X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/e3027f41d0120b4278cca462f397b6619dcd9ac5..5d5c5d0d5b79ade9a973d55186ffda2638ba2b6e:/bsd/kern/uipc_socket2.c diff --git a/bsd/kern/uipc_socket2.c b/bsd/kern/uipc_socket2.c index 2b61d07fd..2a47db5d1 100644 --- a/bsd/kern/uipc_socket2.c +++ b/bsd/kern/uipc_socket2.c @@ -1,23 +1,31 @@ /* * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ + * @APPLE_LICENSE_OSREFERENCE_HEADER_START@ * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. - * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER - * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, - * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. - * - * @APPLE_LICENSE_HEADER_END@ + * This file contains Original Code and/or Modifications of Original Code + * as defined in and that are subject to the Apple Public Source License + * Version 2.0 (the 'License'). You may not use this file except in + * compliance with the License. The rights granted to you under the + * License may not be used to create, or enable the creation or + * redistribution of, unlawful or unlicensed copies of an Apple operating + * system, or to circumvent, violate, or enable the circumvention or + * violation of, any terms of an Apple operating system software license + * agreement. + * + * Please obtain a copy of the License at + * http://www.opensource.apple.com/apsl/ and read it before using this + * file. + * + * The Original Code and all software distributed under the License are + * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, + * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. + * + * @APPLE_LICENSE_OSREFERENCE_HEADER_END@ */ /* Copyright (c) 1998, 1999 Apple Computer, Inc. All Rights Reserved */ /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */ @@ -54,13 +62,15 @@ * SUCH DAMAGE. * * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93 + * $FreeBSD: src/sys/kern/uipc_socket2.c,v 1.55.2.9 2001/07/26 18:53:02 peter Exp $ */ #include #include #include #include -#include +#include +#include #include #include #include @@ -70,6 +80,16 @@ #include #include #include +#include +#include +#include +#include +#include + +#define DBG_FNC_SBDROP NETDBG_CODE(DBG_NETSOCK, 4) +#define DBG_FNC_SBAPPEND NETDBG_CODE(DBG_NETSOCK, 5) + +static int sbcompress(struct sockbuf *, struct mbuf *, struct mbuf *); /* * Primitive routines for operating on sockets and socket buffers @@ -79,8 +99,6 @@ u_long sb_max = SB_MAX; /* XXX should be static */ static u_long sb_efficiency = 8; /* parameter for sbreserve() */ -char netcon[] = "netcon"; - /* * Procedures to manipulate state flags of socket * and do appropriate wakeups. Normal sequence from the @@ -88,7 +106,7 @@ char netcon[] = "netcon"; * called during processing of connect() call, * resulting in an eventual call to soisconnected() if/when the * connection is established. When the connection is torn down - * soisdisconnecting() is called during processing of disconnect() call, + * soisdisconnecting() is called during processing of disconnect() call, * and soisdisconnected() is called when the connection to the peer * is totally severed. The semantics of these routines are such that * connectionless protocols can call soisconnected() and soisdisconnected() @@ -98,19 +116,18 @@ char netcon[] = "netcon"; * From the passive side, a socket is created with * two queues of sockets: so_incomp for connections in progress * and so_comp for connections already made and awaiting user acceptance. - * As a protocol is preparing incoming connections, it creates a socket + * As a protocol is preparing incoming connections, it creates a socket * structure queued on so_incomp by calling sonewconn(). When the connection * is established, soisconnected() is called, and transfers the * socket structure to so_comp, making it available to accept(). * - * If a socket is closed with sockets on either + * If a socket is closed with sockets on either * so_incomp or so_comp, these sockets are dropped. - * + * * If higher level protocols are implemented in * the kernel, the wakeups done here will sometimes * cause software-interrupt process scheduling. */ - void soisconnecting(so) register struct socket *so; @@ -118,36 +135,40 @@ soisconnecting(so) so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); so->so_state |= SS_ISCONNECTING; + + sflt_notify(so, sock_evt_connecting, NULL); } void soisconnected(so) - register struct socket *so; -{ register struct kextcb *kp; - register struct socket *head = so->so_head; - - kp = sotokextcb(so); - while (kp) - { if (kp->e_soif && kp->e_soif->sf_soisconnected) - { if ((*kp->e_soif->sf_soisconnected)(so, kp)) - return; - } - kp = kp->e_next; - } + struct socket *so; +{ + struct socket *head = so->so_head; so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING); so->so_state |= SS_ISCONNECTED; + + sflt_notify(so, sock_evt_connected, NULL); + if (head && (so->so_state & SS_INCOMP)) { - postevent(head,0,EV_RCONN); + so->so_state &= ~SS_INCOMP; + so->so_state |= SS_COMP; + if (head->so_proto->pr_getlock != NULL) { + socket_unlock(so, 0); + socket_lock(head, 1); + } + postevent(head, 0, EV_RCONN); TAILQ_REMOVE(&head->so_incomp, so, so_list); head->so_incqlen--; - so->so_state &= ~SS_INCOMP; TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); - so->so_state |= SS_COMP; sorwakeup(head); - wakeup((caddr_t)&head->so_timeo); + wakeup_one((caddr_t)&head->so_timeo); + if (head->so_proto->pr_getlock != NULL) { + socket_unlock(head, 1); + socket_lock(so, 0); + } } else { - postevent(so,0,EV_WCONN); + postevent(so, 0, EV_WCONN); wakeup((caddr_t)&so->so_timeo); sorwakeup(so); sowwakeup(so); @@ -157,19 +178,10 @@ soisconnected(so) void soisdisconnecting(so) register struct socket *so; -{ register struct kextcb *kp; - - kp = sotokextcb(so); - while (kp) - { if (kp->e_soif && kp->e_soif->sf_soisdisconnecting) - { if ((*kp->e_soif->sf_soisdisconnecting)(so, kp)) - return; - } - kp = kp->e_next; - } - +{ so->so_state &= ~SS_ISCONNECTING; so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); + sflt_notify(so, sock_evt_disconnecting, NULL); wakeup((caddr_t)&so->so_timeo); sowwakeup(so); sorwakeup(so); @@ -178,19 +190,10 @@ soisdisconnecting(so) void soisdisconnected(so) register struct socket *so; -{ register struct kextcb *kp; - - kp = sotokextcb(so); - while (kp) - { if (kp->e_soif && kp->e_soif->sf_soisdisconnected) - { if ((*kp->e_soif->sf_soisdisconnected)(so, kp)) - return; - } - kp = kp->e_next; - } - +{ so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); - so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE); + so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED); + sflt_notify(so, sock_evt_disconnected, NULL); wakeup((caddr_t)&so->so_timeo); sowwakeup(so); sorwakeup(so); @@ -210,7 +213,7 @@ struct socket * sodropablereq(head) register struct socket *head; { - register struct socket *so; + struct socket *so, *sonext = NULL; unsigned int i, j, qlen; static int rnd; static struct timeval old_runtime; @@ -226,18 +229,27 @@ sodropablereq(head) so = TAILQ_FIRST(&head->so_incomp); if (!so) - return (so); + return (NULL); qlen = head->so_incqlen; if (++cur_cnt > qlen || old_cnt > qlen) { rnd = (314159 * rnd + 66329) & 0xffff; j = ((qlen + 1) * rnd) >> 16; - - while (j-- && so) - so = TAILQ_NEXT(so, so_list); +//###LD To clean up + while (j-- && so) { +// if (in_pcb_checkstate(so->so_pcb, WNT_ACQUIRE, 0) != WNT_STOPUSING) { + socket_lock(so, 1); + sonext = TAILQ_NEXT(so, so_list); +// in_pcb_check_state(so->so_pcb, WNT_RELEASE, 0); + socket_unlock(so, 1); + so = sonext; + } } - return (so); +// if (in_pcb_checkstate(so->so_pcb, WNT_ACQUIRE, 0) == WNT_STOPUSING) +// return (NULL); +// else + return (so); } /* @@ -248,27 +260,30 @@ sodropablereq(head) * data structure of the original socket, and return this. * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. */ -struct socket * -sonewconn(head, connstatus) +static struct socket * +sonewconn_internal(head, connstatus) register struct socket *head; int connstatus; -{ int error = 0; +{ + int error = 0; register struct socket *so; - register struct kextcb *kp; + lck_mtx_t *mutex_held; + + if (head->so_proto->pr_getlock != NULL) + mutex_held = (*head->so_proto->pr_getlock)(head, 0); + else + mutex_held = head->so_proto->pr_domain->dom_mtx; + lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); if (head->so_qlen > 3 * head->so_qlimit / 2) return ((struct socket *)0); - so = soalloc(0, head->so_proto->pr_domain->dom_family, head->so_type); + so = soalloc(1, head->so_proto->pr_domain->dom_family, head->so_type); if (so == NULL) return ((struct socket *)0); - - kp = sotokextcb(so); - while (kp) - { if (kp->e_soif && kp->e_soif->sf_sonewconn1) - { if ((*kp->e_soif->sf_sonewconn1)(so, connstatus, kp)) - return; - } - kp = kp->e_next; + /* check if head was closed during the soalloc */ + if (head->so_proto == NULL) { + sodealloc(so); + return ((struct socket *)0); } so->so_head = head; @@ -280,17 +295,39 @@ sonewconn(head, connstatus) so->so_timeo = head->so_timeo; so->so_pgid = head->so_pgid; so->so_uid = head->so_uid; + so->so_usecount = 1; + so->next_lock_lr = 0; + so->next_unlock_lr = 0; + +#ifdef __APPLE__ so->so_rcv.sb_flags |= SB_RECV; /* XXX */ - (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat); + so->so_rcv.sb_so = so->so_snd.sb_so = so; + TAILQ_INIT(&so->so_evlist); +#endif - if (so->so_proto->pr_sfilter.tqh_first) - error = sfilter_init(so); - if (error == 0 && (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) { - sfilter_term(so); + if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat)) { + sflt_termsock(so); sodealloc(so); return ((struct socket *)0); } + + /* + * Must be done with head unlocked to avoid deadlock for protocol with per socket mutexes. + */ + if (head->so_proto->pr_unlock) + socket_unlock(head, 0); + if (((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL) != 0) || error) { + sflt_termsock(so); + sodealloc(so); + if (head->so_proto->pr_unlock) + socket_lock(head, 0); + return ((struct socket *)0); + } + if (head->so_proto->pr_unlock) + socket_lock(head, 0); +#ifdef __APPLE__ so->so_proto->pr_domain->dom_refs++; +#endif if (connstatus) { TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); @@ -301,16 +338,55 @@ sonewconn(head, connstatus) head->so_incqlen++; } head->so_qlen++; + +#ifdef __APPLE__ + /* Attach socket filters for this protocol */ + sflt_initsock(so); +#endif if (connstatus) { + so->so_state |= connstatus; sorwakeup(head); wakeup((caddr_t)&head->so_timeo); - so->so_state |= connstatus; } - so->so_rcv.sb_so = so->so_snd.sb_so = so; - TAILQ_INIT(&so->so_evlist); return (so); } + +struct socket * +sonewconn( + struct socket *head, + int connstatus, + const struct sockaddr *from) +{ + int error = 0; + struct socket_filter_entry *filter; + int filtered = 0; + + error = 0; + for (filter = head->so_filt; filter && (error == 0); + filter = filter->sfe_next_onsocket) { + if (filter->sfe_filter->sf_filter.sf_connect_in) { + if (filtered == 0) { + filtered = 1; + sflt_use(head); + socket_unlock(head, 0); + } + error = filter->sfe_filter->sf_filter.sf_connect_in( + filter->sfe_cookie, head, from); + } + } + if (filtered != 0) { + socket_lock(head, 0); + sflt_unuse(head); + } + + if (error) { + return NULL; + } + + return sonewconn_internal(head, connstatus); +} + /* * Socantsendmore indicates that no more data will be sent on the * socket; it would normally be applied to a socket when the user @@ -324,38 +400,18 @@ sonewconn(head, connstatus) void socantsendmore(so) struct socket *so; -{ register struct kextcb *kp; - - kp = sotokextcb(so); - while (kp) - { if (kp->e_soif && kp->e_soif->sf_socantsendmore) - { if ((*kp->e_soif->sf_socantsendmore)(so, kp)) - return; - } - kp = kp->e_next; - } - - +{ so->so_state |= SS_CANTSENDMORE; + sflt_notify(so, sock_evt_cantsendmore, NULL); sowwakeup(so); } void socantrcvmore(so) struct socket *so; -{ register struct kextcb *kp; - - kp = sotokextcb(so); - while (kp) - { if (kp->e_soif && kp->e_soif->sf_socantrcvmore) - { if ((*kp->e_soif->sf_socantrcvmore)(so, kp)) - return; - } - kp = kp->e_next; - } - - +{ so->so_state |= SS_CANTRCVMORE; + sflt_notify(so, sock_evt_cantrecvmore, NULL); sorwakeup(so); } @@ -366,11 +422,38 @@ int sbwait(sb) struct sockbuf *sb; { + int error = 0, lr_saved; + struct socket *so = sb->sb_so; + lck_mtx_t *mutex_held; + struct timespec ts; + + lr_saved = (unsigned int) __builtin_return_address(0); + + if (so->so_proto->pr_getlock != NULL) + mutex_held = (*so->so_proto->pr_getlock)(so, 0); + else + mutex_held = so->so_proto->pr_domain->dom_mtx; sb->sb_flags |= SB_WAIT; - return (tsleep((caddr_t)&sb->sb_cc, - (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait", - sb->sb_timeo)); + + if (so->so_usecount < 1) + panic("sbwait: so=%x refcount=%d\n", so, so->so_usecount); + ts.tv_sec = sb->sb_timeo.tv_sec; + ts.tv_nsec = sb->sb_timeo.tv_usec * 1000; + error = msleep((caddr_t)&sb->sb_cc, mutex_held, + (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait", + &ts); + + lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); + + if (so->so_usecount < 1) + panic("sbwait: so=%x refcount=%d\n", so, so->so_usecount); + + if ((so->so_state & SS_DRAINING)) { + error = EBADF; + } + + return (error); } /* @@ -381,14 +464,27 @@ int sb_lock(sb) register struct sockbuf *sb; { - int error; + struct socket *so = sb->sb_so; + lck_mtx_t * mutex_held; + int error = 0; + + if (so == NULL) + panic("sb_lock: null so back pointer sb=%x\n", sb); while (sb->sb_flags & SB_LOCK) { sb->sb_flags |= SB_WANT; - error = tsleep((caddr_t)&sb->sb_flags, - (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH, - "sblock", 0); - if (error) + if (so->so_proto->pr_getlock != NULL) + mutex_held = (*so->so_proto->pr_getlock)(so, 0); + else + mutex_held = so->so_proto->pr_domain->dom_mtx; + if (so->so_usecount < 1) + panic("sb_lock: so=%x refcount=%d\n", so, so->so_usecount); + + error = msleep((caddr_t)&sb->sb_flags, mutex_held, + (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sblock", 0); + if (so->so_usecount < 1) + panic("sb_lock: 2 so=%x refcount=%d\n", so, so->so_usecount); + if (error) return (error); } sb->sb_flags |= SB_LOCK; @@ -406,14 +502,8 @@ sowakeup(so, sb) register struct sockbuf *sb; { struct proc *p = current_proc(); - - - - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); - sb->sb_sel.si_flags &= ~SI_SBSEL; + sb->sb_flags &= ~SB_SEL; selwakeup(&sb->sb_sel); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); - if (sb->sb_flags & SB_WAIT) { sb->sb_flags &= ~SB_WAIT; wakeup((caddr_t)&sb->sb_cc); @@ -424,9 +514,14 @@ sowakeup(so, sb) else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) psignal(p, SIGIO); } - - if (sb->sb_flags & SB_UPCALL) + if (sb->sb_flags & SB_KNOTE) { + KNOTE(&sb->sb_sel.si_note, SO_FILT_HINT_LOCKED); + } + if (sb->sb_flags & SB_UPCALL) { + socket_unlock(so, 0); (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT); + socket_lock(so, 0); + } } /* @@ -466,16 +561,6 @@ soreserve(so, sndcc, rcvcc) register struct socket *so; u_long sndcc, rcvcc; { - register struct kextcb *kp; - - kp = sotokextcb(so); - while (kp) - { if (kp->e_soif && kp->e_soif->sf_soreserve) - { if ((*kp->e_soif->sf_soreserve)(so, sndcc, rcvcc, kp)) - return; - } - kp = kp->e_next; - } if (sbreserve(&so->so_snd, sndcc) == 0) goto bad; @@ -489,6 +574,9 @@ soreserve(so, sndcc, rcvcc) so->so_snd.sb_lowat = so->so_snd.sb_hiwat; return (0); bad2: +#ifdef __APPLE__ + selthreadclear(&so->so_snd.sb_sel); +#endif sbrelease(&so->so_snd); bad: return (ENOBUFS); @@ -516,19 +604,16 @@ sbreserve(sb, cc) /* * Free mbufs held by a socket, and reserved mbuf space. */ + /* WARNING needs to do selthreadclear() before calling this */ void sbrelease(sb) struct sockbuf *sb; { sbflush(sb); - sb->sb_hiwat = sb->sb_mbmax = 0; - - { - int oldpri = splimp(); - selthreadclear(&sb->sb_sel); - splx(oldpri); - } + sb->sb_hiwat = 0; + sb->sb_mbmax = 0; + } /* @@ -562,35 +647,59 @@ sbrelease(sb) * the mbuf chain is recorded in sb. Empty mbufs are * discarded and mbufs are compacted where possible. */ -void +int sbappend(sb, m) struct sockbuf *sb; struct mbuf *m; -{ register struct kextcb *kp; - register struct mbuf *n; +{ + register struct mbuf *n, *sb_first; + int result = 0; + int error = 0; + int filtered = 0; - if (m == 0) - return; - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbappend) - { if ((*kp->e_sout->su_sbappend)(sb, m, kp)) - return; - } - kp = kp->e_next; - } - if (n = sb->sb_mb) { + KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_START), sb, m->m_len, 0, 0, 0); + + if (m == 0) + return 0; + +again: + sb_first = n = sb->sb_mb; + if (n) { while (n->m_nextpkt) n = n->m_nextpkt; do { if (n->m_flags & M_EOR) { - sbappendrecord(sb, m); /* XXXXXX!!!! */ - return; + result = sbappendrecord(sb, m); /* XXXXXX!!!! */ + KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_END), sb, sb->sb_cc, 0, 0, 0); + return result; } } while (n->m_next && (n = n->m_next)); } - sbcompress(sb, m, n); + + if (!filtered && (sb->sb_flags & SB_RECV) != 0) { + error = sflt_data_in(sb->sb_so, NULL, &m, NULL, 0, &filtered); + if (error) { + /* no data was appended, caller should not call sowakeup */ + return 0; + } + + /* + If we any filters, the socket lock was dropped. n and sb_first + cached data from the socket buffer. This cache is not valid + since we dropped the lock. We must start over. Since filtered + is set we won't run through the filters a second time. We just + set n and sb_start again. + */ + if (filtered) + goto again; + } + + result = sbcompress(sb, m, n); + + KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_END), sb, sb->sb_cc, 0, 0, 0); + + return result; } #ifdef SOCKBUF_DEBUG @@ -601,22 +710,31 @@ sbcheck(sb) register struct mbuf *m; register struct mbuf *n = 0; register u_long len = 0, mbcnt = 0; + lck_mtx_t *mutex_held; + + if (sb->sb_so->so_proto->pr_getlock != NULL) + mutex_held = (*sb->sb_so->so_proto->pr_getlock)(sb->sb_so, 0); + else + mutex_held = sb->sb_so->so_proto->pr_domain->dom_mtx; + + lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); + + if (sbchecking == 0) + return; for (m = sb->sb_mb; m; m = n) { n = m->m_nextpkt; for (; m; m = m->m_next) { - len += m->m_len; - mbcnt += MSIZE; - if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */ - mbcnt += m->m_ext.ext_size; - if (m->m_nextpkt) - panic("sbcheck nextpkt"); - } - if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) { - printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc, - mbcnt, sb->sb_mbcnt); - panic("sbcheck"); + len += m->m_len; + mbcnt += MSIZE; + if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */ + mbcnt += m->m_ext.ext_size; + } } + if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) { + panic("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc, + mbcnt, sb->sb_mbcnt); + } } #endif @@ -624,24 +742,24 @@ sbcheck(sb) * As above, except the mbuf chain * begins a new record. */ -void +int sbappendrecord(sb, m0) register struct sockbuf *sb; - register struct mbuf *m0; + struct mbuf *m0; { register struct mbuf *m; - register struct kextcb *kp; - + int result = 0; + if (m0 == 0) - return; - - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbappendrecord) - { if ((*kp->e_sout->su_sbappendrecord)(sb, m0, kp)) - return; + return 0; + + if ((sb->sb_flags & SB_RECV) != 0) { + int error = sflt_data_in(sb->sb_so, NULL, &m0, NULL, sock_data_filt_flag_record, NULL); + if (error != 0) { + if (error != EJUSTRETURN) + m_freem(m0); + return 0; } - kp = kp->e_next; } m = sb->sb_mb; @@ -663,7 +781,7 @@ sbappendrecord(sb, m0) m0->m_flags &= ~M_EOR; m->m_flags |= M_EOR; } - sbcompress(sb, m, m0); + return sbcompress(sb, m, m0); } /* @@ -671,25 +789,27 @@ sbappendrecord(sb, m0) * is inserted at the beginning of the sockbuf, * but after any other OOB data. */ -void +int sbinsertoob(sb, m0) - register struct sockbuf *sb; - register struct mbuf *m0; + struct sockbuf *sb; + struct mbuf *m0; { - register struct mbuf *m; - register struct mbuf **mp; - register struct kextcb *kp; + struct mbuf *m; + struct mbuf **mp; if (m0 == 0) - return; - - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbinsertoob) - { if ((*kp->e_sout->su_sbinsertoob)(sb, m0, kp)) - return; + return 0; + + if ((sb->sb_flags & SB_RECV) != 0) { + int error = sflt_data_in(sb->sb_so, NULL, &m0, NULL, + sock_data_filt_flag_oob, NULL); + + if (error) { + if (error != EJUSTRETURN) { + m_freem(m0); + } + return 0; } - kp = kp->e_next; } for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) { @@ -720,7 +840,7 @@ sbinsertoob(sb, m0) m0->m_flags &= ~M_EOR; m->m_flags |= M_EOR; } - sbcompress(sb, m, m0); + return sbcompress(sb, m, m0); } /* @@ -729,28 +849,18 @@ sbinsertoob(sb, m0) * m0 must include a packet header with total length. * Returns 0 if no space in sockbuf or insufficient mbufs. */ -int -sbappendaddr(sb, asa, m0, control) +static int +sbappendaddr_internal(sb, asa, m0, control) register struct sockbuf *sb; struct sockaddr *asa; struct mbuf *m0, *control; { register struct mbuf *m, *n; int space = asa->sa_len; - register struct kextcb *kp; if (m0 && (m0->m_flags & M_PKTHDR) == 0) panic("sbappendaddr"); - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbappendaddr) - { if ((*kp->e_sout->su_sbappendaddr)(sb, asa, m0, control, kp)) - return 0; - } - kp = kp->e_next; - } - if (m0) space += m0->m_pkthdr.len; for (n = control; n; n = n->m_next) { @@ -786,26 +896,55 @@ sbappendaddr(sb, asa, m0, control) } int -sbappendcontrol(sb, m0, control) +sbappendaddr( + struct sockbuf* sb, + struct sockaddr* asa, + struct mbuf *m0, + struct mbuf *control, + int *error_out) +{ + int result = 0; + + if (error_out) *error_out = 0; + + if (m0 && (m0->m_flags & M_PKTHDR) == 0) + panic("sbappendaddrorfree"); + + /* Call socket data in filters */ + if ((sb->sb_flags & SB_RECV) != 0) { + int error; + error = sflt_data_in(sb->sb_so, asa, &m0, &control, 0, NULL); + if (error) { + if (error != EJUSTRETURN) { + if (m0) m_freem(m0); + if (control) m_freem(control); + if (error_out) *error_out = error; + } + return 0; + } + } + + result = sbappendaddr_internal(sb, asa, m0, control); + if (result == 0) { + if (m0) m_freem(m0); + if (control) m_freem(control); + if (error_out) *error_out = ENOBUFS; + } + + return result; +} + +static int +sbappendcontrol_internal(sb, m0, control) struct sockbuf *sb; struct mbuf *control, *m0; { register struct mbuf *m, *n; int space = 0; - register struct kextcb *kp; if (control == 0) panic("sbappendcontrol"); - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbappendcontrol) - { if ((*kp->e_sout->su_sbappendcontrol)(sb, m0, control, kp)) - return 0; - } - kp = kp->e_next; - } - for (m = control; ; m = m->m_next) { space += m->m_len; if (m->m_next == 0) @@ -830,12 +969,46 @@ sbappendcontrol(sb, m0, control) return (1); } +int +sbappendcontrol( + struct sockbuf *sb, + struct mbuf *m0, + struct mbuf *control, + int *error_out) +{ + int result = 0; + + if (error_out) *error_out = 0; + + if (sb->sb_flags & SB_RECV) { + int error; + error = sflt_data_in(sb->sb_so, NULL, &m0, &control, 0, NULL); + if (error) { + if (error != EJUSTRETURN) { + if (m0) m_freem(m0); + if (control) m_freem(control); + if (error_out) *error_out = error; + } + return 0; + } + } + + result = sbappendcontrol_internal(sb, m0, control); + if (result == 0) { + if (m0) m_freem(m0); + if (control) m_freem(control); + if (error_out) *error_out = ENOBUFS; + } + + return result; +} + /* * Compress mbuf chain m into the socket * buffer sb following mbuf n. If n * is null, the buffer is presumed empty. */ -void +static int sbcompress(sb, m, n) register struct sockbuf *sb; register struct mbuf *m, *n; @@ -852,8 +1025,12 @@ sbcompress(sb, m, n) m = m_free(m); continue; } - if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 && - (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] && + if (n && (n->m_flags & M_EOR) == 0 && +#ifndef __APPLE__ + M_WRITABLE(n) && +#endif + m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */ + m->m_len <= M_TRAILINGSPACE(n) && n->m_type == m->m_type) { bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len, (unsigned)m->m_len); @@ -879,6 +1056,7 @@ sbcompress(sb, m, n) printf("semi-panic: sbcompress\n"); } postevent(0,sb, EV_RWBYTES); + return 1; } /* @@ -889,55 +1067,70 @@ void sbflush(sb) register struct sockbuf *sb; { - register struct kextcb *kp; - - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbflush) - { if ((*kp->e_sout->su_sbflush)(sb, kp)) - return; - } - kp = kp->e_next; + if (sb->sb_so == NULL) + panic ("sbflush sb->sb_so already null sb=%x\n", sb); + (void)sblock(sb, M_WAIT); + while (sb->sb_mbcnt) { + /* + * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty: + * we would loop forever. Panic instead. + */ + if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len)) + break; + sbdrop(sb, (int)sb->sb_cc); } + if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt || sb->sb_so == NULL) + panic("sbflush: cc %ld || mb %p || mbcnt %ld sb_so=%x", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt, sb->sb_so); - if (sb->sb_flags & SB_LOCK) - panic("sbflush: locked"); - while (sb->sb_mbcnt && sb->sb_cc) - sbdrop(sb, (int)sb->sb_cc); - if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt) - panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt); postevent(0, sb, EV_RWBYTES); + sbunlock(sb, 1); /* keep socket locked */ + } /* * Drop data from (the front of) a sockbuf. + * use m_freem_list to free the mbuf structures + * under a single lock... this is done by pruning + * the top of the tree from the body by keeping track + * of where we get to in the tree and then zeroing the + * two pertinent pointers m_nextpkt and m_next + * the socket buffer is then updated to point at the new + * top of the tree and the pruned area is released via + * m_freem_list. */ void sbdrop(sb, len) register struct sockbuf *sb; register int len; { - register struct mbuf *m, *mn; - struct mbuf *next; - register struct kextcb *kp; - - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbdrop) - { if ((*kp->e_sout->su_sbdrop)(sb, len, kp)) - return; - } - kp = kp->e_next; - } + register struct mbuf *m, *free_list, *ml; + struct mbuf *next, *last; + + KERNEL_DEBUG((DBG_FNC_SBDROP | DBG_FUNC_START), sb, len, 0, 0, 0); next = (m = sb->sb_mb) ? m->m_nextpkt : 0; + free_list = last = m; + ml = (struct mbuf *)0; + while (len > 0) { if (m == 0) { - if (next == 0) - panic("sbdrop"); - m = next; - next = m->m_nextpkt; - continue; + if (next == 0) { + /* temporarily replacing this panic with printf because + * it occurs occasionally when closing a socket when there + * is no harm in ignoring it. This problem will be investigated + * further. + */ + /* panic("sbdrop"); */ + printf("sbdrop - count not zero\n"); + len = 0; + /* zero the counts. if we have no mbufs, we have no data (PR-2986815) */ + sb->sb_cc = 0; + sb->sb_mbcnt = 0; + break; + } + m = last = next; + next = m->m_nextpkt; + continue; } if (m->m_len > len) { m->m_len -= len; @@ -947,20 +1140,30 @@ sbdrop(sb, len) } len -= m->m_len; sbfree(sb, m); - MFREE(m, mn); - m = mn; + + ml = m; + m = m->m_next; } while (m && m->m_len == 0) { sbfree(sb, m); - MFREE(m, mn); - m = mn; + + ml = m; + m = m->m_next; + } + if (ml) { + ml->m_next = (struct mbuf *)0; + last->m_nextpkt = (struct mbuf *)0; + m_freem_list(free_list); } if (m) { sb->sb_mb = m; m->m_nextpkt = next; } else sb->sb_mb = next; + postevent(0, sb, EV_RWBYTES); + + KERNEL_DEBUG((DBG_FNC_SBDROP | DBG_FUNC_END), sb, 0, 0, 0, 0); } /* @@ -972,16 +1175,6 @@ sbdroprecord(sb) register struct sockbuf *sb; { register struct mbuf *m, *mn; - register struct kextcb *kp; - - kp = sotokextcb(sbtoso(sb)); - while (kp) - { if (kp->e_sout && kp->e_sout->su_sbdroprecord) - { if ((*kp->e_sout->su_sbdroprecord)(sb, kp)) - return; - } - kp = kp->e_next; - } m = sb->sb_mb; if (m) { @@ -989,7 +1182,8 @@ sbdroprecord(sb) do { sbfree(sb, m); MFREE(m, mn); - } while (m = mn); + m = mn; + } while (m); } postevent(0, sb, EV_RWBYTES); } @@ -1007,14 +1201,15 @@ sbcreatecontrol(p, size, type, level) register struct cmsghdr *cp; struct mbuf *m; + if (CMSG_SPACE((u_int)size) > MLEN) + return ((struct mbuf *) NULL); if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL) return ((struct mbuf *) NULL); cp = mtod(m, struct cmsghdr *); /* XXX check size? */ (void)memcpy(CMSG_DATA(cp), p, size); - size += sizeof(*cp); - m->m_len = size; - cp->cmsg_len = size; + m->m_len = CMSG_SPACE(size); + cp->cmsg_len = CMSG_LEN(size); cp->cmsg_level = level; cp->cmsg_type = type; return (m); @@ -1171,13 +1366,156 @@ int pru_soreceive(struct socket *so, } -int pru_sopoll_notsupp(struct socket *so, int events, - struct ucred *cred) +int +pru_sopoll_notsupp(__unused struct socket *so, __unused int events, + __unused kauth_cred_t cred, __unused void *wql) { return EOPNOTSUPP; } +#ifdef __APPLE__ +/* + * The following are macros on BSD and functions on Darwin + */ + +/* + * Do we need to notify the other side when I/O is possible? + */ + +int +sb_notify(struct sockbuf *sb) +{ + return ((sb->sb_flags & (SB_WAIT|SB_SEL|SB_ASYNC|SB_UPCALL|SB_KNOTE)) != 0); +} + +/* + * How much space is there in a socket buffer (so->so_snd or so->so_rcv)? + * This is problematical if the fields are unsigned, as the space might + * still be negative (cc > hiwat or mbcnt > mbmax). Should detect + * overflow and return 0. Should use "lmin" but it doesn't exist now. + */ +long +sbspace(struct sockbuf *sb) +{ + return ((long) imin((int)(sb->sb_hiwat - sb->sb_cc), + (int)(sb->sb_mbmax - sb->sb_mbcnt))); +} + +/* do we have to send all at once on a socket? */ +int +sosendallatonce(struct socket *so) +{ + return (so->so_proto->pr_flags & PR_ATOMIC); +} + +/* can we read something from so? */ +int +soreadable(struct socket *so) +{ + return (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat || + (so->so_state & SS_CANTRCVMORE) || + so->so_comp.tqh_first || so->so_error); +} + +/* can we write something to so? */ + +int +sowriteable(struct socket *so) +{ + return ((sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && + ((so->so_state&SS_ISCONNECTED) || + (so->so_proto->pr_flags&PR_CONNREQUIRED)==0)) || + (so->so_state & SS_CANTSENDMORE) || + so->so_error); +} + +/* adjust counters in sb reflecting allocation of m */ + +void +sballoc(struct sockbuf *sb, struct mbuf *m) +{ + sb->sb_cc += m->m_len; + sb->sb_mbcnt += MSIZE; + if (m->m_flags & M_EXT) + sb->sb_mbcnt += m->m_ext.ext_size; +} + +/* adjust counters in sb reflecting freeing of m */ +void +sbfree(struct sockbuf *sb, struct mbuf *m) +{ + sb->sb_cc -= m->m_len; + sb->sb_mbcnt -= MSIZE; + if (m->m_flags & M_EXT) + sb->sb_mbcnt -= m->m_ext.ext_size; +} + +/* + * Set lock on sockbuf sb; sleep if lock is already held. + * Unless SB_NOINTR is set on sockbuf, sleep is interruptible. + * Returns error without lock if sleep is interrupted. + */ +int +sblock(struct sockbuf *sb, int wf) +{ + return(sb->sb_flags & SB_LOCK ? + ((wf == M_WAIT) ? sb_lock(sb) : EWOULDBLOCK) : + (sb->sb_flags |= SB_LOCK), 0); +} + +/* release lock on sockbuf sb */ +void +sbunlock(struct sockbuf *sb, int keeplocked) +{ + struct socket *so = sb->sb_so; + int lr_saved; + lck_mtx_t *mutex_held; + + + lr_saved = (unsigned int) __builtin_return_address(0); + + sb->sb_flags &= ~SB_LOCK; + + if (so->so_proto->pr_getlock != NULL) + mutex_held = (*so->so_proto->pr_getlock)(so, 0); + else + mutex_held = so->so_proto->pr_domain->dom_mtx; + + if (keeplocked == 0) + lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); + + if (sb->sb_flags & SB_WANT) { + sb->sb_flags &= ~SB_WANT; + if (so->so_usecount < 0) + panic("sbunlock: b4 wakeup so=%x ref=%d lr=%x sb_flags=%x\n", sb->sb_so, so->so_usecount, lr_saved, sb->sb_flags); + + wakeup((caddr_t)&(sb)->sb_flags); + } + if (keeplocked == 0) { /* unlock on exit */ + so->so_usecount--; + if (so->so_usecount < 0) + panic("sbunlock: unlock on exit so=%x lr=%x sb_flags=%x\n", so, so->so_usecount,lr_saved, sb->sb_flags); + so->unlock_lr[so->next_unlock_lr] = (void *)lr_saved; + so->next_unlock_lr = (so->next_unlock_lr+1) % SO_LCKDBG_MAX; + lck_mtx_unlock(mutex_held); + } +} + +void +sorwakeup(struct socket * so) +{ + if (sb_notify(&so->so_rcv)) + sowakeup(so, &so->so_rcv); +} + +void +sowwakeup(struct socket * so) +{ + if (sb_notify(&so->so_snd)) + sowakeup(so, &so->so_snd); +} +#endif __APPLE__ /* * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. @@ -1214,8 +1552,12 @@ sotoxsocket(struct socket *so, struct xsocket *xso) xso->so_linger = so->so_linger; xso->so_state = so->so_state; xso->so_pcb = so->so_pcb; - xso->xso_protocol = so->so_proto->pr_protocol; - xso->xso_family = so->so_proto->pr_domain->dom_family; + if (so->so_proto) { + xso->xso_protocol = so->so_proto->pr_protocol; + xso->xso_family = so->so_proto->pr_domain->dom_family; + } + else + xso->xso_protocol = xso->xso_family = 0; xso->so_qlen = so->so_qlen; xso->so_incqlen = so->so_incqlen; xso->so_qlimit = so->so_qlimit; @@ -1243,23 +1585,25 @@ sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb) xsb->sb_mbmax = sb->sb_mbmax; xsb->sb_lowat = sb->sb_lowat; xsb->sb_flags = sb->sb_flags; - xsb->sb_timeo = sb->sb_timeo; + xsb->sb_timeo = (u_long)(sb->sb_timeo.tv_sec * hz) + sb->sb_timeo.tv_usec / tick; + if (xsb->sb_timeo == 0 && sb->sb_timeo.tv_usec != 0) + xsb->sb_timeo = 1; } /* * Here is the definition of some of the basic objects in the kern.ipc * branch of the MIB. */ - - SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC"); /* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */ static int dummy; SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, ""); -SYSCTL_INT(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, ""); -SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, &maxsockets, 0, ""); +SYSCTL_INT(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, + &sb_max, 0, "Maximum socket buffer size"); +SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, + &maxsockets, 0, "Maximum number of sockets avaliable"); SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW, &sb_efficiency, 0, ""); SYSCTL_INT(_kern_ipc, KIPC_NMBCLUSTERS, nmbclusters, CTLFLAG_RD, &nmbclusters, 0, "");