:                                (void) soabort(so2);
                    707:                                solock(so);
1.172     ad        708:                                continue;
1.160     ad        709:                        }
1.172     ad        710:                        if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.160     ad        711:                                KASSERT(solocked2(so, so2));
                    712:                                (void) soqremque(so2, 1);
                    713:                                /* soabort drops the lock. */
                    714:                                (void) soabort(so2);
                    715:                                solock(so);
1.172     ad        716:                                continue;
1.160     ad        717:                        }
1.172     ad        718:                        break;
                    719:                }
1.1       cgd       720:        }
                    721:        if (so->so_pcb == 0)
                    722:                goto discard;
                    723:        if (so->so_state & SS_ISCONNECTED) {
                    724:                if ((so->so_state & SS_ISDISCONNECTING) == 0) {
                    725:                        error = sodisconnect(so);
                    726:                        if (error)
                    727:                                goto drop;
                    728:                }
                    729:                if (so->so_options & SO_LINGER) {
1.206     christos  730:                        if ((so->so_state & (SS_ISDISCONNECTING|SS_NBIO)) ==
                    731:                            (SS_ISDISCONNECTING|SS_NBIO))
1.1       cgd       732:                                goto drop;
1.21      christos  733:                        while (so->so_state & SS_ISCONNECTED) {
1.185     yamt      734:                                error = sowait(so, true, so->so_linger * hz);
1.21      christos  735:                                if (error)
1.1       cgd       736:                                        break;
1.21      christos  737:                        }
1.1       cgd       738:                }
                    739:        }
1.54      lukem     740:  drop:
1.1       cgd       741:        if (so->so_pcb) {
1.160     ad        742:                error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140     dyoung    743:                    NULL, NULL, NULL, NULL);
1.1       cgd       744:                if (error == 0)
                    745:                        error = error2;
                    746:        }
1.54      lukem     747:  discard:
1.1       cgd       748:        if (so->so_state & SS_NOFDREF)
                    749:                panic("soclose: NOFDREF");
1.198     elad      750:        kauth_cred_free(so->so_cred);
1.1       cgd       751:        so->so_state |= SS_NOFDREF;
                    752:        sofree(so);
                    753:        return (error);
                    754: }
                    755:
                    756: /*
1.160     ad        757:  * Must be called with the socket locked..  Will return with it unlocked.
1.1       cgd       758:  */
1.3       andrew    759: int
1.54      lukem     760: soabort(struct socket *so)
1.1       cgd       761: {
1.161     ad        762:        u_int refs;
1.139     yamt      763:        int error;
1.160     ad        764:
                    765:        KASSERT(solocked(so));
                    766:        KASSERT(so->so_head == NULL);
1.1       cgd       767:
1.161     ad        768:        so->so_aborting++;              /* XXX */
1.140     dyoung    769:        error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
                    770:            NULL, NULL, NULL);
1.161     ad        771:        refs = --so->so_aborting;       /* XXX */
1.164     drochner  772:        if (error || (refs == 0)) {
1.139     yamt      773:                sofree(so);
1.160     ad        774:        } else {
                    775:                sounlock(so);
1.139     yamt      776:        }
                    777:        return error;
1.1       cgd       778: }
                    779:
1.3       andrew    780: int
1.54      lukem     781: soaccept(struct socket *so, struct mbuf *nam)
1.1       cgd       782: {
1.160     ad        783:        int     error;
                    784:
                    785:        KASSERT(solocked(so));
1.1       cgd       786:
1.54      lukem     787:        error = 0;
1.1       cgd       788:        if ((so->so_state & SS_NOFDREF) == 0)
                    789:                panic("soaccept: !NOFDREF");
                    790:        so->so_state &= ~SS_NOFDREF;
1.55      thorpej   791:        if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
                    792:            (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41      mycroft   793:                error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140     dyoung    794:                    NULL, nam, NULL, NULL);
1.41      mycroft   795:        else
1.53      itojun    796:                error = ECONNABORTED;
1.52      itojun    797:
1.1       cgd       798:        return (error);
                    799: }
                    800:
1.3       andrew    801: int
1.114     christos  802: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1       cgd       803: {
1.160     ad        804:        int             error;
                    805:
                    806:        KASSERT(solocked(so));
1.1       cgd       807:
                    808:        if (so->so_options & SO_ACCEPTCONN)
                    809:                return (EOPNOTSUPP);
                    810:        /*
                    811:         * If protocol is connection-based, can only connect once.
                    812:         * Otherwise, if connected, try to disconnect first.
                    813:         * This allows user to disconnect by connecting to, e.g.,
                    814:         * a null address.
                    815:         */
                    816:        if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
                    817:            ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
                    818:            (error = sodisconnect(so))))
                    819:                error = EISCONN;
                    820:        else
                    821:                error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140     dyoung    822:                    NULL, nam, NULL, l);
1.1       cgd       823:        return (error);
                    824: }
                    825:
1.3       andrew    826: int
1.54      lukem     827: soconnect2(struct socket *so1, struct socket *so2)
1.1       cgd       828: {
1.160     ad        829:        int     error;
                    830:
                    831:        KASSERT(solocked2(so1, so2));
1.1       cgd       832:
1.22      mycroft   833:        error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140     dyoung    834:            NULL, (struct mbuf *)so2, NULL, NULL);
1.1       cgd       835:        return (error);
                    836: }
                    837:
1.3       andrew    838: int
1.54      lukem     839: sodisconnect(struct socket *so)
1.1       cgd       840: {
1.160     ad        841:        int     error;
                    842:
                    843:        KASSERT(solocked(so));
1.1       cgd       844:
                    845:        if ((so->so_state & SS_ISCONNECTED) == 0) {
                    846:                error = ENOTCONN;
1.160     ad        847:        } else if (so->so_state & SS_ISDISCONNECTING) {
1.1       cgd       848:                error = EALREADY;
1.160     ad        849:        } else {
                    850:                error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
                    851:                    NULL, NULL, NULL, NULL);
1.1       cgd       852:        }
                    853:        return (error);
                    854: }
                    855:
1.15      mycroft   856: #define        SBLOCKWAIT(f)   (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1       cgd       857: /*
                    858:  * Send on a socket.
                    859:  * If send must go all at once and message is larger than
                    860:  * send buffering, then hard error.
                    861:  * Lock against other senders.
                    862:  * If must go all at once and not enough room now, then
                    863:  * inform user that this would block and do nothing.
                    864:  * Otherwise, if nonblocking, send as much as possible.
                    865:  * The data to be sent is described by "uio" if nonzero,
                    866:  * otherwise by the mbuf chain "top" (which must be null
                    867:  * if uio is not).  Data provided in mbuf chain must be small
                    868:  * enough to send all at once.
                    869:  *
                    870:  * Returns nonzero on error, timeout or signal; callers
                    871:  * must check for short counts if EINTR/ERESTART are returned.
                    872:  * Data and control buffers are freed on return.
                    873:  */
1.3       andrew    874: int
1.54      lukem     875: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114     christos  876:        struct mbuf *control, int flags, struct lwp *l)
1.1       cgd       877: {
1.54      lukem     878:        struct mbuf     **mp, *m;
1.114     christos  879:        struct proc     *p;
1.58      jdolecek  880:        long            space, len, resid, clen, mlen;
                    881:        int             error, s, dontroute, atomic;
1.196     dsl       882:        short           wakeup_state = 0;
1.54      lukem     883:
1.114     christos  884:        p = l->l_proc;
1.160     ad        885:        clen = 0;
1.64      thorpej   886:
1.160     ad        887:        /*
                    888:         * solock() provides atomicity of access.  splsoftnet() prevents
                    889:         * protocol processing soft interrupts from interrupting us and
                    890:         * blocking (expensive).
                    891:         */
                    892:        s = splsoftnet();
                    893:        solock(so);
1.54      lukem     894:        atomic = sosendallatonce(so) || top;
1.1       cgd       895:        if (uio)
                    896:                resid = uio->uio_resid;
                    897:        else
                    898:                resid = top->m_pkthdr.len;
1.7       cgd       899:        /*
                    900:         * In theory resid should be unsigned.
                    901:         * However, space must be signed, as it might be less than 0
                    902:         * if we over-committed, and we must use a signed comparison
                    903:         * of space and resid.  On the other hand, a negative resid
                    904:         * causes us to loop sending 0-length segments to the protocol.
                    905:         */
1.29      mycroft   906:        if (resid < 0) {
                    907:                error = EINVAL;
                    908:                goto out;
                    909:        }
1.1       cgd       910:        dontroute =
                    911:            (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
                    912:            (so->so_proto->pr_flags & PR_ATOMIC);
1.165     christos  913:        l->l_ru.ru_msgsnd++;
1.1       cgd       914:        if (control)
                    915:                clen = control->m_len;
1.54      lukem     916:  restart:
1.21      christos  917:        if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1       cgd       918:                goto out;
                    919:        do {
1.160     ad        920:                if (so->so_state & SS_CANTSENDMORE) {
                    921:                        error = EPIPE;
                    922:                        goto release;
                    923:                }
1.48      thorpej   924:                if (so->so_error) {
                    925:                        error = so->so_error;
                    926:                        so->so_error = 0;
                    927:                        goto release;
                    928:                }
1.1       cgd       929:                if ((so->so_state & SS_ISCONNECTED) == 0) {
                    930:                        if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
                    931:                                if ((so->so_state & SS_ISCONFIRMING) == 0 &&
1.160     ad        932:                                    !(resid == 0 && clen != 0)) {
                    933:                                        error = ENOTCONN;
                    934:                                        goto release;
                    935:                                }
                    936:                        } else if (addr == 0) {
                    937:                                error = EDESTADDRREQ;
                    938:                                goto release;
                    939:                        }
1.1       cgd       940:                }
                    941:                space = sbspace(&so->so_snd);
                    942:                if (flags & MSG_OOB)
                    943:                        space += 1024;
1.21      christos  944:                if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.160     ad        945:                    clen > so->so_snd.sb_hiwat) {
                    946:                        error = EMSGSIZE;
                    947:                        goto release;
                    948:                }
1.96      mycroft   949:                if (space < resid + clen &&
1.1       cgd       950:                    (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.206     christos  951:                        if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) {
1.160     ad        952:                                error = EWOULDBLOCK;
                    953:                                goto release;
                    954:                        }
1.1       cgd       955:                        sbunlock(&so->so_snd);
1.196     dsl       956:                        if (wakeup_state & SS_RESTARTSYS) {
                    957:                                error = ERESTART;
                    958:                                goto out;
                    959:                        }
1.1       cgd       960:                        error = sbwait(&so->so_snd);
                    961:                        if (error)
                    962:                                goto out;
1.196     dsl       963:                        wakeup_state = so->so_state;
1.1       cgd       964:                        goto restart;
                    965:                }
1.196     dsl       966:                wakeup_state = 0;
1.1       cgd       967:                mp = &top;
                    968:                space -= clen;
                    969:                do {
1.45      tv        970:                        if (uio == NULL) {
                    971:                                /*
                    972:                                 * Data is prepackaged in "top".
                    973:                                 */
                    974:                                resid = 0;
                    975:                                if (flags & MSG_EOR)
                    976:                                        top->m_flags |= M_EOR;
                    977:                        } else do {
1.160     ad        978:                                sounlock(so);
                    979:                                splx(s);
1.144     dyoung    980:                                if (top == NULL) {
1.78      matt      981:                                        m = m_gethdr(M_WAIT, MT_DATA);
1.45      tv        982:                                        mlen = MHLEN;
                    983:                                        m->m_pkthdr.len = 0;
1.140     dyoung    984:                                        m->m_pkthdr.rcvif = NULL;
1.45      tv        985:                                } else {
1.78      matt      986:                                        m = m_get(M_WAIT, MT_DATA);
1.45      tv        987:                                        mlen = MLEN;
                    988:                                }
1.78      matt      989:                                MCLAIM(m, so->so_snd.sb_mowner);
1.121     yamt      990:                                if (sock_loan_thresh >= 0 &&
                    991:                                    uio->uio_iov->iov_len >= sock_loan_thresh &&
                    992:                                    space >= sock_loan_thresh &&
1.64      thorpej   993:                                    (len = sosend_loan(so, uio, m,
                    994:                                                       space)) != 0) {
                    995:                                        SOSEND_COUNTER_INCR(&sosend_loan_big);
                    996:                                        space -= len;
                    997:                                        goto have_data;
                    998:                                }
1.45      tv        999:                                if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64      thorpej  1000:                                        SOSEND_COUNTER_INCR(&sosend_copy_big);
1.201     oki      1001:                                        m_clget(m, M_DONTWAIT);
1.45      tv       1002:                                        if ((m->m_flags & M_EXT) == 0)
                   1003:                                                goto nopages;
                   1004:                                        mlen = MCLBYTES;
                   1005:                                        if (atomic && top == 0) {
1.58      jdolecek 1006:                                                len = lmin(MCLBYTES - max_hdr,
1.54      lukem    1007:                                                    resid);
1.45      tv       1008:                                                m->m_data += max_hdr;
                   1009:                                        } else
1.58      jdolecek 1010:                                                len = lmin(MCLBYTES, resid);
1.45      tv       1011:                                        space -= len;
                   1012:                                } else {
1.64      thorpej  1013:  nopages:
                   1014:                                        SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58      jdolecek 1015:                                        len = lmin(lmin(mlen, resid), space);
1.45      tv       1016:                                        space -= len;
                   1017:                                        /*
                   1018:                                         * For datagram protocols, leave room
                   1019:                                         * for protocol headers in first mbuf.
                   1020:                                         */
                   1021:                                        if (atomic && top == 0 && len < mlen)
                   1022:                                                MH_ALIGN(m, len);
                   1023:                                }
1.144     dyoung   1024:                                error = uiomove(mtod(m, void *), (int)len, uio);
1.64      thorpej  1025:  have_data:
1.45      tv       1026:                                resid = uio->uio_resid;
                   1027:                                m->m_len = len;
                   1028:                                *mp = m;
                   1029:                                top->m_pkthdr.len += len;
1.160     ad       1030:                                s = splsoftnet();
                   1031:                                solock(so);
1.144     dyoung   1032:                                if (error != 0)
1.45      tv       1033:                                        goto release;
                   1034:                                mp = &m->m_next;
                   1035:                                if (resid <= 0) {
                   1036:                                        if (flags & MSG_EOR)
                   1037:                                                top->m_flags |= M_EOR;
                   1038:                                        break;
                   1039:                                }
                   1040:                        } while (space > 0 && atomic);
1.108     perry    1041:
1.160     ad       1042:                        if (so->so_state & SS_CANTSENDMORE) {
                   1043:                                error = EPIPE;
                   1044:                                goto release;
                   1045:                        }
1.45      tv       1046:                        if (dontroute)
                   1047:                                so->so_options |= SO_DONTROUTE;
                   1048:                        if (resid > 0)
                   1049:                                so->so_state |= SS_MORETOCOME;
1.46      sommerfe 1050:                        error = (*so->so_proto->pr_usrreq)(so,
                   1051:                            (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.160     ad       1052:                            top, addr, control, curlwp);
1.45      tv       1053:                        if (dontroute)
                   1054:                                so->so_options &= ~SO_DONTROUTE;
                   1055:                        if (resid > 0)
                   1056:                                so->so_state &= ~SS_MORETOCOME;
                   1057:                        clen = 0;
1.144     dyoung   1058:                        control = NULL;
                   1059:                        top = NULL;
1.45      tv       1060:                        mp = &top;
1.144     dyoung   1061:                        if (error != 0)
1.1       cgd      1062:                                goto release;
                   1063:                } while (resid && space > 0);
                   1064:        } while (resid);
                   1065:
1.54      lukem    1066:  release:
1.1       cgd      1067:        sbunlock(&so->so_snd);
1.54      lukem    1068:  out:
1.160     ad       1069:        sounlock(so);
                   1070:        splx(s);
1.1       cgd      1071:        if (top)
                   1072:                m_freem(top);
                   1073:        if (control)
                   1074:                m_freem(control);
                   1075:        return (error);
                   1076: }
                   1077:
                   1078: /*
1.159     ad       1079:  * Following replacement or removal of the first mbuf on the first
                   1080:  * mbuf chain of a socket buffer, push necessary state changes back
                   1081:  * into the socket buffer so that other consumers see the values
                   1082:  * consistently.  'nextrecord' is the callers locally stored value of
                   1083:  * the original value of sb->sb_mb->m_nextpkt which must be restored
                   1084:  * when the lead mbuf changes.  NOTE: 'nextrecord' may be NULL.
                   1085:  */
                   1086: static void
                   1087: sbsync(struct sockbuf *sb, struct mbuf *nextrecord)
                   1088: {
                   1089:
1.160     ad       1090:        KASSERT(solocked(sb->sb_so));
                   1091:
1.159     ad       1092:        /*
                   1093:         * First, update for the new value of nextrecord.  If necessary,
                   1094:         * make it the first record.
                   1095:         */
                   1096:        if (sb->sb_mb != NULL)
                   1097:                sb->sb_mb->m_nextpkt = nextrecord;
                   1098:        else
                   1099:                sb->sb_mb = nextrecord;
                   1100:
                   1101:         /*
                   1102:          * Now update any dependent socket buffer fields to reflect
                   1103:          * the new state.  This is an inline of SB_EMPTY_FIXUP, with
                   1104:          * the addition of a second clause that takes care of the
                   1105:          * case where sb_mb has been updated, but remains the last
                   1106:          * record.
                   1107:          */
                   1108:         if (sb->sb_mb == NULL) {
                   1109:                 sb->sb_mbtail = NULL;
                   1110:                 sb->sb_lastrecord = NULL;
                   1111:         } else if (sb->sb_mb->m_nextpkt == NULL)
                   1112:                 sb->sb_lastrecord = sb->sb_mb;
                   1113: }
                   1114:
                   1115: /*
1.1       cgd      1116:  * Implement receive operations on a socket.
                   1117:  * We depend on the way that records are added to the sockbuf
                   1118:  * by sbappend*.  In particular, each record (mbufs linked through m_next)
                   1119:  * must begin with an address if the protocol so specifies,
                   1120:  * followed by an optional mbuf or mbufs containing ancillary data,
                   1121:  * and then zero or more mbufs of data.
                   1122:  * In order to avoid blocking network interrupts for the entire time here,
                   1123:  * we splx() while doing the actual copy to user space.
                   1124:  * Although the sockbuf is locked, new data may still be appended,
                   1125:  * and thus we must maintain consistency of the sockbuf during that time.
                   1126:  *
                   1127:  * The caller may receive the data as a single mbuf chain by supplying
                   1128:  * an mbuf **mp0 for use in returning the chain.  The uio is then used
                   1129:  * only for the count in uio_resid.
                   1130:  */
1.3       andrew   1131: int
1.54      lukem    1132: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
                   1133:        struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1       cgd      1134: {
1.116     yamt     1135:        struct lwp *l = curlwp;
1.160     ad       1136:        struct mbuf     *m, **mp, *mt;
1.209.2.1  riz      1137:        size_t len, offset, moff, orig_resid;
                   1138:        int atomic, flags, error, s, type;
1.99      matt     1139:        const struct protosw    *pr;
1.54      lukem    1140:        struct mbuf     *nextrecord;
1.67      he       1141:        int             mbuf_removed = 0;
1.146     dyoung   1142:        const struct domain *dom;
1.196     dsl      1143:        short           wakeup_state = 0;
1.64      thorpej  1144:
1.54      lukem    1145:        pr = so->so_proto;
1.146     dyoung   1146:        atomic = pr->pr_flags & PR_ATOMIC;
                   1147:        dom = pr->pr_domain;
1.1       cgd      1148:        mp = mp0;
1.54      lukem    1149:        type = 0;
                   1150:        orig_resid = uio->uio_resid;
1.102     jonathan 1151:
1.144     dyoung   1152:        if (paddr != NULL)
                   1153:                *paddr = NULL;
                   1154:        if (controlp != NULL)
                   1155:                *controlp = NULL;
                   1156:        if (flagsp != NULL)
1.1       cgd      1157:                flags = *flagsp &~ MSG_EOR;
                   1158:        else
                   1159:                flags = 0;
1.66      enami    1160:
1.1       cgd      1161:        if (flags & MSG_OOB) {
                   1162:                m = m_get(M_WAIT, MT_DATA);
1.160     ad       1163:                solock(so);
1.17      cgd      1164:                error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140     dyoung   1165:                    (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.160     ad       1166:                sounlock(so);
1.1       cgd      1167:                if (error)
                   1168:                        goto bad;
                   1169:                do {
1.134     christos 1170:                        error = uiomove(mtod(m, void *),
1.209.2.1  riz      1171:                            MIN(uio->uio_resid, m->m_len), uio);
1.1       cgd      1172:                        m = m_free(m);
1.144     dyoung   1173:                } while (uio->uio_resid > 0 && error == 0 && m);
1.54      lukem    1174:  bad:
1.144     dyoung   1175:                if (m != NULL)
1.1       cgd      1176:                        m_freem(m);
1.144     dyoung   1177:                return error;
1.1       cgd      1178:        }
1.144     dyoung   1179:        if (mp != NULL)
1.140     dyoung   1180:                *mp = NULL;
1.160     ad       1181:
                   1182:        /*
                   1183:         * solock() provides atomicity of access.  splsoftnet() prevents
                   1184:         * protocol processing soft interrupts from interrupting us and
                   1185:         * blocking (expensive).
                   1186:         */
                   1187:        s = splsoftnet();
                   1188:        solock(so);
1.1       cgd      1189:        if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140     dyoung   1190:                (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1       cgd      1191:
1.54      lukem    1192:  restart:
1.160     ad       1193:        if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) {
                   1194:                sounlock(so);
                   1195:                splx(s);
1.144     dyoung   1196:                return error;
1.160     ad       1197:        }
1.1       cgd      1198:
                   1199:        m = so->so_rcv.sb_mb;
                   1200:        /*
                   1201:         * If we have less data than requested, block awaiting more
                   1202:         * (subject to any timeout) if:
1.15      mycroft  1203:         *   1. the current count is less than the low water mark,
1.1       cgd      1204:         *   2. MSG_WAITALL is set, and it is possible to do the entire
1.15      mycroft  1205:         *      receive operation at once if we block (resid <= hiwat), or
                   1206:         *   3. MSG_DONTWAIT is not set.
1.1       cgd      1207:         * If MSG_WAITALL is set but resid is larger than the receive buffer,
                   1208:         * we have to do the receive in sections, and thus risk returning
                   1209:         * a short count if a timeout or signal occurs after we start.
                   1210:         */
1.144     dyoung   1211:        if (m == NULL ||
                   1212:            ((flags & MSG_DONTWAIT) == 0 &&
                   1213:             so->so_rcv.sb_cc < uio->uio_resid &&
                   1214:             (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
                   1215:              ((flags & MSG_WAITALL) &&
                   1216:               uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146     dyoung   1217:             m->m_nextpkt == NULL && !atomic)) {
1.1       cgd      1218: #ifdef DIAGNOSTIC
1.144     dyoung   1219:                if (m == NULL && so->so_rcv.sb_cc)
1.1       cgd      1220:                        panic("receive 1");
                   1221: #endif
                   1222:                if (so->so_error) {
1.144     dyoung   1223:                        if (m != NULL)
1.15      mycroft  1224:                                goto dontblock;
1.1       cgd      1225:                        error = so->so_error;
                   1226:                        if ((flags & MSG_PEEK) == 0)
                   1227:                                so->so_error = 0;
                   1228:                        goto release;
                   1229:                }
                   1230:                if (so->so_state & SS_CANTRCVMORE) {
1.144     dyoung   1231:                        if (m != NULL)
1.15      mycroft  1232:                                goto dontblock;
1.1       cgd      1233:                        else
                   1234:                                goto release;
                   1235:                }
1.144     dyoung   1236:                for (; m != NULL; m = m->m_next)
1.1       cgd      1237:                        if (m->m_type == MT_OOBDATA  || (m->m_flags & M_EOR)) {
                   1238:                                m = so->so_rcv.sb_mb;
                   1239:                                goto dontblock;
                   1240:                        }
                   1241:                if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
                   1242:                    (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
                   1243:                        error = ENOTCONN;
                   1244:                        goto release;
                   1245:                }
                   1246:                if (uio->uio_resid == 0)
                   1247:                        goto release;
1.206     christos 1248:                if ((so->so_state & SS_NBIO) ||
                   1249:                    (flags & (MSG_DONTWAIT|MSG_NBIO))) {
1.1       cgd      1250:                        error = EWOULDBLOCK;
                   1251:                        goto release;
                   1252:                }
1.69      thorpej  1253:                SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
                   1254:                SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1       cgd      1255:                sbunlock(&so->so_rcv);
1.196     dsl      1256:                if (wakeup_state & SS_RESTARTSYS)
                   1257:                        error = ERESTART;
                   1258:                else
                   1259:                        error = sbwait(&so->so_rcv);
1.160     ad       1260:                if (error != 0) {
                   1261:                        sounlock(so);
                   1262:                        splx(s);
1.144     dyoung   1263:                        return error;
1.160     ad       1264:                }
1.196     dsl      1265:                wakeup_state = so->so_state;
1.1       cgd      1266:                goto restart;
                   1267:        }
1.54      lukem    1268:  dontblock:
1.69      thorpej  1269:        /*
                   1270:         * On entry here, m points to the first record of the socket buffer.
1.159     ad       1271:         * From this point onward, we maintain 'nextrecord' as a cache of the
                   1272:         * pointer to the next record in the socket buffer.  We must keep the
                   1273:         * various socket buffer pointers and local stack versions of the
                   1274:         * pointers in sync, pushing out modifications before dropping the
1.160     ad       1275:         * socket lock, and re-reading them when picking it up.
1.159     ad       1276:         *
                   1277:         * Otherwise, we will race with the network stack appending new data
                   1278:         * or records onto the socket buffer by using inconsistent/stale
                   1279:         * versions of the field, possibly resulting in socket buffer
                   1280:         * corruption.
                   1281:         *
                   1282:         * By holding the high-level sblock(), we prevent simultaneous
                   1283:         * readers from pulling off the front of the socket buffer.
1.69      thorpej  1284:         */
1.144     dyoung   1285:        if (l != NULL)
1.157     ad       1286:                l->l_ru.ru_msgrcv++;
1.69      thorpej  1287:        KASSERT(m == so->so_rcv.sb_mb);
                   1288:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
                   1289:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1       cgd      1290:        nextrecord = m->m_nextpkt;
                   1291:        if (pr->pr_flags & PR_ADDR) {
                   1292: #ifdef DIAGNOSTIC
                   1293:                if (m->m_type != MT_SONAME)
                   1294:                        panic("receive 1a");
                   1295: #endif
1.3       andrew   1296:                orig_resid = 0;
1.1       cgd      1297:                if (flags & MSG_PEEK) {
                   1298:                        if (paddr)
                   1299:                                *paddr = m_copy(m, 0, m->m_len);
                   1300:                        m = m->m_next;
                   1301:                } else {
                   1302:                        sbfree(&so->so_rcv, m);
1.67      he       1303:                        mbuf_removed = 1;
1.144     dyoung   1304:                        if (paddr != NULL) {
1.1       cgd      1305:                                *paddr = m;
                   1306:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1307:                                m->m_next = NULL;
1.1       cgd      1308:                                m = so->so_rcv.sb_mb;
                   1309:                        } else {
                   1310:                                MFREE(m, so->so_rcv.sb_mb);
                   1311:                                m = so->so_rcv.sb_mb;
                   1312:                        }
1.159     ad       1313:                        sbsync(&so->so_rcv, nextrecord);
1.1       cgd      1314:                }
                   1315:        }
1.159     ad       1316:
                   1317:        /*
                   1318:         * Process one or more MT_CONTROL mbufs present before any data mbufs
                   1319:         * in the first mbuf chain on the socket buffer.  If MSG_PEEK, we
                   1320:         * just copy the data; if !MSG_PEEK, we call into the protocol to
                   1321:         * perform externalization (or freeing if controlp == NULL).
                   1322:         */
                   1323:        if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) {
                   1324:                struct mbuf *cm = NULL, *cmn;
                   1325:                struct mbuf **cme = &cm;
                   1326:
                   1327:                do {
                   1328:                        if (flags & MSG_PEEK) {
                   1329:                                if (controlp != NULL) {
                   1330:                                        *controlp = m_copy(m, 0, m->m_len);
                   1331:                                        controlp = &(*controlp)->m_next;
                   1332:                                }
                   1333:                                m = m->m_next;
                   1334:                        } else {
                   1335:                                sbfree(&so->so_rcv, m);
1.1       cgd      1336:                                so->so_rcv.sb_mb = m->m_next;
1.144     dyoung   1337:                                m->m_next = NULL;
1.159     ad       1338:                                *cme = m;
                   1339:                                cme = &(*cme)->m_next;
1.1       cgd      1340:                                m = so->so_rcv.sb_mb;
1.159     ad       1341:                        }
                   1342:                } while (m != NULL && m->m_type == MT_CONTROL);
                   1343:                if ((flags & MSG_PEEK) == 0)
                   1344:                        sbsync(&so->so_rcv, nextrecord);
                   1345:                for (; cm != NULL; cm = cmn) {
                   1346:                        cmn = cm->m_next;
                   1347:                        cm->m_next = NULL;
                   1348:                        type = mtod(cm, struct cmsghdr *)->cmsg_type;
                   1349:                        if (controlp != NULL) {
                   1350:                                if (dom->dom_externalize != NULL &&
                   1351:                                    type == SCM_RIGHTS) {
1.160     ad       1352:                                        sounlock(so);
1.159     ad       1353:                                        splx(s);
1.204     christos 1354:                                        error = (*dom->dom_externalize)(cm, l,
                   1355:                                            (flags & MSG_CMSG_CLOEXEC) ?
                   1356:                                            O_CLOEXEC : 0);
1.159     ad       1357:                                        s = splsoftnet();
1.160     ad       1358:                                        solock(so);
1.159     ad       1359:                                }
                   1360:                                *controlp = cm;
                   1361:                                while (*controlp != NULL)
                   1362:                                        controlp = &(*controlp)->m_next;
1.1       cgd      1363:                        } else {
1.106     itojun   1364:                                /*
                   1365:                                 * Dispose of any SCM_RIGHTS message that went
                   1366:                                 * through the read path rather than recv.
                   1367:                                 */
1.159     ad       1368:                                if (dom->dom_dispose != NULL &&
                   1369:                                    type == SCM_RIGHTS) {
1.160     ad       1370:                                        sounlock(so);
1.159     ad       1371:                                        (*dom->dom_dispose)(cm);
1.160     ad       1372:                                        solock(so);
1.159     ad       1373:                                }
                   1374:                                m_freem(cm);
1.1       cgd      1375:                        }
                   1376:                }
1.159     ad       1377:                if (m != NULL)
                   1378:                        nextrecord = so->so_rcv.sb_mb->m_nextpkt;
                   1379:                else
                   1380:                        nextrecord = so->so_rcv.sb_mb;
                   1381:                orig_resid = 0;
1.1       cgd      1382:        }
1.69      thorpej  1383:
1.159     ad       1384:        /* If m is non-NULL, we have some data to read. */
                   1385:        if (__predict_true(m != NULL)) {
1.1       cgd      1386:                type = m->m_type;
                   1387:                if (type == MT_OOBDATA)
                   1388:                        flags |= MSG_OOB;
                   1389:        }
1.69      thorpej  1390:        SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
                   1391:        SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
                   1392:
1.1       cgd      1393:        moff = 0;
                   1394:        offset = 0;
1.144     dyoung   1395:        while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1       cgd      1396:                if (m->m_type == MT_OOBDATA) {
                   1397:                        if (type != MT_OOBDATA)
                   1398:                                break;
                   1399:                } else if (type == MT_OOBDATA)
                   1400:                        break;
                   1401: #ifdef DIAGNOSTIC
                   1402:                else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
                   1403:                        panic("receive 3");
                   1404: #endif
                   1405:                so->so_state &= ~SS_RCVATMARK;
1.196     dsl      1406:                wakeup_state = 0;
1.1       cgd      1407:                len = uio->uio_resid;
                   1408:                if (so->so_oobmark && len > so->so_oobmark - offset)
                   1409:                        len = so->so_oobmark - offset;
                   1410:                if (len > m->m_len - moff)
                   1411:                        len = m->m_len - moff;
                   1412:                /*
                   1413:                 * If mp is set, just pass back the mbufs.
                   1414:                 * Otherwise copy them out via the uio, then free.
                   1415:                 * Sockbuf must be consistent here (points to current mbuf,
                   1416:                 * it points to next record) when we drop priority;
                   1417:                 * we must note any additions to the sockbuf when we
                   1418:                 * block interrupts again.
                   1419:                 */
1.144     dyoung   1420:                if (mp == NULL) {
1.69      thorpej  1421:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
                   1422:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.160     ad       1423:                        sounlock(so);
1.1       cgd      1424:                        splx(s);
1.209.2.1  riz      1425:                        error = uiomove(mtod(m, char *) + moff, len, uio);
1.20      mycroft  1426:                        s = splsoftnet();
1.160     ad       1427:                        solock(so);
1.144     dyoung   1428:                        if (error != 0) {
1.67      he       1429:                                /*
                   1430:                                 * If any part of the record has been removed
                   1431:                                 * (such as the MT_SONAME mbuf, which will
                   1432:                                 * happen when PR_ADDR, and thus also
                   1433:                                 * PR_ATOMIC, is set), then drop the entire
                   1434:                                 * record to maintain the atomicity of the
                   1435:                                 * receive operation.
                   1436:                                 *
                   1437:                                 * This avoids a later panic("receive 1a")
                   1438:                                 * when compiled with DIAGNOSTIC.
                   1439:                                 */
1.146     dyoung   1440:                                if (m && mbuf_removed && atomic)
1.67      he       1441:                                        (void) sbdroprecord(&so->so_rcv);
                   1442:
1.57      jdolecek 1443:                                goto release;
1.67      he       1444:                        }
1.1       cgd      1445:                } else
                   1446:                        uio->uio_resid -= len;
                   1447:                if (len == m->m_len - moff) {
                   1448:                        if (m->m_flags & M_EOR)
                   1449:                                flags |= MSG_EOR;
                   1450:                        if (flags & MSG_PEEK) {
                   1451:                                m = m->m_next;
                   1452:                                moff = 0;
                   1453:                        } else {
                   1454:                                nextrecord = m->m_nextpkt;
                   1455:                                sbfree(&so->so_rcv, m);
                   1456:                                if (mp) {
                   1457:                                        *mp = m;
                   1458:                                        mp = &m->m_next;
                   1459:                                        so->so_rcv.sb_mb = m = m->m_next;
1.140     dyoung   1460:                                        *mp = NULL;
1.1       cgd      1461:                                } else {
                   1462:                                        MFREE(m, so->so_rcv.sb_mb);
                   1463:                                        m = so->so_rcv.sb_mb;
                   1464:                                }
1.69      thorpej  1465:                                /*
                   1466:                                 * If m != NULL, we also know that
                   1467:                                 * so->so_rcv.sb_mb != NULL.
                   1468:                                 */
                   1469:                                KASSERT(so->so_rcv.sb_mb == m);
                   1470:                                if (m) {
1.1       cgd      1471:                                        m->m_nextpkt = nextrecord;
1.69      thorpej  1472:                                        if (nextrecord == NULL)
                   1473:                                                so->so_rcv.sb_lastrecord = m;
                   1474:                                } else {
                   1475:                                        so->so_rcv.sb_mb = nextrecord;
1.70      thorpej  1476:                                        SB_EMPTY_FIXUP(&so->so_rcv);
1.69      thorpej  1477:                                }
                   1478:                                SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
                   1479:                                SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1       cgd      1480:                        }
1.144     dyoung   1481:                } else if (flags & MSG_PEEK)
                   1482:                        moff += len;
                   1483:                else {
1.160     ad       1484:                        if (mp != NULL) {
                   1485:                                mt = m_copym(m, 0, len, M_NOWAIT);
                   1486:                                if (__predict_false(mt == NULL)) {
                   1487:                                        sounlock(so);
                   1488:                                        mt = m_copym(m, 0, len, M_WAIT);
                   1489:                                        solock(so);
                   1490:                                }
                   1491:                                *mp = mt;
                   1492:                        }
1.144     dyoung   1493:                        m->m_data += len;
                   1494:                        m->m_len -= len;
                   1495:                        so->so_rcv.sb_cc -= len;
1.1       cgd      1496:                }
                   1497:                if (so->so_oobmark) {
                   1498:                        if ((flags & MSG_PEEK) == 0) {
                   1499:                                so->so_oobmark -= len;
                   1500:                                if (so->so_oobmark == 0) {
                   1501:                                        so->so_state |= SS_RCVATMARK;
                   1502:                                        break;
                   1503:                                }
1.7       cgd      1504:                        } else {
1.1       cgd      1505:                                offset += len;
1.7       cgd      1506:                                if (offset == so->so_oobmark)
                   1507:                                        break;
                   1508:                        }
1.1       cgd      1509:                }
                   1510:                if (flags & MSG_EOR)
                   1511:                        break;
                   1512:                /*
                   1513:                 * If the MSG_WAITALL flag is set (for non-atomic socket),
                   1514:                 * we must not quit until "uio->uio_resid == 0" or an error
                   1515:                 * termination.  If a signal/timeout occurs, return
                   1516:                 * with a short count but without error.
                   1517:                 * Keep sockbuf locked against other readers.
                   1518:                 */
1.144     dyoung   1519:                while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3       andrew   1520:                    !sosendallatonce(so) && !nextrecord) {
1.1       cgd      1521:                        if (so->so_error || so->so_state & SS_CANTRCVMORE)
                   1522:                                break;
1.68      matt     1523:                        /*
                   1524:                         * If we are peeking and the socket receive buffer is
                   1525:                         * full, stop since we can't get more data to peek at.
                   1526:                         */
                   1527:                        if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
                   1528:                                break;
                   1529:                        /*
                   1530:                         * If we've drained the socket buffer, tell the
                   1531:                         * protocol in case it needs to do something to
                   1532:                         * get it filled again.
                   1533:                         */
                   1534:                        if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
                   1535:                                (*pr->pr_usrreq)(so, PRU_RCVD,
1.140     dyoung   1536:                                    NULL, (struct mbuf *)(long)flags, NULL, l);
1.69      thorpej  1537:                        SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
                   1538:                        SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.196     dsl      1539:                        if (wakeup_state & SS_RESTARTSYS)
                   1540:                                error = ERESTART;
                   1541:                        else
                   1542:                                error = sbwait(&so->so_rcv);
1.144     dyoung   1543:                        if (error != 0) {
1.1       cgd      1544:                                sbunlock(&so->so_rcv);
1.160     ad       1545:                                sounlock(so);
1.1       cgd      1546:                                splx(s);
1.144     dyoung   1547:                                return 0;
1.1       cgd      1548:                        }
1.21      christos 1549:                        if ((m = so->so_rcv.sb_mb) != NULL)
1.1       cgd      1550:                                nextrecord = m->m_nextpkt;
1.196     dsl      1551:                        wakeup_state = so->so_state;
1.1       cgd      1552:                }
                   1553:        }
1.3       andrew   1554:
1.146     dyoung   1555:        if (m && atomic) {
1.3       andrew   1556:                flags |= MSG_TRUNC;
                   1557:                if ((flags & MSG_PEEK) == 0)
                   1558:                        (void) sbdroprecord(&so->so_rcv);
                   1559:        }
1.1       cgd      1560:        if ((flags & MSG_PEEK) == 0) {
1.144     dyoung   1561:                if (m == NULL) {
1.69      thorpej  1562:                        /*
1.70      thorpej  1563:                         * First part is an inline SB_EMPTY_FIXUP().  Second
1.69      thorpej  1564:                         * part makes sure sb_lastrecord is up-to-date if
                   1565:                         * there is still data in the socket buffer.
                   1566:                         */
1.1       cgd      1567:                        so->so_rcv.sb_mb = nextrecord;
1.69      thorpej  1568:                        if (so->so_rcv.sb_mb == NULL) {
                   1569:                                so->so_rcv.sb_mbtail = NULL;
                   1570:                                so->so_rcv.sb_lastrecord = NULL;
                   1571:                        } else if (nextrecord->m_nextpkt == NULL)
                   1572:                                so->so_rcv.sb_lastrecord = nextrecord;
                   1573:                }
                   1574:                SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
                   1575:                SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1       cgd      1576:                if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140     dyoung   1577:                        (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
                   1578:                            (struct mbuf *)(long)flags, NULL, l);
1.1       cgd      1579:        }
1.3       andrew   1580:        if (orig_resid == uio->uio_resid && orig_resid &&
                   1581:            (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
                   1582:                sbunlock(&so->so_rcv);
                   1583:                goto restart;
                   1584:        }
1.108     perry    1585:
1.144     dyoung   1586:        if (flagsp != NULL)
1.1       cgd      1587:                *flagsp |= flags;
1.54      lukem    1588:  release:
1.1       cgd      1589:        sbunlock(&so->so_rcv);
1.160     ad       1590:        sounlock(so);
1.1       cgd      1591:        splx(s);
1.144     dyoung   1592:        return error;
1.1       cgd      1593: }
                   1594:
1.14      mycroft  1595: int
1.54      lukem    1596: soshutdown(struct socket *so, int how)
1.1       cgd      1597: {
1.99      matt     1598:        const struct protosw    *pr;
1.160     ad       1599:        int     error;
                   1600:
                   1601:        KASSERT(solocked(so));
1.34      kleink   1602:
1.54      lukem    1603:        pr = so->so_proto;
1.34      kleink   1604:        if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
                   1605:                return (EINVAL);
1.1       cgd      1606:
1.160     ad       1607:        if (how == SHUT_RD || how == SHUT_RDWR) {
1.1       cgd      1608:                sorflush(so);
1.160     ad       1609:                error = 0;
                   1610:        }
1.34      kleink   1611:        if (how == SHUT_WR || how == SHUT_RDWR)
1.160     ad       1612:                error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1.140     dyoung   1613:                    NULL, NULL, NULL);
1.160     ad       1614:
                   1615:        return error;
1.1       cgd      1616: }
                   1617:
1.195     dsl      1618: void
1.196     dsl      1619: sorestart(struct socket *so)
1.188     ad       1620: {
1.196     dsl      1621:        /*
                   1622:         * An application has called close() on an fd on which another
                   1623:         * of its threads has called a socket system call.
                   1624:         * Mark this and wake everyone up, and code that would block again
                   1625:         * instead returns ERESTART.
                   1626:         * On system call re-entry the fd is validated and EBADF returned.
                   1627:         * Any other fd will block again on the 2nd syscall.
                   1628:         */
1.188     ad       1629:        solock(so);
1.196     dsl      1630:        so->so_state |= SS_RESTARTSYS;
1.188     ad       1631:        cv_broadcast(&so->so_cv);
1.196     dsl      1632:        cv_broadcast(&so->so_snd.sb_cv);
                   1633:        cv_broadcast(&so->so_rcv.sb_cv);
1.188     ad       1634:        sounlock(so);
                   1635: }
                   1636:
1.14      mycroft  1637: void
1.54      lukem    1638: sorflush(struct socket *so)
1.1       cgd      1639: {
1.54      lukem    1640:        struct sockbuf  *sb, asb;
1.99      matt     1641:        const struct protosw    *pr;
1.160     ad       1642:
                   1643:        KASSERT(solocked(so));
1.1       cgd      1644:
1.54      lukem    1645:        sb = &so->so_rcv;
                   1646:        pr = so->so_proto;
1.160     ad       1647:        socantrcvmore(so);
1.1       cgd      1648:        sb->sb_flags |= SB_NOINTR;
1.160     ad       1649:        (void )sblock(sb, M_WAITOK);
1.1       cgd      1650:        sbunlock(sb);
                   1651:        asb = *sb;
1.86      wrstuden 1652:        /*
                   1653:         * Clear most of the sockbuf structure, but leave some of the
                   1654:         * fields valid.
                   1655:         */
                   1656:        memset(&sb->sb_startzero, 0,
                   1657:            sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.160     ad       1658:        if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) {
                   1659:                sounlock(so);
1.1       cgd      1660:                (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.160     ad       1661:                solock(so);
                   1662:        }
1.98      christos 1663:        sbrelease(&asb, so);
1.1       cgd      1664: }
                   1665:
1.171     plunky   1666: /*
                   1667:  * internal set SOL_SOCKET options
                   1668:  */
1.142     dyoung   1669: static int
1.171     plunky   1670: sosetopt1(struct socket *so, const struct sockopt *sopt)
1.1       cgd      1671: {
1.182     christos 1672:        int error = EINVAL, optval, opt;
1.171     plunky   1673:        struct linger l;
                   1674:        struct timeval tv;
1.142     dyoung   1675:
1.179     christos 1676:        switch ((opt = sopt->sopt_name)) {
1.142     dyoung   1677:
1.170     tls      1678:        case SO_ACCEPTFILTER:
1.177     ad       1679:                error = accept_filt_setopt(so, sopt);
                   1680:                KASSERT(solocked(so));
1.170     tls      1681:                break;
                   1682:
1.171     plunky   1683:        case SO_LINGER:
                   1684:                error = sockopt_get(sopt, &l, sizeof(l));
1.177     ad       1685:                solock(so);
1.171     plunky   1686:                if (error)
1.177     ad       1687:                        break;
1.171     plunky   1688:                if (l.l_linger < 0 || l.l_linger > USHRT_MAX ||
1.177     ad       1689:                    l.l_linger > (INT_MAX / hz)) {
                   1690:                        error = EDOM;
                   1691:                        break;
                   1692:                }
1.171     plunky   1693:                so->so_linger = l.l_linger;
                   1694:                if (l.l_onoff)
                   1695:                        so->so_options |= SO_LINGER;
                   1696:                else
                   1697:                        so->so_options &= ~SO_LINGER;
1.177     ad       1698:                break;
1.1       cgd      1699:
1.142     dyoung   1700:        case SO_DEBUG:
                   1701:        case SO_KEEPALIVE:
                   1702:        case SO_DONTROUTE:
                   1703:        case SO_USELOOPBACK:
                   1704:        case SO_BROADCAST:
                   1705:        case SO_REUSEADDR:
                   1706:        case SO_REUSEPORT:
                   1707:        case SO_OOBINLINE:
                   1708:        case SO_TIMESTAMP:
1.207     christos 1709:        case SO_NOSIGPIPE:
1.184     christos 1710: #ifdef SO_OTIMESTAMP
                   1711:        case SO_OTIMESTAMP:
                   1712: #endif
1.171     plunky   1713:                error = sockopt_getint(sopt, &optval);
1.177     ad       1714:                solock(so);
1.171     plunky   1715:                if (error)
1.177     ad       1716:                        break;
1.171     plunky   1717:                if (optval)
1.179     christos 1718:                        so->so_options |= opt;
1.142     dyoung   1719:                else
1.179     christos 1720:                        so->so_options &= ~opt;
1.142     dyoung   1721:                break;
                   1722:
                   1723:        case SO_SNDBUF:
                   1724:        case SO_RCVBUF:
                   1725:        case SO_SNDLOWAT:
                   1726:        case SO_RCVLOWAT:
1.171     plunky   1727:                error = sockopt_getint(sopt, &optval);
1.177     ad       1728:                solock(so);
1.171     plunky   1729:                if (error)
1.177     ad       1730:                        break;
1.1       cgd      1731:
1.142     dyoung   1732:                /*
                   1733:                 * Values < 1 make no sense for any of these
                   1734:                 * options, so disallow them.
                   1735:                 */
1.177     ad       1736:                if (optval < 1) {
                   1737:                        error = EINVAL;
                   1738:                        break;
                   1739:                }
1.1       cgd      1740:
1.179     christos 1741:                switch (opt) {
1.171     plunky   1742:                case SO_SNDBUF:
1.177     ad       1743:                        if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) {
                   1744:                                error = ENOBUFS;
                   1745:                                break;
                   1746:                        }
1.171     plunky   1747:                        so->so_snd.sb_flags &= ~SB_AUTOSIZE;
                   1748:                        break;
1.1       cgd      1749:
                   1750:                case SO_RCVBUF:
1.177     ad       1751:                        if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) {
                   1752:                                error = ENOBUFS;
                   1753:                                break;
                   1754:                        }
1.171     plunky   1755:                        so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1.142     dyoung   1756:                        break;
                   1757:
                   1758:                /*
                   1759:                 * Make sure the low-water is never greater than
                   1760:                 * the high-water.
                   1761:                 */
1.1       cgd      1762:                case SO_SNDLOWAT:
1.171     plunky   1763:                        if (optval > so->so_snd.sb_hiwat)
                   1764:                                optval = so->so_snd.sb_hiwat;
                   1765:
                   1766:                        so->so_snd.sb_lowat = optval;
1.142     dyoung   1767:                        break;
1.171     plunky   1768:
1.1       cgd      1769:                case SO_RCVLOWAT:
1.171     plunky   1770:                        if (optval > so->so_rcv.sb_hiwat)
                   1771:                                optval = so->so_rcv.sb_hiwat;
                   1772:
                   1773:                        so->so_rcv.sb_lowat = optval;
1.142     dyoung   1774:                        break;
                   1775:                }
                   1776:                break;
1.28      thorpej  1777:
1.179     christos 1778: #ifdef COMPAT_50
                   1779:        case SO_OSNDTIMEO:
                   1780:        case SO_ORCVTIMEO: {
                   1781:                struct timeval50 otv;
                   1782:                error = sockopt_get(sopt, &otv, sizeof(otv));
1.186     pooka    1783:                if (error) {
                   1784:                        solock(so);
1.183     christos 1785:                        break;
1.186     pooka    1786:                }
1.179     christos 1787:                timeval50_to_timeval(&otv, &tv);
                   1788:                opt = opt == SO_OSNDTIMEO ? SO_SNDTIMEO : SO_RCVTIMEO;
1.182     christos 1789:                error = 0;
1.179     christos 1790:                /*FALLTHROUGH*/
                   1791:        }
                   1792: #endif /* COMPAT_50 */
                   1793:
1.142     dyoung   1794:        case SO_SNDTIMEO:
                   1795:        case SO_RCVTIMEO:
1.182     christos 1796:                if (error)
1.179     christos 1797:                        error = sockopt_get(sopt, &tv, sizeof(tv));
1.177     ad       1798:                solock(so);
1.171     plunky   1799:                if (error)
1.177     ad       1800:                        break;
1.171     plunky   1801:
1.177     ad       1802:                if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) {
                   1803:                        error = EDOM;
                   1804:                        break;
                   1805:                }
1.28      thorpej  1806:
1.171     plunky   1807:                optval = tv.tv_sec * hz + tv.tv_usec / tick;
                   1808:                if (optval == 0 && tv.tv_usec != 0)
                   1809:                        optval = 1;
1.28      thorpej  1810:
1.179     christos 1811:                switch (opt) {
1.142     dyoung   1812:                case SO_SNDTIMEO:
1.171     plunky   1813:                        so->so_snd.sb_timeo = optval;
1.1       cgd      1814:                        break;
                   1815:                case SO_RCVTIMEO:
1.171     plunky   1816:                        so->so_rcv.sb_timeo = optval;
1.142     dyoung   1817:                        break;
                   1818:                }
                   1819:                break;
1.1       cgd      1820:
1.142     dyoung   1821:        default:
1.177     ad       1822:                solock(so);
                   1823:                error = ENOPROTOOPT;
                   1824:                break;
1.142     dyoung   1825:        }
1.177     ad       1826:        KASSERT(solocked(so));
                   1827:        return error;
1.142     dyoung   1828: }
1.1       cgd      1829:
1.142     dyoung   1830: int
1.171     plunky   1831: sosetopt(struct socket *so, struct sockopt *sopt)
1.142     dyoung   1832: {
                   1833:        int error, prerr;
1.1       cgd      1834:
1.177     ad       1835:        if (sopt->sopt_level == SOL_SOCKET) {
1.171     plunky   1836:                error = sosetopt1(so, sopt);
1.177     ad       1837:                KASSERT(solocked(so));
                   1838:        } else {
1.142     dyoung   1839:                error = ENOPROTOOPT;
1.177     ad       1840:                solock(so);
                   1841:        }
1.1       cgd      1842:
1.142     dyoung   1843:        if ((error == 0 || error == ENOPROTOOPT) &&
                   1844:            so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
                   1845:                /* give the protocol stack a shot */
1.171     plunky   1846:                prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt);
1.142     dyoung   1847:                if (prerr == 0)
                   1848:                        error = 0;
                   1849:                else if (prerr != ENOPROTOOPT)
                   1850:                        error = prerr;
1.171     plunky   1851:        }
1.160     ad       1852:        sounlock(so);
1.142     dyoung   1853:        return error;
1.1       cgd      1854: }
                   1855:
1.171     plunky   1856: /*
                   1857:  * so_setsockopt() is a wrapper providing a sockopt structure for sosetopt()
                   1858:  */
                   1859: int
                   1860: so_setsockopt(struct lwp *l, struct socket *so, int level, int name,
                   1861:     const void *val, size_t valsize)
                   1862: {
                   1863:        struct sockopt sopt;
                   1864:        int error;
                   1865:
                   1866:        KASSERT(valsize == 0 || val != NULL);
                   1867:
                   1868:        sockopt_init(&sopt, level, name, valsize);
                   1869:        sockopt_set(&sopt, val, valsize);
                   1870:
                   1871:        error = sosetopt(so, &sopt);
                   1872:
                   1873:        sockopt_destroy(&sopt);
                   1874:
                   1875:        return error;
                   1876: }
                   1877:
                   1878: /*
                   1879:  * internal get SOL_SOCKET options
                   1880:  */
                   1881: static int
                   1882: sogetopt1(struct socket *so, struct sockopt *sopt)
                   1883: {
1.179     christos 1884:        int error, optval, opt;
1.171     plunky   1885:        struct linger l;
                   1886:        struct timeval tv;
                   1887:
1.179     christos 1888:        switch ((opt = sopt->sopt_name)) {
1.171     plunky   1889:
                   1890:        case SO_ACCEPTFILTER:
1.177     ad       1891:                error = accept_filt_getopt(so, sopt);
1.171     plunky   1892:                break;
                   1893:
                   1894:        case SO_LINGER:
                   1895:                l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0;
                   1896:                l.l_linger = so->so_linger;
                   1897:
                   1898:                error = sockopt_set(sopt, &l, sizeof(l));
                   1899:                break;
                   1900:
                   1901:        case SO_USELOOPBACK:
                   1902:        case SO_DONTROUTE:
                   1903:        case SO_DEBUG:
                   1904:        case SO_KEEPALIVE:
                   1905:        case SO_REUSEADDR:
                   1906:        case SO_REUSEPORT:
                   1907:        case SO_BROADCAST:
                   1908:        case SO_OOBINLINE:
                   1909:        case SO_TIMESTAMP:
1.207     christos 1910:        case SO_NOSIGPIPE:
1.184     christos 1911: #ifdef SO_OTIMESTAMP
                   1912:        case SO_OTIMESTAMP:
                   1913: #endif
1.179     christos 1914:                error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0);
1.171     plunky   1915:                break;
                   1916:
                   1917:        case SO_TYPE:
                   1918:                error = sockopt_setint(sopt, so->so_type);
                   1919:                break;
                   1920:
                   1921:        case SO_ERROR:
                   1922:                error = sockopt_setint(sopt, so->so_error);
                   1923:                so->so_error = 0;
                   1924:                break;
                   1925:
                   1926:        case SO_SNDBUF:
                   1927:                error = sockopt_setint(sopt, so->so_snd.sb_hiwat);
                   1928:                break;
                   1929:
                   1930:        case SO_RCVBUF:
                   1931:                error = sockopt_setint(sopt, so->so_rcv.sb_hiwat);
                   1932:                break;
                   1933:
                   1934:        case SO_SNDLOWAT:
                   1935:                error = sockopt_setint(sopt, so->so_snd.sb_lowat);
                   1936:                break;
                   1937:
                   1938:        case SO_RCVLOWAT:
                   1939:                error = sockopt_setint(sopt, so->so_rcv.sb_lowat);
                   1940:                break;
                   1941:
1.179     christos 1942: #ifdef COMPAT_50
                   1943:        case SO_OSNDTIMEO:
                   1944:        case SO_ORCVTIMEO: {
                   1945:                struct timeval50 otv;
                   1946:
                   1947:                optval = (opt == SO_OSNDTIMEO ?
                   1948:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1949:
                   1950:                otv.tv_sec = optval / hz;
                   1951:                otv.tv_usec = (optval % hz) * tick;
                   1952:
                   1953:                error = sockopt_set(sopt, &otv, sizeof(otv));
                   1954:                break;
                   1955:        }
                   1956: #endif /* COMPAT_50 */
                   1957:
1.171     plunky   1958:        case SO_SNDTIMEO:
                   1959:        case SO_RCVTIMEO:
1.179     christos 1960:                optval = (opt == SO_SNDTIMEO ?
1.171     plunky   1961:                     so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
                   1962:
                   1963:                tv.tv_sec = optval / hz;
                   1964:                tv.tv_usec = (optval % hz) * tick;
                   1965:
                   1966:                error = sockopt_set(sopt, &tv, sizeof(tv));
                   1967:                break;
                   1968:
                   1969:        case SO_OVERFLOWED:
                   1970:                error = sockopt_setint(sopt, so->so_rcv.sb_overflowed);
                   1971:                break;
                   1972:
                   1973:        default:
                   1974:                error = ENOPROTOOPT;
                   1975:                break;
                   1976:        }
                   1977:
                   1978:        return (error);
                   1979: }
                   1980:
1.14      mycroft  1981: int
1.171     plunky   1982: sogetopt(struct socket *so, struct sockopt *sopt)
1.1       cgd      1983: {
1.160     ad       1984:        int             error;
1.1       cgd      1985:
1.160     ad       1986:        solock(so);
1.171     plunky   1987:        if (sopt->sopt_level != SOL_SOCKET) {
1.1       cgd      1988:                if (so->so_proto && so->so_proto->pr_ctloutput) {
1.160     ad       1989:                        error = ((*so->so_proto->pr_ctloutput)
1.171     plunky   1990:                            (PRCO_GETOPT, so, sopt));
1.1       cgd      1991:                } else
1.160     ad       1992:                        error = (ENOPROTOOPT);
1.1       cgd      1993:        } else {
1.171     plunky   1994:                error = sogetopt1(so, sopt);
                   1995:        }
                   1996:        sounlock(so);
                   1997:        return (error);
                   1998: }
                   1999:
                   2000: /*
                   2001:  * alloc sockopt data buffer buffer
                   2002:  *     - will be released at destroy
                   2003:  */
1.176     plunky   2004: static int
                   2005: sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag)
1.171     plunky   2006: {
                   2007:
                   2008:        KASSERT(sopt->sopt_size == 0);
                   2009:
1.176     plunky   2010:        if (len > sizeof(sopt->sopt_buf)) {
                   2011:                sopt->sopt_data = kmem_zalloc(len, kmflag);
                   2012:                if (sopt->sopt_data == NULL)
                   2013:                        return ENOMEM;
                   2014:        } else
1.171     plunky   2015:                sopt->sopt_data = sopt->sopt_buf;
                   2016:
                   2017:        sopt->sopt_size = len;
1.176     plunky   2018:        return 0;
1.171     plunky   2019: }
                   2020:
                   2021: /*
                   2022:  * initialise sockopt storage
1.176     plunky   2023:  *     - MAY sleep during allocation
1.171     plunky   2024:  */
                   2025: void
                   2026: sockopt_init(struct sockopt *sopt, int level, int name, size_t size)
                   2027: {
1.1       cgd      2028:
1.171     plunky   2029:        memset(sopt, 0, sizeof(*sopt));
1.1       cgd      2030:
1.171     plunky   2031:        sopt->sopt_level = level;
                   2032:        sopt->sopt_name = name;
1.176     plunky   2033:        (void)sockopt_alloc(sopt, size, KM_SLEEP);
1.171     plunky   2034: }
                   2035:
                   2036: /*
                   2037:  * destroy sockopt storage
                   2038:  *     - will release any held memory references
                   2039:  */
                   2040: void
                   2041: sockopt_destroy(struct sockopt *sopt)
                   2042: {
                   2043:
                   2044:        if (sopt->sopt_data != sopt->sopt_buf)
1.173     plunky   2045:                kmem_free(sopt->sopt_data, sopt->sopt_size);
1.171     plunky   2046:
                   2047:        memset(sopt, 0, sizeof(*sopt));
                   2048: }
                   2049:
                   2050: /*
                   2051:  * set sockopt value
                   2052:  *     - value is copied into sockopt
1.176     plunky   2053:  *     - memory is allocated when necessary, will not sleep
1.171     plunky   2054:  */
                   2055: int
                   2056: sockopt_set(struct sockopt *sopt, const void *buf, size_t len)
                   2057: {
1.176     plunky   2058:        int error;
1.171     plunky   2059:
1.176     plunky   2060:        if (sopt->sopt_size == 0) {
                   2061:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2062:                if (error)
                   2063:                        return error;
                   2064:        }
1.171     plunky   2065:
                   2066:        KASSERT(sopt->sopt_size == len);
                   2067:        memcpy(sopt->sopt_data, buf, len);
                   2068:        return 0;
                   2069: }
                   2070:
                   2071: /*
                   2072:  * common case of set sockopt integer value
                   2073:  */
                   2074: int
                   2075: sockopt_setint(struct sockopt *sopt, int val)
                   2076: {
                   2077:
                   2078:        return sockopt_set(sopt, &val, sizeof(int));
                   2079: }
                   2080:
                   2081: /*
                   2082:  * get sockopt value
                   2083:  *     - correct size must be given
                   2084:  */
                   2085: int
                   2086: sockopt_get(const struct sockopt *sopt, void *buf, size_t len)
                   2087: {
1.170     tls      2088:
1.171     plunky   2089:        if (sopt->sopt_size != len)
                   2090:                return EINVAL;
1.1       cgd      2091:
1.171     plunky   2092:        memcpy(buf, sopt->sopt_data, len);
                   2093:        return 0;
                   2094: }
1.1       cgd      2095:
1.171     plunky   2096: /*
                   2097:  * common case of get sockopt integer value
                   2098:  */
                   2099: int
                   2100: sockopt_getint(const struct sockopt *sopt, int *valp)
                   2101: {
1.1       cgd      2102:
1.171     plunky   2103:        return sockopt_get(sopt, valp, sizeof(int));
                   2104: }
1.1       cgd      2105:
1.171     plunky   2106: /*
                   2107:  * set sockopt value from mbuf
                   2108:  *     - ONLY for legacy code
                   2109:  *     - mbuf is released by sockopt
1.176     plunky   2110:  *     - will not sleep
1.171     plunky   2111:  */
                   2112: int
                   2113: sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m)
                   2114: {
                   2115:        size_t len;
1.176     plunky   2116:        int error;
1.1       cgd      2117:
1.171     plunky   2118:        len = m_length(m);
1.1       cgd      2119:
1.176     plunky   2120:        if (sopt->sopt_size == 0) {
                   2121:                error = sockopt_alloc(sopt, len, KM_NOSLEEP);
                   2122:                if (error)
                   2123:                        return error;
                   2124:        }
1.1       cgd      2125:
1.171     plunky   2126:        KASSERT(sopt->sopt_size == len);
                   2127:        m_copydata(m, 0, len, sopt->sopt_data);
                   2128:        m_freem(m);
1.1       cgd      2129:
1.171     plunky   2130:        return 0;
                   2131: }
1.1       cgd      2132:
1.171     plunky   2133: /*
                   2134:  * get sockopt value into mbuf
                   2135:  *     - ONLY for legacy code
                   2136:  *     - mbuf to be released by the caller
1.176     plunky   2137:  *     - will not sleep
1.171     plunky   2138:  */
                   2139: struct mbuf *
                   2140: sockopt_getmbuf(const struct sockopt *sopt)
                   2141: {
                   2142:        struct mbuf *m;
1.107     darrenr  2143:
1.176     plunky   2144:        if (sopt->sopt_size > MCLBYTES)
                   2145:                return NULL;
                   2146:
                   2147:        m = m_get(M_DONTWAIT, MT_SOOPTS);
1.171     plunky   2148:        if (m == NULL)
                   2149:                return NULL;
                   2150:
1.176     plunky   2151:        if (sopt->sopt_size > MLEN) {
                   2152:                MCLGET(m, M_DONTWAIT);
                   2153:                if ((m->m_flags & M_EXT) == 0) {
                   2154:                        m_free(m);
                   2155:                        return NULL;
                   2156:                }
1.1       cgd      2157:        }
1.176     plunky   2158:
                   2159:        memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size);
                   2160:        m->m_len = sopt->sopt_size;
1.160     ad       2161:
1.171     plunky   2162:        return m;
1.1       cgd      2163: }
                   2164:
1.14      mycroft  2165: void
1.54      lukem    2166: sohasoutofband(struct socket *so)
1.1       cgd      2167: {
1.153     rmind    2168:
1.90      christos 2169:        fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.189     ad       2170:        selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT);
1.1       cgd      2171: }
1.72      jdolecek 2172:
                   2173: static void
                   2174: filt_sordetach(struct knote *kn)
                   2175: {
                   2176:        struct socket   *so;
                   2177:
1.155     ad       2178:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2179:        solock(so);
1.73      christos 2180:        SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2181:        if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72      jdolecek 2182:                so->so_rcv.sb_flags &= ~SB_KNOTE;
1.160     ad       2183:        sounlock(so);
1.72      jdolecek 2184: }
                   2185:
                   2186: /*ARGSUSED*/
                   2187: static int
1.129     yamt     2188: filt_soread(struct knote *kn, long hint)
1.72      jdolecek 2189: {
                   2190:        struct socket   *so;
1.160     ad       2191:        int rv;
1.72      jdolecek 2192:
1.155     ad       2193:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2194:        if (hint != NOTE_SUBMIT)
                   2195:                solock(so);
1.72      jdolecek 2196:        kn->kn_data = so->so_rcv.sb_cc;
                   2197:        if (so->so_state & SS_CANTRCVMORE) {
1.108     perry    2198:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2199:                kn->kn_fflags = so->so_error;
1.160     ad       2200:                rv = 1;
                   2201:        } else if (so->so_error)        /* temporary udp error */
                   2202:                rv = 1;
                   2203:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2204:                rv = (kn->kn_data >= kn->kn_sdata);
                   2205:        else
                   2206:                rv = (kn->kn_data >= so->so_rcv.sb_lowat);
                   2207:        if (hint != NOTE_SUBMIT)
                   2208:                sounlock(so);
                   2209:        return rv;
1.72      jdolecek 2210: }
                   2211:
                   2212: static void
                   2213: filt_sowdetach(struct knote *kn)
                   2214: {
                   2215:        struct socket   *so;
                   2216:
1.155     ad       2217:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2218:        solock(so);
1.73      christos 2219:        SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
                   2220:        if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72      jdolecek 2221:                so->so_snd.sb_flags &= ~SB_KNOTE;
1.160     ad       2222:        sounlock(so);
1.72      jdolecek 2223: }
                   2224:
                   2225: /*ARGSUSED*/
                   2226: static int
1.129     yamt     2227: filt_sowrite(struct knote *kn, long hint)
1.72      jdolecek 2228: {
                   2229:        struct socket   *so;
1.160     ad       2230:        int rv;
1.72      jdolecek 2231:
1.155     ad       2232:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2233:        if (hint != NOTE_SUBMIT)
                   2234:                solock(so);
1.72      jdolecek 2235:        kn->kn_data = sbspace(&so->so_snd);
                   2236:        if (so->so_state & SS_CANTSENDMORE) {
1.108     perry    2237:                kn->kn_flags |= EV_EOF;
1.72      jdolecek 2238:                kn->kn_fflags = so->so_error;
1.160     ad       2239:                rv = 1;
                   2240:        } else if (so->so_error)        /* temporary udp error */
                   2241:                rv = 1;
                   2242:        else if (((so->so_state & SS_ISCONNECTED) == 0) &&
1.72      jdolecek 2243:            (so->so_proto->pr_flags & PR_CONNREQUIRED))
1.160     ad       2244:                rv = 0;
                   2245:        else if (kn->kn_sfflags & NOTE_LOWAT)
                   2246:                rv = (kn->kn_data >= kn->kn_sdata);
                   2247:        else
                   2248:                rv = (kn->kn_data >= so->so_snd.sb_lowat);
                   2249:        if (hint != NOTE_SUBMIT)
                   2250:                sounlock(so);
                   2251:        return rv;
1.72      jdolecek 2252: }
                   2253:
                   2254: /*ARGSUSED*/
                   2255: static int
1.129     yamt     2256: filt_solisten(struct knote *kn, long hint)
1.72      jdolecek 2257: {
                   2258:        struct socket   *so;
1.160     ad       2259:        int rv;
1.72      jdolecek 2260:
1.155     ad       2261:        so = ((file_t *)kn->kn_obj)->f_data;
1.72      jdolecek 2262:
                   2263:        /*
                   2264:         * Set kn_data to number of incoming connections, not
                   2265:         * counting partial (incomplete) connections.
1.108     perry    2266:         */
1.160     ad       2267:        if (hint != NOTE_SUBMIT)
                   2268:                solock(so);
1.72      jdolecek 2269:        kn->kn_data = so->so_qlen;
1.160     ad       2270:        rv = (kn->kn_data > 0);
                   2271:        if (hint != NOTE_SUBMIT)
                   2272:                sounlock(so);
                   2273:        return rv;
1.72      jdolecek 2274: }
                   2275:
                   2276: static const struct filterops solisten_filtops =
                   2277:        { 1, NULL, filt_sordetach, filt_solisten };
                   2278: static const struct filterops soread_filtops =
                   2279:        { 1, NULL, filt_sordetach, filt_soread };
                   2280: static const struct filterops sowrite_filtops =
                   2281:        { 1, NULL, filt_sowdetach, filt_sowrite };
                   2282:
                   2283: int
1.129     yamt     2284: soo_kqfilter(struct file *fp, struct knote *kn)
1.72      jdolecek 2285: {
                   2286:        struct socket   *so;
                   2287:        struct sockbuf  *sb;
                   2288:
1.155     ad       2289:        so = ((file_t *)kn->kn_obj)->f_data;
1.160     ad       2290:        solock(so);
1.72      jdolecek 2291:        switch (kn->kn_filter) {
                   2292:        case EVFILT_READ:
                   2293:                if (so->so_options & SO_ACCEPTCONN)
                   2294:                        kn->kn_fop = &solisten_filtops;
                   2295:                else
                   2296:                        kn->kn_fop = &soread_filtops;
                   2297:                sb = &so->so_rcv;
                   2298:                break;
                   2299:        case EVFILT_WRITE:
                   2300:                kn->kn_fop = &sowrite_filtops;
                   2301:                sb = &so->so_snd;
                   2302:                break;
                   2303:        default:
1.160     ad       2304:                sounlock(so);
1.149     pooka    2305:                return (EINVAL);
1.72      jdolecek 2306:        }
1.73      christos 2307:        SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72      jdolecek 2308:        sb->sb_flags |= SB_KNOTE;
1.160     ad       2309:        sounlock(so);
1.72      jdolecek 2310:        return (0);
                   2311: }
                   2312:
1.154     ad       2313: static int
                   2314: sodopoll(struct socket *so, int events)
                   2315: {
                   2316:        int revents;
                   2317:
                   2318:        revents = 0;
                   2319:
                   2320:        if (events & (POLLIN | POLLRDNORM))
                   2321:                if (soreadable(so))
                   2322:                        revents |= events & (POLLIN | POLLRDNORM);
                   2323:
                   2324:        if (events & (POLLOUT | POLLWRNORM))
                   2325:                if (sowritable(so))
                   2326:                        revents |= events & (POLLOUT | POLLWRNORM);
                   2327:
                   2328:        if (events & (POLLPRI | POLLRDBAND))
                   2329:                if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
                   2330:                        revents |= events & (POLLPRI | POLLRDBAND);
                   2331:
                   2332:        return revents;
                   2333: }
                   2334:
                   2335: int
                   2336: sopoll(struct socket *so, int events)
                   2337: {
                   2338:        int revents = 0;
                   2339:
1.160     ad       2340: #ifndef DIAGNOSTIC
                   2341:        /*
                   2342:         * Do a quick, unlocked check in expectation that the socket
                   2343:         * will be ready for I/O.  Don't do this check if DIAGNOSTIC,
                   2344:         * as the solocked() assertions will fail.
                   2345:         */
1.154     ad       2346:        if ((revents = sodopoll(so, events)) != 0)
                   2347:                return revents;
1.160     ad       2348: #endif
1.154     ad       2349:
1.160     ad       2350:        solock(so);
1.154     ad       2351:        if ((revents = sodopoll(so, events)) == 0) {
                   2352:                if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
                   2353:                        selrecord(curlwp, &so->so_rcv.sb_sel);
1.160     ad       2354:                        so->so_rcv.sb_flags |= SB_NOTIFY;
1.154     ad       2355:                }
                   2356:
                   2357:                if (events & (POLLOUT | POLLWRNORM)) {
                   2358:                        selrecord(curlwp, &so->so_snd.sb_sel);
1.160     ad       2359:                        so->so_snd.sb_flags |= SB_NOTIFY;
1.154     ad       2360:                }
                   2361:        }
1.160     ad       2362:        sounlock(so);
1.154     ad       2363:
                   2364:        return revents;
                   2365: }
                   2366:
                   2367:
1.94      yamt     2368: #include <sys/sysctl.h>
                   2369:
                   2370: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
                   2371:
                   2372: /*
                   2373:  * sysctl helper routine for kern.somaxkva.  ensures that the given
                   2374:  * value is not too small.
                   2375:  * (XXX should we maybe make sure it's not too large as well?)
                   2376:  */
                   2377: static int
                   2378: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
                   2379: {
                   2380:        int error, new_somaxkva;
                   2381:        struct sysctlnode node;
                   2382:
                   2383:        new_somaxkva = somaxkva;
                   2384:        node = *rnode;
                   2385:        node.sysctl_data = &new_somaxkva;
                   2386:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                   2387:        if (error || newp == NULL)
                   2388:                return (error);
                   2389:
                   2390:        if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
                   2391:                return (EINVAL);
                   2392:
1.136     ad       2393:        mutex_enter(&so_pendfree_lock);
1.94      yamt     2394:        somaxkva = new_somaxkva;
1.136     ad       2395:        cv_broadcast(&socurkva_cv);
                   2396:        mutex_exit(&so_pendfree_lock);
1.94      yamt     2397:
                   2398:        return (error);
                   2399: }
                   2400:
1.178     pooka    2401: static void
1.187     cegger   2402: sysctl_kern_somaxkva_setup(void)
1.94      yamt     2403: {
                   2404:
1.178     pooka    2405:        KASSERT(socket_sysctllog == NULL);
                   2406:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2407:                       CTLFLAG_PERMANENT,
                   2408:                       CTLTYPE_NODE, "kern", NULL,
                   2409:                       NULL, 0, NULL, 0,
                   2410:                       CTL_KERN, CTL_EOL);
                   2411:
1.178     pooka    2412:        sysctl_createv(&socket_sysctllog, 0, NULL, NULL,
1.97      atatat   2413:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103     atatat   2414:                       CTLTYPE_INT, "somaxkva",
                   2415:                       SYSCTL_DESCR("Maximum amount of kernel memory to be "
                   2416:                                    "used for socket buffers"),
1.94      yamt     2417:                       sysctl_kern_somaxkva, 0, NULL, 0,
                   2418:                       CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
                   2419: }

CVSweb <webmaster@jp.NetBSD.org>