X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/8f6c56a50524aa785f7e596d52dddfb331e18961..c18c124eaa464aaaa5549e99e5a70fc9cbb50944:/bsd/kern/uipc_socket2.c diff --git a/bsd/kern/uipc_socket2.c b/bsd/kern/uipc_socket2.c index 1b2ced9a9..5cbf06334 100644 --- a/bsd/kern/uipc_socket2.c +++ b/bsd/kern/uipc_socket2.c @@ -1,8 +1,8 @@ /* - * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. + * Copyright (c) 1998-2014 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ - * + * * 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 @@ -11,10 +11,10 @@ * 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, @@ -22,10 +22,9 @@ * 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_OSREFERENCE_LICENSE_HEADER_END@ */ -/* Copyright (c) 1998, 1999 Apple Computer, Inc. All Rights Reserved */ /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */ /* * Copyright (c) 1982, 1986, 1988, 1990, 1993 @@ -60,7 +59,12 @@ * 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 $ + */ +/* + * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce + * support for mandatory and extensible security protections. This notice + * is included in support of clause 2.2 (b) of the Apple Public License, + * Version 2.0. */ #include @@ -71,30 +75,72 @@ #include #include #include +#include #include #include #include #include #include #include +#include #include #include #include +#include #include #include #include +#include + +#if CONFIG_MACF +#include +#endif -#define DBG_FNC_SBDROP NETDBG_CODE(DBG_NETSOCK, 4) -#define DBG_FNC_SBAPPEND NETDBG_CODE(DBG_NETSOCK, 5) +#include +#if MPTCP +#include +#endif + +#define DBG_FNC_SBDROP NETDBG_CODE(DBG_NETSOCK, 4) +#define DBG_FNC_SBAPPEND NETDBG_CODE(DBG_NETSOCK, 5) + +static inline void sbcompress(struct sockbuf *, struct mbuf *, struct mbuf *); +static struct socket *sonewconn_internal(struct socket *, int); +static int sbappendaddr_internal(struct sockbuf *, struct sockaddr *, + struct mbuf *, struct mbuf *); +static int sbappendcontrol_internal(struct sockbuf *, struct mbuf *, + struct mbuf *); +static void soevent_ifdenied(struct socket *); /* * Primitive routines for operating on sockets and socket buffers */ +static int soqlimitcompat = 1; +static int soqlencomp = 0; + +/* + * Based on the number of mbuf clusters configured, high_sb_max and sb_max can + * get scaled up or down to suit that memory configuration. high_sb_max is a + * higher limit on sb_max that is checked when sb_max gets set through sysctl. + */ + +u_int32_t sb_max = SB_MAX; /* XXX should be static */ +u_int32_t high_sb_max = SB_MAX; + +static u_int32_t sb_efficiency = 8; /* parameter for sbreserve() */ +int32_t total_sbmb_cnt __attribute__((aligned(8))) = 0; +int32_t total_sbmb_cnt_peak __attribute__((aligned(8))) = 0; +int64_t sbmb_limreached __attribute__((aligned(8))) = 0; -u_long sb_max = SB_MAX; /* XXX should be static */ +/* Control whether to throttle sockets eligible to be throttled */ +__private_extern__ u_int32_t net_io_policy_throttled = 0; +static int sysctl_io_policy_throttled SYSCTL_HANDLER_ARGS; -static u_long sb_efficiency = 8; /* parameter for sbreserve() */ +u_int32_t net_io_policy_log = 0; /* log socket policy changes */ +#if CONFIG_PROC_UUID_POLICY +u_int32_t net_io_policy_uuid = 1; /* enable UUID socket policy */ +#endif /* CONFIG_PROC_UUID_POLICY */ /* * Procedures to manipulate state flags of socket @@ -126,27 +172,25 @@ static u_long sb_efficiency = 8; /* parameter for sbreserve() */ * cause software-interrupt process scheduling. */ void -soisconnecting(so) - register struct socket *so; +soisconnecting(struct socket *so) { so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING); so->so_state |= SS_ISCONNECTING; - + sflt_notify(so, sock_evt_connecting, NULL); } void -soisconnected(so) - struct socket *so; +soisconnected(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)) { so->so_state &= ~SS_INCOMP; so->so_state |= SS_COMP; @@ -169,15 +213,17 @@ soisconnected(so) wakeup((caddr_t)&so->so_timeo); sorwakeup(so); sowwakeup(so); + soevent(so, SO_FILT_HINT_LOCKED | SO_FILT_HINT_CONNECTED | + SO_FILT_HINT_CONNINFO_UPDATED); } } void -soisdisconnecting(so) - register struct socket *so; +soisdisconnecting(struct socket *so) { so->so_state &= ~SS_ISCONNECTING; so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE); + soevent(so, SO_FILT_HINT_LOCKED); sflt_notify(so, sock_evt_disconnecting, NULL); wakeup((caddr_t)&so->so_timeo); sowwakeup(so); @@ -185,68 +231,43 @@ soisdisconnecting(so) } void -soisdisconnected(so) - register struct socket *so; +soisdisconnected(struct socket *so) { so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED); + soevent(so, SO_FILT_HINT_LOCKED | SO_FILT_HINT_DISCONNECTED | + SO_FILT_HINT_CONNINFO_UPDATED); sflt_notify(so, sock_evt_disconnected, NULL); wakeup((caddr_t)&so->so_timeo); sowwakeup(so); sorwakeup(so); + +#if CONTENT_FILTER + /* Notify content filters as soon as we cannot send/receive data */ + cfil_sock_notify_shutdown(so, SHUT_RDWR); +#endif /* CONTENT_FILTER */ } /* - * Return a random connection that hasn't been serviced yet and - * is eligible for discard. There is a one in qlen chance that - * we will return a null, saying that there are no dropable - * requests. In this case, the protocol specific code should drop - * the new request. This insures fairness. - * - * This may be used in conjunction with protocol specific queue - * congestion routines. + * This function will issue a wakeup like soisdisconnected but it will not + * notify the socket filters. This will avoid unlocking the socket + * in the midst of closing it. */ -struct socket * -sodropablereq(head) - register struct socket *head; +void +sodisconnectwakeup(struct socket *so) { - struct socket *so, *sonext = NULL; - unsigned int i, j, qlen; - static int rnd; - static struct timeval old_runtime; - static unsigned int cur_cnt, old_cnt; - struct timeval tv; - - microtime(&tv); - if ((i = (tv.tv_sec - old_runtime.tv_sec)) != 0) { - old_runtime = tv; - old_cnt = cur_cnt / i; - cur_cnt = 0; - } - - so = TAILQ_FIRST(&head->so_incomp); - if (!so) - return (NULL); - - qlen = head->so_incqlen; - if (++cur_cnt > qlen || old_cnt > qlen) { - rnd = (314159 * rnd + 66329) & 0xffff; - j = ((qlen + 1) * rnd) >> 16; -//###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; - } - } + so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING); + so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED); + soevent(so, SO_FILT_HINT_LOCKED | SO_FILT_HINT_DISCONNECTED | + SO_FILT_HINT_CONNINFO_UPDATED); + wakeup((caddr_t)&so->so_timeo); + sowwakeup(so); + sorwakeup(so); -// if (in_pcb_checkstate(so->so_pcb, WNT_ACQUIRE, 0) == WNT_STOPUSING) -// return (NULL); -// else - return (so); +#if CONTENT_FILTER + /* Notify content filters as soon as we cannot send/receive data */ + cfil_sock_notify_shutdown(so, SHUT_RDWR); +#endif /* CONTENT_FILTER */ } /* @@ -258,32 +279,50 @@ sodropablereq(head) * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED. */ static struct socket * -sonewconn_internal(head, connstatus) - register struct socket *head; - int connstatus; +sonewconn_internal(struct socket *head, int connstatus) { - int error = 0; - register struct socket *so; + int so_qlen, error = 0; + struct socket *so; lck_mtx_t *mutex_held; - if (head->so_proto->pr_getlock != NULL) + if (head->so_proto->pr_getlock != NULL) mutex_held = (*head->so_proto->pr_getlock)(head, 0); - else + 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) + if (!soqlencomp) { + /* + * This is the default case; so_qlen represents the + * sum of both incomplete and completed queues. + */ + so_qlen = head->so_qlen; + } else { + /* + * When kern.ipc.soqlencomp is set to 1, so_qlen + * represents only the completed queue. Since we + * cannot let the incomplete queue goes unbounded + * (in case of SYN flood), we cap the incomplete + * queue length to at most somaxconn, and use that + * as so_qlen so that we fail immediately below. + */ + so_qlen = head->so_qlen - head->so_incqlen; + if (head->so_incqlen > somaxconn) + so_qlen = somaxconn; + } + + if (so_qlen >= + (soqlimitcompat ? head->so_qlimit : (3 * head->so_qlimit / 2))) return ((struct socket *)0); - so = soalloc(1, head->so_proto->pr_domain->dom_family, head->so_type); + so = soalloc(1, SOCK_DOM(head), head->so_type); if (so == NULL) return ((struct socket *)0); /* check if head was closed during the soalloc */ if (head->so_proto == NULL) { - sodealloc(so); - return ((struct socket *)0); + sodealloc(so); + return ((struct socket *)0); } - so->so_head = head; so->so_type = head->so_type; so->so_options = head->so_options &~ SO_ACCEPTCONN; so->so_linger = head->so_linger; @@ -291,38 +330,86 @@ sonewconn_internal(head, connstatus) so->so_proto = head->so_proto; so->so_timeo = head->so_timeo; so->so_pgid = head->so_pgid; - so->so_uid = head->so_uid; + kauth_cred_ref(head->so_cred); + so->so_cred = head->so_cred; + so->last_pid = head->last_pid; + so->last_upid = head->last_upid; + memcpy(so->last_uuid, head->last_uuid, sizeof (so->last_uuid)); + if (head->so_flags & SOF_DELEGATED) { + so->e_pid = head->e_pid; + so->e_upid = head->e_upid; + memcpy(so->e_uuid, head->e_uuid, sizeof (so->e_uuid)); + } + /* inherit socket options stored in so_flags */ + so->so_flags = head->so_flags & + (SOF_NOSIGPIPE | SOF_NOADDRAVAIL | SOF_REUSESHAREUID | + SOF_NOTIFYCONFLICT | SOF_BINDRANDOMPORT | SOF_NPX_SETOPTSHUT | + SOF_NODEFUNCT | SOF_PRIVILEGED_TRAFFIC_CLASS| SOF_NOTSENT_LOWAT | + SOF_USELRO | SOF_DELEGATED); so->so_usecount = 1; + so->next_lock_lr = 0; + so->next_unlock_lr = 0; -#ifdef __APPLE__ so->so_rcv.sb_flags |= SB_RECV; /* XXX */ so->so_rcv.sb_so = so->so_snd.sb_so = so; TAILQ_INIT(&so->so_evlist); + +#if CONFIG_MACF_SOCKET + mac_socket_label_associate_accept(head, so); #endif + /* inherit traffic management properties of listener */ + so->so_traffic_mgt_flags = + head->so_traffic_mgt_flags & (TRAFFIC_MGT_SO_BACKGROUND); + so->so_background_thread = head->so_background_thread; + so->so_traffic_class = head->so_traffic_class; + if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat)) { - sflt_termsock(so); sodealloc(so); return ((struct socket *)0); } + so->so_rcv.sb_flags |= (head->so_rcv.sb_flags & SB_USRSIZE); + so->so_snd.sb_flags |= (head->so_snd.sb_flags & SB_USRSIZE); /* - * Must be done with head unlocked to avoid deadlock for protocol with per socket mutexes. + * 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); + if (((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL) != 0) || + error) { sodealloc(so); if (head->so_proto->pr_unlock) socket_lock(head, 0); return ((struct socket *)0); } - if (head->so_proto->pr_unlock) + if (head->so_proto->pr_unlock) { socket_lock(head, 0); -#ifdef __APPLE__ - so->so_proto->pr_domain->dom_refs++; -#endif + /* + * Radar 7385998 Recheck that the head is still accepting + * to avoid race condition when head is getting closed. + */ + if ((head->so_options & SO_ACCEPTCONN) == 0) { + so->so_state &= ~SS_NOFDREF; + soclose(so); + return ((struct socket *)0); + } + } + + atomic_add_32(&so->so_proto->pr_domain->dom_refs, 1); + + /* Insert in head appropriate lists */ + so->so_head = head; + + /* + * Since this socket is going to be inserted into the incomp + * queue, it can be picked up by another thread in + * tcp_dropdropablreq to get dropped before it is setup.. + * To prevent this race, set in-progress flag which can be + * cleared later + */ + so->so_flags |= SOF_INCOMP_INPROGRESS; if (connstatus) { TAILQ_INSERT_TAIL(&head->so_comp, so, so_list); @@ -334,10 +421,9 @@ sonewconn_internal(head, connstatus) } head->so_qlen++; -#ifdef __APPLE__ - /* Attach socket filters for this protocol */ - sflt_initsock(so); -#endif + /* Attach socket filters for this protocol */ + sflt_initsock(so); + if (connstatus) { so->so_state |= connstatus; sorwakeup(head); @@ -348,38 +434,14 @@ sonewconn_internal(head, connstatus) struct socket * -sonewconn( - struct socket *head, - int connstatus, - const struct sockaddr *from) +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); - } - + int error = sflt_connectin(head, from); if (error) { - return NULL; + return (NULL); } - - return sonewconn_internal(head, connstatus); + + return (sonewconn_internal(head, connstatus)); } /* @@ -393,19 +455,19 @@ sonewconn( */ void -socantsendmore(so) - struct socket *so; +socantsendmore(struct socket *so) { so->so_state |= SS_CANTSENDMORE; + soevent(so, SO_FILT_HINT_LOCKED | SO_FILT_HINT_CANTSENDMORE); sflt_notify(so, sock_evt_cantsendmore, NULL); sowwakeup(so); } void -socantrcvmore(so) - struct socket *so; +socantrcvmore(struct socket *so) { so->so_state |= SS_CANTRCVMORE; + soevent(so, SO_FILT_HINT_LOCKED | SO_FILT_HINT_CANTRCVMORE); sflt_notify(so, sock_evt_cantrecvmore, NULL); sorwakeup(so); } @@ -414,84 +476,71 @@ socantrcvmore(so) * Wait for data to arrive at/drain from a socket buffer. */ int -sbwait(sb) - struct sockbuf *sb; +sbwait(struct sockbuf *sb) { - int error = 0, lr, lr_saved; + boolean_t nointr = (sb->sb_flags & SB_NOINTR); + void *lr_saved = __builtin_return_address(0); struct socket *so = sb->sb_so; lck_mtx_t *mutex_held; struct timespec ts; + int error = 0; -#ifdef __ppc__ - __asm__ volatile("mflr %0" : "=r" (lr)); - lr_saved = lr; -#endif - + if (so == NULL) { + panic("%s: null so, sb=%p sb_flags=0x%x lr=%p\n", + __func__, sb, sb->sb_flags, lr_saved); + /* NOTREACHED */ + } else if (so->so_usecount < 1) { + panic("%s: sb=%p sb_flags=0x%x sb_so=%p usecount=%d lr=%p " + "lrh= %s\n", __func__, sb, sb->sb_flags, so, + so->so_usecount, lr_saved, solockhistory_nr(so)); + /* NOTREACHED */ + } - if (so->so_proto->pr_getlock != NULL) + if (so->so_proto->pr_getlock != NULL) mutex_held = (*so->so_proto->pr_getlock)(so, 0); - else + else mutex_held = so->so_proto->pr_domain->dom_mtx; - sb->sb_flags |= SB_WAIT; + 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); ts.tv_sec = sb->sb_timeo.tv_sec; ts.tv_nsec = sb->sb_timeo.tv_usec * 1000; + + sb->sb_waiters++; + VERIFY(sb->sb_waiters != 0); + error = msleep((caddr_t)&sb->sb_cc, mutex_held, - (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait", - &ts); + nointr ? PSOCK : PSOCK | PCATCH, + nointr ? "sbwait_nointr" : "sbwait", &ts); - lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); + VERIFY(sb->sb_waiters != 0); + sb->sb_waiters--; - if (so->so_usecount < 1) - panic("sbwait: so=%x refcount=%d\n", so, so->so_usecount); + if (so->so_usecount < 1) { + panic("%s: 2 sb=%p sb_flags=0x%x sb_so=%p usecount=%d lr=%p " + "lrh= %s\n", __func__, sb, sb->sb_flags, so, + so->so_usecount, lr_saved, solockhistory_nr(so)); + /* NOTREACHED */ + } - if ((so->so_state & SS_DRAINING)) { + if ((so->so_state & SS_DRAINING) || (so->so_flags & SOF_DEFUNCT)) { error = EBADF; + if (so->so_flags & SOF_DEFUNCT) { + SODEFUNCTLOG(("%s[%d]: defunct so 0x%llx [%d,%d] " + "(%d)\n", __func__, proc_selfpid(), + (uint64_t)VM_KERNEL_ADDRPERM(so), + SOCK_DOM(so), SOCK_TYPE(so), error)); + } } return (error); } -/* - * Lock a sockbuf already known to be locked; - * return any error returned from sleep (EINTR). - */ -int -sb_lock(sb) - register struct sockbuf *sb; +void +sbwakeup(struct sockbuf *sb) { - struct socket *so = sb->sb_so; - lck_mtx_t * mutex_held; - int error = 0, lr, lr_saved; - -#ifdef __ppc__ - __asm__ volatile("mflr %0" : "=r" (lr)); - lr_saved = lr; -#endif - - 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; - 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; - return (0); + if (sb->sb_waiters > 0) + wakeup((caddr_t)&sb->sb_cc); } /* @@ -500,31 +549,61 @@ sb_lock(sb) * if the socket has the SS_ASYNC flag set. */ void -sowakeup(so, sb) - register struct socket *so; - register struct sockbuf *sb; -{ - struct proc *p = current_proc(); +sowakeup(struct socket *so, struct sockbuf *sb) +{ + if (so->so_flags & SOF_DEFUNCT) { + SODEFUNCTLOG(("%s[%d]: defunct so 0x%llx [%d,%d] si 0x%x, " + "fl 0x%x [%s]\n", __func__, proc_selfpid(), + (uint64_t)VM_KERNEL_ADDRPERM(so), SOCK_DOM(so), + SOCK_TYPE(so), (uint32_t)sb->sb_sel.si_flags, sb->sb_flags, + (sb->sb_flags & SB_RECV) ? "rcv" : "snd")); + } + sb->sb_flags &= ~SB_SEL; selwakeup(&sb->sb_sel); - if (sb->sb_flags & SB_WAIT) { - sb->sb_flags &= ~SB_WAIT; - wakeup((caddr_t)&sb->sb_cc); - } + sbwakeup(sb); if (so->so_state & SS_ASYNC) { if (so->so_pgid < 0) gsignal(-so->so_pgid, SIGIO); - else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0) - psignal(p, SIGIO); + else if (so->so_pgid > 0) + proc_signal(so->so_pgid, SIGIO); } if (sb->sb_flags & SB_KNOTE) { KNOTE(&sb->sb_sel.si_note, SO_FILT_HINT_LOCKED); } if (sb->sb_flags & SB_UPCALL) { + void (*sb_upcall)(struct socket *, void *, int); + caddr_t sb_upcallarg; + + sb_upcall = sb->sb_upcall; + sb_upcallarg = sb->sb_upcallarg; + /* Let close know that we're about to do an upcall */ + so->so_upcallusecount++; + socket_unlock(so, 0); - (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT); + (*sb_upcall)(so, sb_upcallarg, M_DONTWAIT); socket_lock(so, 0); + + so->so_upcallusecount--; + /* Tell close that it's safe to proceed */ + if ((so->so_flags & SOF_CLOSEWAIT) && + so->so_upcallusecount == 0) + wakeup((caddr_t)&so->so_upcallusecount); + } +#if CONTENT_FILTER + /* + * Trap disconnection events for content filters + */ + if ((so->so_flags & SOF_CONTENT_FILTER) != 0) { + if ((sb->sb_flags & SB_RECV)) { + if (so->so_state & (SS_CANTRCVMORE)) + cfil_sock_notify_shutdown(so, SHUT_RD); + } else { + if (so->so_state & (SS_CANTSENDMORE)) + cfil_sock_notify_shutdown(so, SHUT_WR); + } } +#endif /* CONTENT_FILTER */ } /* @@ -559,16 +638,24 @@ sowakeup(so, sb) * should be released by calling sbrelease() when the socket is destroyed. */ +/* + * Returns: 0 Success + * ENOBUFS + */ int -soreserve(so, sndcc, rcvcc) - register struct socket *so; - u_long sndcc, rcvcc; +soreserve(struct socket *so, u_int32_t sndcc, u_int32_t rcvcc) { if (sbreserve(&so->so_snd, sndcc) == 0) goto bad; + else + so->so_snd.sb_idealsize = sndcc; + if (sbreserve(&so->so_rcv, rcvcc) == 0) goto bad2; + else + so->so_rcv.sb_idealsize = rcvcc; + if (so->so_rcv.sb_lowat == 0) so->so_rcv.sb_lowat = 1; if (so->so_snd.sb_lowat == 0) @@ -577,9 +664,8 @@ soreserve(so, sndcc, rcvcc) so->so_snd.sb_lowat = so->so_snd.sb_hiwat; return (0); bad2: -#ifdef __APPLE__ + so->so_snd.sb_flags &= ~SB_SEL; selthreadclear(&so->so_snd.sb_sel); -#endif sbrelease(&so->so_snd); bad: return (ENOBUFS); @@ -591,9 +677,7 @@ bad: * if buffering efficiency is near the normal case. */ int -sbreserve(sb, cc) - struct sockbuf *sb; - u_long cc; +sbreserve(struct sockbuf *sb, u_int32_t cc) { if ((u_quad_t)cc > (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES)) return (0); @@ -607,16 +691,13 @@ sbreserve(sb, cc) /* * Free mbufs held by a socket, and reserved mbuf space. */ - /* WARNING needs to do selthreadclear() before calling this */ +/* WARNING needs to do selthreadclear() before calling this */ void -sbrelease(sb) - struct sockbuf *sb; +sbrelease(struct sockbuf *sb) { - sbflush(sb); sb->sb_hiwat = 0; sb->sb_mbmax = 0; - } /* @@ -651,73 +732,106 @@ sbrelease(sb) * discarded and mbufs are compacted where possible. */ int -sbappend(sb, m) - struct sockbuf *sb; - struct mbuf *m; +sbappend(struct sockbuf *sb, struct mbuf *m) { - register struct mbuf *n, *sb_first; - int result = 0; - int error = 0; - int filtered = 0; + struct socket *so = sb->sb_so; + + if (m == NULL || (sb->sb_flags & SB_DROP)) { + if (m != NULL) + m_freem(m); + return (0); + } + SBLASTRECORDCHK(sb, "sbappend 1"); - KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_START), sb, m->m_len, 0, 0, 0); + if (sb->sb_lastrecord != NULL && (sb->sb_mbtail->m_flags & M_EOR)) + return (sbappendrecord(sb, m)); - 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) { - 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)); - } - - 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 (sb->sb_flags & SB_RECV && !(m && m->m_flags & M_SKIPCFIL)) { + int error = sflt_data_in(so, NULL, &m, NULL, 0); + SBLASTRECORDCHK(sb, "sbappend 2"); + +#if CONTENT_FILTER + if (error == 0) + error = cfil_sock_data_in(so, NULL, m, NULL, 0); +#endif /* CONTENT_FILTER */ + + if (error != 0) { + if (error != EJUSTRETURN) + m_freem(m); + 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; + } else if (m) { + m->m_flags &= ~M_SKIPCFIL; + } + + /* If this is the first record, it's also the last record */ + if (sb->sb_lastrecord == NULL) + sb->sb_lastrecord = m; + + sbcompress(sb, m, sb->sb_mbtail); + SBLASTRECORDCHK(sb, "sbappend 3"); + return (1); +} + +/* + * Similar to sbappend, except that this is optimized for stream sockets. + */ +int +sbappendstream(struct sockbuf *sb, struct mbuf *m) +{ + struct socket *so = sb->sb_so; + + if (m == NULL || (sb->sb_flags & SB_DROP)) { + if (m != NULL) + m_freem(m); + return (0); + } + + if (m->m_nextpkt != NULL || (sb->sb_mb != sb->sb_lastrecord)) { + panic("sbappendstream: nexpkt %p || mb %p != lastrecord %p\n", + m->m_nextpkt, sb->sb_mb, sb->sb_lastrecord); + /* NOTREACHED */ } - result = sbcompress(sb, m, n); + SBLASTMBUFCHK(sb, __func__); + + if (sb->sb_flags & SB_RECV && !(m && m->m_flags & M_SKIPCFIL)) { + int error = sflt_data_in(so, NULL, &m, NULL, 0); + SBLASTRECORDCHK(sb, "sbappendstream 1"); + +#if CONTENT_FILTER + if (error == 0) + error = cfil_sock_data_in(so, NULL, m, NULL, 0); +#endif /* CONTENT_FILTER */ + + if (error != 0) { + if (error != EJUSTRETURN) + m_freem(m); + return (0); + } + } else if (m) { + m->m_flags &= ~M_SKIPCFIL; + } - KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_END), sb, sb->sb_cc, 0, 0, 0); - - return result; + sbcompress(sb, m, sb->sb_mbtail); + sb->sb_lastrecord = sb->sb_mb; + SBLASTRECORDCHK(sb, "sbappendstream 2"); + return (1); } #ifdef SOCKBUF_DEBUG void -sbcheck(sb) - register struct sockbuf *sb; +sbcheck(struct sockbuf *sb) { - register struct mbuf *m; - register struct mbuf *n = 0; - register u_long len = 0, mbcnt = 0; + struct mbuf *m; + struct mbuf *n = 0; + u_int32_t len = 0, mbcnt = 0; lck_mtx_t *mutex_held; - if (sb->sb_so->so_proto->pr_getlock != NULL) + if (sb->sb_so->so_proto->pr_getlock != NULL) mutex_held = (*sb->sb_so->so_proto->pr_getlock)(sb->sb_so, 0); - else + else mutex_held = sb->sb_so->so_proto->pr_domain->dom_mtx; lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); @@ -726,65 +840,135 @@ sbcheck(sb) 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 (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) { - panic("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc, - mbcnt, sb->sb_mbcnt); - } + n = m->m_nextpkt; + for (; m; m = m->m_next) { + len += m->m_len; + mbcnt += MSIZE; + /* XXX pretty sure this is bogus */ + if (m->m_flags & M_EXT) + 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 +void +sblastrecordchk(struct sockbuf *sb, const char *where) +{ + struct mbuf *m = sb->sb_mb; + + while (m && m->m_nextpkt) + m = m->m_nextpkt; + + if (m != sb->sb_lastrecord) { + printf("sblastrecordchk: mb 0x%llx lastrecord 0x%llx " + "last 0x%llx\n", + (uint64_t)VM_KERNEL_ADDRPERM(sb->sb_mb), + (uint64_t)VM_KERNEL_ADDRPERM(sb->sb_lastrecord), + (uint64_t)VM_KERNEL_ADDRPERM(m)); + printf("packet chain:\n"); + for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt) + printf("\t0x%llx\n", (uint64_t)VM_KERNEL_ADDRPERM(m)); + panic("sblastrecordchk from %s", where); + } +} + +void +sblastmbufchk(struct sockbuf *sb, const char *where) +{ + struct mbuf *m = sb->sb_mb; + struct mbuf *n; + + while (m && m->m_nextpkt) + m = m->m_nextpkt; + + while (m && m->m_next) + m = m->m_next; + + if (m != sb->sb_mbtail) { + printf("sblastmbufchk: mb 0x%llx mbtail 0x%llx last 0x%llx\n", + (uint64_t)VM_KERNEL_ADDRPERM(sb->sb_mb), + (uint64_t)VM_KERNEL_ADDRPERM(sb->sb_mbtail), + (uint64_t)VM_KERNEL_ADDRPERM(m)); + printf("packet tree:\n"); + for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt) { + printf("\t"); + for (n = m; n != NULL; n = n->m_next) + printf("0x%llx ", (uint64_t)VM_KERNEL_ADDRPERM(n)); + printf("\n"); + } + panic("sblastmbufchk from %s", where); + } +} + /* - * As above, except the mbuf chain - * begins a new record. + * Similar to sbappend, except the mbuf chain begins a new record. */ int -sbappendrecord(sb, m0) - register struct sockbuf *sb; - register struct mbuf *m0; +sbappendrecord(struct sockbuf *sb, struct mbuf *m0) { - register struct mbuf *m; - int result = 0; + struct mbuf *m; + int space = 0; + + if (m0 == NULL || (sb->sb_flags & SB_DROP)) { + if (m0 != NULL) + m_freem(m0); + return (0); + } + + for (m = m0; m != NULL; m = m->m_next) + space += m->m_len; + + if (space > sbspace(sb) && !(sb->sb_flags & SB_UNIX)) { + m_freem(m0); + return (0); + } + + if (sb->sb_flags & SB_RECV && !(m0 && m0->m_flags & M_SKIPCFIL)) { + int error = sflt_data_in(sb->sb_so, NULL, &m0, NULL, + sock_data_filt_flag_record); + +#if CONTENT_FILTER + if (error == 0) + error = cfil_sock_data_in(sb->sb_so, NULL, m0, NULL, 0); +#endif /* CONTENT_FILTER */ - if (m0 == 0) - 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) { + SBLASTRECORDCHK(sb, "sbappendrecord 1"); if (error != EJUSTRETURN) m_freem(m0); - return 0; + return (0); } + } else if (m0) { + m0->m_flags &= ~M_SKIPCFIL; } - - m = sb->sb_mb; - if (m) - while (m->m_nextpkt) - m = m->m_nextpkt; + /* - * Put the first mbuf on the queue. * Note this permits zero length records. */ sballoc(sb, m0); - if (m) - m->m_nextpkt = m0; - else + SBLASTRECORDCHK(sb, "sbappendrecord 2"); + if (sb->sb_lastrecord != NULL) { + sb->sb_lastrecord->m_nextpkt = m0; + } else { sb->sb_mb = m0; + } + sb->sb_lastrecord = m0; + sb->sb_mbtail = m0; + m = m0->m_next; m0->m_next = 0; if (m && (m0->m_flags & M_EOR)) { m0->m_flags &= ~M_EOR; m->m_flags |= M_EOR; } - return sbcompress(sb, m, m0); + sbcompress(sb, m, m0); + SBLASTRECORDCHK(sb, "sbappendrecord 3"); + return (1); } /* @@ -793,31 +977,40 @@ sbappendrecord(sb, m0) * but after any other OOB data. */ int -sbinsertoob(sb, m0) - struct sockbuf *sb; - struct mbuf *m0; +sbinsertoob(struct sockbuf *sb, struct mbuf *m0) { struct mbuf *m; struct mbuf **mp; if (m0 == 0) - return 0; - - if ((sb->sb_flags & SB_RECV) != 0) { + return (0); + + SBLASTRECORDCHK(sb, "sbinsertoob 1"); + + if ((sb->sb_flags & SB_RECV && !(m0->m_flags & M_SKIPCFIL)) != 0) { int error = sflt_data_in(sb->sb_so, NULL, &m0, NULL, - sock_data_filt_flag_oob, NULL); - + sock_data_filt_flag_oob); + + SBLASTRECORDCHK(sb, "sbinsertoob 2"); + +#if CONTENT_FILTER + if (error == 0) + error = cfil_sock_data_in(sb->sb_so, NULL, m0, NULL, 0); +#endif /* CONTENT_FILTER */ + if (error) { if (error != EJUSTRETURN) { m_freem(m0); } - return 0; + return (0); } + } else if (m0) { + m0->m_flags &= ~M_SKIPCFIL; } - - for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) { - m = *mp; - again: + + for (mp = &sb->sb_mb; *mp; mp = &((*mp)->m_nextpkt)) { + m = *mp; +again: switch (m->m_type) { case MT_OOBDATA: @@ -836,6 +1029,10 @@ sbinsertoob(sb, m0) */ sballoc(sb, m0); m0->m_nextpkt = *mp; + if (*mp == NULL) { + /* m0 is actually the new tail */ + sb->sb_lastrecord = m0; + } *mp = m0; m = m0->m_next; m0->m_next = 0; @@ -843,7 +1040,9 @@ sbinsertoob(sb, m0) m0->m_flags &= ~M_EOR; m->m_flags |= M_EOR; } - return sbcompress(sb, m, m0); + sbcompress(sb, m, m0); + SBLASTRECORDCHK(sb, "sbinsertoob 3"); + return (1); } /* @@ -851,14 +1050,15 @@ sbinsertoob(sb, m0) * to the receive queue of a socket. If present, * m0 must include a packet header with total length. * Returns 0 if no space in sockbuf or insufficient mbufs. + * + * Returns: 0 No space/out of mbufs + * 1 Success */ static int -sbappendaddr_internal(sb, asa, m0, control) - register struct sockbuf *sb; - struct sockaddr *asa; - struct mbuf *m0, *control; +sbappendaddr_internal(struct sockbuf *sb, struct sockaddr *asa, + struct mbuf *m0, struct mbuf *control) { - register struct mbuf *m, *n; + struct mbuf *m, *n, *nlast; int space = asa->sa_len; if (m0 && (m0->m_flags & M_PKTHDR) == 0) @@ -885,64 +1085,104 @@ sbappendaddr_internal(sb, asa, m0, control) else control = m0; m->m_next = control; - for (n = m; n; n = n->m_next) + + SBLASTRECORDCHK(sb, "sbappendadddr 1"); + + for (n = m; n->m_next != NULL; n = n->m_next) sballoc(sb, n); - n = sb->sb_mb; - if (n) { - while (n->m_nextpkt) - n = n->m_nextpkt; - n->m_nextpkt = m; - } else + sballoc(sb, n); + nlast = n; + + if (sb->sb_lastrecord != NULL) { + sb->sb_lastrecord->m_nextpkt = m; + } else { sb->sb_mb = m; - postevent(0,sb,EV_RWBYTES); + } + sb->sb_lastrecord = m; + sb->sb_mbtail = nlast; + + SBLASTMBUFCHK(sb, __func__); + SBLASTRECORDCHK(sb, "sbappendadddr 2"); + + postevent(0, sb, EV_RWBYTES); return (1); } +/* + * Returns: 0 Error: No space/out of mbufs/etc. + * 1 Success + * + * Imputed: (*error_out) errno for error + * ENOBUFS + * sflt_data_in:??? [whatever a filter author chooses] + */ int -sbappendaddr( - struct sockbuf* sb, - struct sockaddr* asa, - struct mbuf *m0, - struct mbuf *control, - int *error_out) +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; - + boolean_t sb_unix = (sb->sb_flags & SB_UNIX); + + if (error_out) + *error_out = 0; + if (m0 && (m0->m_flags & M_PKTHDR) == 0) panic("sbappendaddrorfree"); - + + if (sb->sb_flags & SB_DROP) { + if (m0 != NULL) + m_freem(m0); + if (control != NULL && !sb_unix) + m_freem(control); + if (error_out != NULL) + *error_out = EINVAL; + return (0); + } + /* Call socket data in filters */ - if ((sb->sb_flags & SB_RECV) != 0) { + if (sb->sb_flags & SB_RECV && !(m0 && m0->m_flags & M_SKIPCFIL)) { int error; - error = sflt_data_in(sb->sb_so, asa, &m0, &control, 0, NULL); + error = sflt_data_in(sb->sb_so, asa, &m0, &control, 0); + SBLASTRECORDCHK(sb, __func__); + +#if CONTENT_FILTER + if (error == 0) + error = cfil_sock_data_in(sb->sb_so, asa, m0, control, 0); +#endif /* CONTENT_FILTER */ + if (error) { if (error != EJUSTRETURN) { - if (m0) m_freem(m0); - if (control) m_freem(control); - if (error_out) *error_out = error; + if (m0) + m_freem(m0); + if (control != NULL && !sb_unix) + m_freem(control); + if (error_out) + *error_out = error; } - return 0; + return (0); } + } else if (m0) { + m0->m_flags &= ~M_SKIPCFIL; } - + 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; + if (m0) + m_freem(m0); + if (control != NULL && !sb_unix) + m_freem(control); + if (error_out) + *error_out = ENOBUFS; } - - return result; + + return (result); } static int -sbappendcontrol_internal(sb, m0, control) - struct sockbuf *sb; - struct mbuf *control, *m0; +sbappendcontrol_internal(struct sockbuf *sb, struct mbuf *m0, + struct mbuf *control) { - register struct mbuf *m, *n; + struct mbuf *m, *mlast, *n; int space = 0; if (control == 0) @@ -956,79 +1196,457 @@ sbappendcontrol_internal(sb, m0, control) n = m; /* save pointer to last control buffer */ for (m = m0; m; m = m->m_next) space += m->m_len; - if (space > sbspace(sb)) + if (space > sbspace(sb) && !(sb->sb_flags & SB_UNIX)) return (0); n->m_next = m0; /* concatenate data to control */ - for (m = control; m; m = m->m_next) + SBLASTRECORDCHK(sb, "sbappendcontrol 1"); + + for (m = control; m->m_next != NULL; m = m->m_next) sballoc(sb, m); - n = sb->sb_mb; - if (n) { - while (n->m_nextpkt) - n = n->m_nextpkt; - n->m_nextpkt = control; - } else + sballoc(sb, m); + mlast = m; + + if (sb->sb_lastrecord != NULL) { + sb->sb_lastrecord->m_nextpkt = control; + } else { sb->sb_mb = control; - postevent(0,sb,EV_RWBYTES); + } + sb->sb_lastrecord = control; + sb->sb_mbtail = mlast; + + SBLASTMBUFCHK(sb, __func__); + SBLASTRECORDCHK(sb, "sbappendcontrol 2"); + + postevent(0, sb, EV_RWBYTES); return (1); } int -sbappendcontrol( - struct sockbuf *sb, - struct mbuf *m0, - struct mbuf *control, - int *error_out) +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) { + boolean_t sb_unix = (sb->sb_flags & SB_UNIX); + + if (error_out) + *error_out = 0; + + if (sb->sb_flags & SB_DROP) { + if (m0 != NULL) + m_freem(m0); + if (control != NULL && !sb_unix) + m_freem(control); + if (error_out != NULL) + *error_out = EINVAL; + return (0); + } + + if (sb->sb_flags & SB_RECV && !(m0 && m0->m_flags & M_SKIPCFIL)) { int error; - error = sflt_data_in(sb->sb_so, NULL, &m0, &control, 0, NULL); + + error = sflt_data_in(sb->sb_so, NULL, &m0, &control, 0); + SBLASTRECORDCHK(sb, __func__); + +#if CONTENT_FILTER + if (error == 0) + error = cfil_sock_data_in(sb->sb_so, NULL, m0, control, 0); +#endif /* CONTENT_FILTER */ + if (error) { if (error != EJUSTRETURN) { - if (m0) m_freem(m0); - if (control) m_freem(control); - if (error_out) *error_out = error; + if (m0) + m_freem(m0); + if (control != NULL && !sb_unix) + m_freem(control); + if (error_out) + *error_out = error; } - return 0; + return (0); } + } else if (m0) { + m0->m_flags &= ~M_SKIPCFIL; } - + 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; + if (m0) + m_freem(m0); + if (control != NULL && !sb_unix) + m_freem(control); + if (error_out) + *error_out = ENOBUFS; } - - return result; + + return (result); } /* - * Compress mbuf chain m into the socket - * buffer sb following mbuf n. If n - * is null, the buffer is presumed empty. + * Append a contiguous TCP data blob with TCP sequence number as control data + * as a new msg to the receive socket buffer. */ -static int -sbcompress(sb, m, n) - register struct sockbuf *sb; - register struct mbuf *m, *n; +int +sbappendmsgstream_rcv(struct sockbuf *sb, struct mbuf *m, uint32_t seqnum, + int unordered) +{ + struct mbuf *m_eor = NULL; + u_int32_t data_len = 0; + int ret = 0; + struct socket *so = sb->sb_so; + + VERIFY((m->m_flags & M_PKTHDR) && m_pktlen(m) > 0); + VERIFY(so->so_msg_state != NULL); + VERIFY(sb->sb_flags & SB_RECV); + + /* Keep the TCP sequence number in the mbuf pkthdr */ + m->m_pkthdr.msg_seq = seqnum; + + /* find last mbuf and set M_EOR */ + for (m_eor = m; ; m_eor = m_eor->m_next) { + /* + * If the msg is unordered, we need to account for + * these bytes in receive socket buffer size. Otherwise, + * the receive window advertised will shrink because + * of the additional unordered bytes added to the + * receive buffer. + */ + if (unordered) { + m_eor->m_flags |= M_UNORDERED_DATA; + data_len += m_eor->m_len; + so->so_msg_state->msg_uno_bytes += m_eor->m_len; + } else { + m_eor->m_flags &= ~M_UNORDERED_DATA; + } + if (m_eor->m_next == NULL) + break; + } + + /* set EOR flag at end of byte blob */ + m_eor->m_flags |= M_EOR; + + /* expand the receive socket buffer to allow unordered data */ + if (unordered && !sbreserve(sb, sb->sb_hiwat + data_len)) { + /* + * Could not allocate memory for unordered data, it + * means this packet will have to be delivered in order + */ + printf("%s: could not reserve space for unordered data\n", + __func__); + } + + if (!unordered && (sb->sb_mbtail != NULL) && + !(sb->sb_mbtail->m_flags & M_UNORDERED_DATA)) { + sb->sb_mbtail->m_flags &= ~M_EOR; + sbcompress(sb, m, sb->sb_mbtail); + ret = 1; + } else { + ret = sbappendrecord(sb, m); + } + VERIFY(sb->sb_mbtail->m_flags & M_EOR); + return (ret); +} + +/* + * TCP streams have message based out of order delivery support, or have + * Multipath TCP support, or are regular TCP sockets + */ +int +sbappendstream_rcvdemux(struct socket *so, struct mbuf *m, uint32_t seqnum, + int unordered) +{ + int ret = 0; + + if ((m != NULL) && (m_pktlen(m) <= 0)) { + m_freem(m); + return (ret); + } + + if (so->so_flags & SOF_ENABLE_MSGS) { + ret = sbappendmsgstream_rcv(&so->so_rcv, m, seqnum, unordered); + } +#if MPTCP + else if (so->so_flags & SOF_MPTCP_TRUE) { + ret = sbappendmptcpstream_rcv(&so->so_rcv, m); + } +#endif /* MPTCP */ + else { + ret = sbappendstream(&so->so_rcv, m); + } + return (ret); +} + +#if MPTCP +int +sbappendmptcpstream_rcv(struct sockbuf *sb, struct mbuf *m) +{ + struct socket *so = sb->sb_so; + + VERIFY(m == NULL || (m->m_flags & M_PKTHDR)); + /* SB_NOCOMPRESS must be set prevent loss of M_PKTHDR data */ + VERIFY((sb->sb_flags & (SB_RECV|SB_NOCOMPRESS)) == + (SB_RECV|SB_NOCOMPRESS)); + + if (m == NULL || m_pktlen(m) == 0 || (sb->sb_flags & SB_DROP) || + (so->so_state & SS_CANTRCVMORE)) { + if (m != NULL) + m_freem(m); + return (0); + } + /* the socket is not closed, so SOF_MP_SUBFLOW must be set */ + VERIFY(so->so_flags & SOF_MP_SUBFLOW); + + if (m->m_nextpkt != NULL || (sb->sb_mb != sb->sb_lastrecord)) { + panic("%s: nexpkt %p || mb %p != lastrecord %p\n", __func__, + m->m_nextpkt, sb->sb_mb, sb->sb_lastrecord); + /* NOTREACHED */ + } + + SBLASTMBUFCHK(sb, __func__); + + if (mptcp_adj_rmap(so, m) != 0) + return (0); + + /* No filter support (SB_RECV) on mptcp subflow sockets */ + + sbcompress(sb, m, sb->sb_mbtail); + sb->sb_lastrecord = sb->sb_mb; + SBLASTRECORDCHK(sb, __func__); + return (1); +} +#endif /* MPTCP */ + +/* + * Append message to send socket buffer based on priority. + */ +int +sbappendmsg_snd(struct sockbuf *sb, struct mbuf *m) { - register int eor = 0; - register struct mbuf *o; + struct socket *so = sb->sb_so; + struct msg_priq *priq; + int set_eor = 0; + + VERIFY(so->so_msg_state != NULL); + + if (m->m_nextpkt != NULL || (sb->sb_mb != sb->sb_lastrecord)) + panic("sbappendstream: nexpkt %p || mb %p != lastrecord %p\n", + m->m_nextpkt, sb->sb_mb, sb->sb_lastrecord); + + SBLASTMBUFCHK(sb, __func__); + + if (m == NULL || (sb->sb_flags & SB_DROP) || so->so_msg_state == NULL) { + if (m != NULL) + m_freem(m); + return (0); + } + + priq = &so->so_msg_state->msg_priq[m->m_pkthdr.msg_pri]; + + /* note if we need to propogate M_EOR to the last mbuf */ + if (m->m_flags & M_EOR) { + set_eor = 1; + + /* Reset M_EOR from the first mbuf */ + m->m_flags &= ~(M_EOR); + } + + if (priq->msgq_head == NULL) { + VERIFY(priq->msgq_tail == NULL && priq->msgq_lastmsg == NULL); + priq->msgq_head = priq->msgq_lastmsg = m; + } else { + VERIFY(priq->msgq_tail->m_next == NULL); + + /* Check if the last message has M_EOR flag set */ + if (priq->msgq_tail->m_flags & M_EOR) { + /* Insert as a new message */ + priq->msgq_lastmsg->m_nextpkt = m; + + /* move the lastmsg pointer */ + priq->msgq_lastmsg = m; + } else { + /* Append to the existing message */ + priq->msgq_tail->m_next = m; + } + } + + /* Update accounting and the queue tail pointer */ + + while (m->m_next != NULL) { + sballoc(sb, m); + priq->msgq_bytes += m->m_len; + m = m->m_next; + } + sballoc(sb, m); + priq->msgq_bytes += m->m_len; + + if (set_eor) { + m->m_flags |= M_EOR; + + /* + * Since the user space can not write a new msg + * without completing the previous one, we can + * reset this flag to start sending again. + */ + priq->msgq_flags &= ~(MSGQ_MSG_NOTDONE); + } + + priq->msgq_tail = m; + + SBLASTRECORDCHK(sb, "sbappendstream 2"); + postevent(0, sb, EV_RWBYTES); + return (1); +} + +/* + * Pull data from priority queues to the serial snd queue + * right before sending. + */ +void +sbpull_unordered_data(struct socket *so, int32_t off, int32_t len) +{ + int32_t topull, i; + struct msg_priq *priq = NULL; + + VERIFY(so->so_msg_state != NULL); + + topull = (off + len) - so->so_msg_state->msg_serial_bytes; + + i = MSG_PRI_MAX; + while (i >= MSG_PRI_MIN && topull > 0) { + struct mbuf *m = NULL, *mqhead = NULL, *mend = NULL; + priq = &so->so_msg_state->msg_priq[i]; + if ((priq->msgq_flags & MSGQ_MSG_NOTDONE) && + priq->msgq_head == NULL) { + /* + * We were in the middle of sending + * a message and we have not seen the + * end of it. + */ + VERIFY(priq->msgq_lastmsg == NULL && + priq->msgq_tail == NULL); + return; + } + if (priq->msgq_head != NULL) { + int32_t bytes = 0, topull_tmp = topull; + /* + * We found a msg while scanning the priority + * queue from high to low priority. + */ + m = priq->msgq_head; + mqhead = m; + mend = m; + + /* + * Move bytes from the priority queue to the + * serial queue. Compute the number of bytes + * being added. + */ + while (mqhead->m_next != NULL && topull_tmp > 0) { + bytes += mqhead->m_len; + topull_tmp -= mqhead->m_len; + mend = mqhead; + mqhead = mqhead->m_next; + } + + if (mqhead->m_next == NULL) { + /* + * If we have only one more mbuf left, + * move the last mbuf of this message to + * serial queue and set the head of the + * queue to be the next message. + */ + bytes += mqhead->m_len; + mend = mqhead; + mqhead = m->m_nextpkt; + if (!(mend->m_flags & M_EOR)) { + /* + * We have not seen the end of + * this message, so we can not + * pull anymore. + */ + priq->msgq_flags |= MSGQ_MSG_NOTDONE; + } else { + /* Reset M_EOR */ + mend->m_flags &= ~(M_EOR); + } + } else { + /* propogate the next msg pointer */ + mqhead->m_nextpkt = m->m_nextpkt; + } + priq->msgq_head = mqhead; + + /* + * if the lastmsg pointer points to + * the mbuf that is being dequeued, update + * it to point to the new head. + */ + if (priq->msgq_lastmsg == m) + priq->msgq_lastmsg = priq->msgq_head; + + m->m_nextpkt = NULL; + mend->m_next = NULL; + + if (priq->msgq_head == NULL) { + /* Moved all messages, update tail */ + priq->msgq_tail = NULL; + VERIFY(priq->msgq_lastmsg == NULL); + } + + /* Move it to serial sb_mb queue */ + if (so->so_snd.sb_mb == NULL) { + so->so_snd.sb_mb = m; + } else { + so->so_snd.sb_mbtail->m_next = m; + } + + priq->msgq_bytes -= bytes; + VERIFY(priq->msgq_bytes >= 0); + sbwakeup(&so->so_snd); - while (m) { + so->so_msg_state->msg_serial_bytes += bytes; + so->so_snd.sb_mbtail = mend; + so->so_snd.sb_lastrecord = so->so_snd.sb_mb; + + topull = + (off + len) - so->so_msg_state->msg_serial_bytes; + + if (priq->msgq_flags & MSGQ_MSG_NOTDONE) + break; + } else { + --i; + } + } + sblastrecordchk(&so->so_snd, "sbpull_unordered_data"); + sblastmbufchk(&so->so_snd, "sbpull_unordered_data"); +} + +/* + * Compress mbuf chain m into the socket + * buffer sb following mbuf n. If n + * is null, the buffer is presumed empty. + */ +static inline void +sbcompress(struct sockbuf *sb, struct mbuf *m, struct mbuf *n) +{ + int eor = 0, compress = (!(sb->sb_flags & SB_NOCOMPRESS)); + struct mbuf *o; + + if (m == NULL) { + /* There is nothing to compress; just update the tail */ + for (; n->m_next != NULL; n = n->m_next) + ; + sb->sb_mbtail = n; + goto done; + } + + while (m != NULL) { eor |= m->m_flags & M_EOR; - if (m->m_len == 0 && - (eor == 0 || - (((o = m->m_next) || (o = n)) && - o->m_type == m->m_type))) { + if (compress && m->m_len == 0 && (eor == 0 || + (((o = m->m_next) || (o = n)) && o->m_type == m->m_type))) { + if (sb->sb_lastrecord == m) + sb->sb_lastrecord = m->m_next; m = m_free(m); continue; } - if (n && (n->m_flags & M_EOR) == 0 && + if (compress && n != NULL && (n->m_flags & M_EOR) == 0 && #ifndef __APPLE__ M_WRITABLE(n) && #endif @@ -1039,27 +1657,58 @@ sbcompress(sb, m, n) (unsigned)m->m_len); n->m_len += m->m_len; sb->sb_cc += m->m_len; + if (m->m_type != MT_DATA && m->m_type != MT_HEADER && + m->m_type != MT_OOBDATA) { + /* XXX: Probably don't need */ + sb->sb_ctl += m->m_len; + } m = m_free(m); continue; } - if (n) + if (n != NULL) n->m_next = m; else sb->sb_mb = m; + sb->sb_mbtail = m; sballoc(sb, m); n = m; m->m_flags &= ~M_EOR; m = m->m_next; - n->m_next = 0; + n->m_next = NULL; } - if (eor) { - if (n) + if (eor != 0) { + if (n != NULL) n->m_flags |= eor; else printf("semi-panic: sbcompress\n"); } - postevent(0,sb, EV_RWBYTES); - return 1; +done: + SBLASTMBUFCHK(sb, __func__); + postevent(0, sb, EV_RWBYTES); +} + +void +sb_empty_assert(struct sockbuf *sb, const char *where) +{ + if (!(sb->sb_cc == 0 && sb->sb_mb == NULL && sb->sb_mbcnt == 0 && + sb->sb_mbtail == NULL && sb->sb_lastrecord == NULL)) { + panic("%s: sb %p so %p cc %d mbcnt %d mb %p mbtail %p " + "lastrecord %p\n", where, sb, sb->sb_so, sb->sb_cc, + sb->sb_mbcnt, sb->sb_mb, sb->sb_mbtail, + sb->sb_lastrecord); + /* NOTREACHED */ + } +} + +static void +sbflush_priq(struct msg_priq *priq) +{ + struct mbuf *m; + m = priq->msgq_head; + if (m != NULL) + m_freem_list(m); + priq->msgq_head = priq->msgq_tail = priq->msgq_lastmsg = NULL; + priq->msgq_bytes = priq->msgq_flags = 0; } /* @@ -1067,13 +1716,51 @@ sbcompress(sb, m, n) * Check that all resources are reclaimed. */ void -sbflush(sb) - register struct sockbuf *sb; +sbflush(struct sockbuf *sb) { - 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) { + void *lr_saved = __builtin_return_address(0); + struct socket *so = sb->sb_so; +#ifdef notyet + lck_mtx_t *mutex_held; +#endif + u_int32_t i; + + /* so_usecount may be 0 if we get here from sofreelastref() */ + if (so == NULL) { + panic("%s: null so, sb=%p sb_flags=0x%x lr=%p\n", + __func__, sb, sb->sb_flags, lr_saved); + /* NOTREACHED */ + } else if (so->so_usecount < 0) { + panic("%s: sb=%p sb_flags=0x%x sb_so=%p usecount=%d lr=%p " + "lrh= %s\n", __func__, sb, sb->sb_flags, so, + so->so_usecount, lr_saved, solockhistory_nr(so)); + /* NOTREACHED */ + } +#ifdef notyet + /* + * XXX: This code is currently commented out, because we may get here + * as part of sofreelastref(), and at that time, pr_getlock() may no + * longer be able to return us the lock; this will be fixed in future. + */ + 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; + + lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); +#endif + + /* + * Obtain lock on the socket buffer (SB_LOCK). This is required + * to prevent the socket buffer from being unexpectedly altered + * while it is used by another thread in socket send/receive. + * + * sblock() must not fail here, hence the assertion. + */ + (void) sblock(sb, SBL_WAIT | SBL_NOINTR | SBL_IGNDEFUNCT); + VERIFY(sb->sb_flags & SB_LOCK); + + while (sb->sb_mbcnt > 0) { /* * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty: * we would loop forever. Panic instead. @@ -1082,12 +1769,20 @@ sbflush(sb) 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_RECV) && (so->so_flags & SOF_ENABLE_MSGS)) { + VERIFY(so->so_msg_state != NULL); + for (i = MSG_PRI_MIN; i <= MSG_PRI_MAX; ++i) { + sbflush_priq(&so->so_msg_state->msg_priq[i]); + } + so->so_msg_state->msg_serial_bytes = 0; + so->so_msg_state->msg_uno_bytes = 0; + } + + sb_empty_assert(sb, __func__); postevent(0, sb, EV_RWBYTES); - sbunlock(sb, 1); /* keep socket locked */ + sbunlock(sb, TRUE); /* keep socket locked */ } /* @@ -1102,43 +1797,64 @@ sbflush(sb) * m_freem_list. */ void -sbdrop(sb, len) - register struct sockbuf *sb; - register int len; +sbdrop(struct sockbuf *sb, int len) { - register struct mbuf *m, *free_list, *ml; + struct mbuf *m, *free_list, *ml; struct mbuf *next, *last; + next = (m = sb->sb_mb) ? m->m_nextpkt : 0; +#if MPTCP + if ((m != NULL) && (len > 0) && + (!(sb->sb_flags & SB_RECV)) && + ((sb->sb_so->so_flags & SOF_MP_SUBFLOW) || + ((SOCK_CHECK_DOM(sb->sb_so, PF_MULTIPATH)) && + (SOCK_CHECK_PROTO(sb->sb_so, IPPROTO_TCP)))) && + (!(sb->sb_so->so_flags1 & SOF1_POST_FALLBACK_SYNC))) { + mptcp_preproc_sbdrop(m, (unsigned int)len); + } +#endif /* MPTCP */ 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) { - /* 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 (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; + if (!(sb->sb_flags & SB_RECV) && + (sb->sb_so->so_flags & SOF_ENABLE_MSGS)) { + sb->sb_so->so_msg_state-> + msg_serial_bytes = 0; + } + break; + } + m = last = next; + next = m->m_nextpkt; + continue; } if (m->m_len > len) { m->m_len -= len; m->m_data += len; sb->sb_cc -= len; + if (m->m_type != MT_DATA && m->m_type != MT_HEADER && + m->m_type != MT_OOBDATA) + sb->sb_ctl -= len; break; } len -= m->m_len; @@ -1154,15 +1870,33 @@ sbdrop(sb, len) m = m->m_next; } if (ml) { - ml->m_next = (struct mbuf *)0; + ml->m_next = (struct mbuf *)0; last->m_nextpkt = (struct mbuf *)0; - m_freem_list(free_list); + m_freem_list(free_list); } if (m) { sb->sb_mb = m; m->m_nextpkt = next; - } else + } else { sb->sb_mb = next; + } + + /* + * First part is an inline SB_EMPTY_FIXUP(). Second part + * makes sure sb_lastrecord is up-to-date if we dropped + * part of the last record. + */ + m = sb->sb_mb; + if (m == NULL) { + sb->sb_mbtail = NULL; + sb->sb_lastrecord = NULL; + } else if (m->m_nextpkt == NULL) { + sb->sb_lastrecord = m; + } + +#if CONTENT_FILTER + cfil_sock_buf_update(sb); +#endif /* CONTENT_FILTER */ postevent(0, sb, EV_RWBYTES); @@ -1174,10 +1908,9 @@ sbdrop(sb, len) * and move the next record to the front. */ void -sbdroprecord(sb) - register struct sockbuf *sb; +sbdroprecord(struct sockbuf *sb) { - register struct mbuf *m, *mn; + struct mbuf *m, *mn; m = sb->sb_mb; if (m) { @@ -1188,6 +1921,7 @@ sbdroprecord(sb) m = mn; } while (m); } + SB_EMPTY_FIXUP(sb); postevent(0, sb, EV_RWBYTES); } @@ -1196,21 +1930,19 @@ sbdroprecord(sb) * with the specified type for presentation on a socket buffer. */ struct mbuf * -sbcreatecontrol(p, size, type, level) - caddr_t p; - register int size; - int type, level; +sbcreatecontrol(caddr_t p, int size, int type, int level) { - register struct cmsghdr *cp; + struct cmsghdr *cp; struct mbuf *m; if (CMSG_SPACE((u_int)size) > MLEN) - return ((struct mbuf *) NULL); + return ((struct mbuf *)NULL); if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL) - return ((struct mbuf *) NULL); + return ((struct mbuf *)NULL); cp = mtod(m, struct cmsghdr *); + VERIFY(IS_P2ALIGNED(cp, sizeof (u_int32_t))); /* XXX check size? */ - (void)memcpy(CMSG_DATA(cp), p, size); + (void) memcpy(CMSG_DATA(cp), p, size); m->m_len = CMSG_SPACE(size); cp->cmsg_len = CMSG_LEN(size); cp->cmsg_level = level; @@ -1218,6 +1950,40 @@ sbcreatecontrol(p, size, type, level) return (m); } +struct mbuf ** +sbcreatecontrol_mbuf(caddr_t p, int size, int type, int level, struct mbuf **mp) +{ + struct mbuf *m; + struct cmsghdr *cp; + + if (*mp == NULL) { + *mp = sbcreatecontrol(p, size, type, level); + return (mp); + } + + if (CMSG_SPACE((u_int)size) + (*mp)->m_len > MLEN) { + mp = &(*mp)->m_next; + *mp = sbcreatecontrol(p, size, type, level); + return (mp); + } + + m = *mp; + + cp = (struct cmsghdr *)(void *)(mtod(m, char *) + m->m_len); + /* CMSG_SPACE ensures 32-bit alignment */ + VERIFY(IS_P2ALIGNED(cp, sizeof (u_int32_t))); + m->m_len += CMSG_SPACE(size); + + /* XXX check size? */ + (void) memcpy(CMSG_DATA(cp), p, size); + cp->cmsg_len = CMSG_LEN(size); + cp->cmsg_level = level; + cp->cmsg_type = type; + + return (mp); +} + + /* * Some routines that return EOPNOTSUPP for entry points that are not * supported by a protocol. Fill in as needed. @@ -1225,159 +1991,255 @@ sbcreatecontrol(p, size, type, level) int pru_abort_notsupp(struct socket *so) { - return EOPNOTSUPP; +#pragma unused(so) + return (EOPNOTSUPP); } - int pru_accept_notsupp(struct socket *so, struct sockaddr **nam) { - return EOPNOTSUPP; +#pragma unused(so, nam) + return (EOPNOTSUPP); } int pru_attach_notsupp(struct socket *so, int proto, struct proc *p) { - return EOPNOTSUPP; +#pragma unused(so, proto, p) + return (EOPNOTSUPP); } int pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p) { - return EOPNOTSUPP; +#pragma unused(so, nam, p) + return (EOPNOTSUPP); } int pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p) { - return EOPNOTSUPP; +#pragma unused(so, nam, p) + return (EOPNOTSUPP); } int pru_connect2_notsupp(struct socket *so1, struct socket *so2) { - return EOPNOTSUPP; +#pragma unused(so1, so2) + return (EOPNOTSUPP); +} + +int +pru_connectx_notsupp(struct socket *so, struct sockaddr_list **src_sl, + struct sockaddr_list **dst_sl, struct proc *p, uint32_t ifscope, + associd_t aid, connid_t *pcid, uint32_t flags, void *arg, + uint32_t arglen) +{ +#pragma unused(so, src_sl, dst_sl, p, ifscope, aid, pcid, flags, arg, arglen) + return (EOPNOTSUPP); } int pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data, - struct ifnet *ifp, struct proc *p) + struct ifnet *ifp, struct proc *p) { - return EOPNOTSUPP; +#pragma unused(so, cmd, data, ifp, p) + return (EOPNOTSUPP); } int pru_detach_notsupp(struct socket *so) { - return EOPNOTSUPP; +#pragma unused(so) + return (EOPNOTSUPP); } int pru_disconnect_notsupp(struct socket *so) { - return EOPNOTSUPP; +#pragma unused(so) + return (EOPNOTSUPP); +} + +int +pru_disconnectx_notsupp(struct socket *so, associd_t aid, connid_t cid) +{ +#pragma unused(so, aid, cid) + return (EOPNOTSUPP); } int pru_listen_notsupp(struct socket *so, struct proc *p) { - return EOPNOTSUPP; +#pragma unused(so, p) + return (EOPNOTSUPP); +} + +int +pru_peeloff_notsupp(struct socket *so, associd_t aid, struct socket **psop) +{ +#pragma unused(so, aid, psop) + return (EOPNOTSUPP); } int pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam) { - return EOPNOTSUPP; +#pragma unused(so, nam) + return (EOPNOTSUPP); } int pru_rcvd_notsupp(struct socket *so, int flags) { - return EOPNOTSUPP; +#pragma unused(so, flags) + return (EOPNOTSUPP); } int pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags) { - return EOPNOTSUPP; +#pragma unused(so, m, flags) + return (EOPNOTSUPP); } int pru_send_notsupp(struct socket *so, int flags, struct mbuf *m, - struct sockaddr *addr, struct mbuf *control, - struct proc *p) - + struct sockaddr *addr, struct mbuf *control, struct proc *p) { - return EOPNOTSUPP; +#pragma unused(so, flags, m, addr, control, p) + return (EOPNOTSUPP); } +int +pru_send_list_notsupp(struct socket *so, int flags, struct mbuf *m, + struct sockaddr *addr, struct mbuf *control, struct proc *p) +{ +#pragma unused(so, flags, m, addr, control, p) + return (EOPNOTSUPP); +} /* * This isn't really a ``null'' operation, but it's the default one * and doesn't do anything destructive. */ int -pru_sense_null(struct socket *so, struct stat *sb) +pru_sense_null(struct socket *so, void *ub, int isstat64) { - sb->st_blksize = so->so_snd.sb_hiwat; - return 0; -} + if (isstat64 != 0) { + struct stat64 *sb64; + sb64 = (struct stat64 *)ub; + sb64->st_blksize = so->so_snd.sb_hiwat; + } else { + struct stat *sb; -int pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, - struct uio *uio, struct mbuf *top, - struct mbuf *control, int flags) + sb = (struct stat *)ub; + sb->st_blksize = so->so_snd.sb_hiwat; + } -{ - return EOPNOTSUPP; + return (0); } -int pru_soreceive_notsupp(struct socket *so, - struct sockaddr **paddr, - struct uio *uio, struct mbuf **mp0, - struct mbuf **controlp, int *flagsp) + +int +pru_sosend_notsupp(struct socket *so, struct sockaddr *addr, struct uio *uio, + struct mbuf *top, struct mbuf *control, int flags) { - return EOPNOTSUPP; +#pragma unused(so, addr, uio, top, control, flags) + return (EOPNOTSUPP); } int +pru_sosend_list_notsupp(struct socket *so, struct sockaddr *addr, struct uio **uio, + u_int uiocnt, struct mbuf *top, struct mbuf *control, int flags) +{ +#pragma unused(so, addr, uio, uiocnt, top, control, flags) + return (EOPNOTSUPP); +} -pru_shutdown_notsupp(struct socket *so) +int +pru_soreceive_notsupp(struct socket *so, struct sockaddr **paddr, + struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { - return EOPNOTSUPP; +#pragma unused(so, paddr, uio, mp0, controlp, flagsp) + return (EOPNOTSUPP); } int -pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam) +pru_soreceive_list_notsupp(struct socket *so, struct sockaddr **paddr, + struct uio **uio, u_int uiocnt, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { - return EOPNOTSUPP; +#pragma unused(so, paddr, uio, uiocnt, mp0, controlp, flagsp) + return (EOPNOTSUPP); } -int pru_sosend(struct socket *so, struct sockaddr *addr, - struct uio *uio, struct mbuf *top, - struct mbuf *control, int flags) +int +pru_shutdown_notsupp(struct socket *so) { - return EOPNOTSUPP; +#pragma unused(so) + return (EOPNOTSUPP); } -int pru_soreceive(struct socket *so, - struct sockaddr **paddr, - struct uio *uio, struct mbuf **mp0, - struct mbuf **controlp, int *flagsp) +int +pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam) { - return EOPNOTSUPP; +#pragma unused(so, nam) + return (EOPNOTSUPP); } +int +pru_sopoll_notsupp(struct socket *so, int events, kauth_cred_t cred, void *wql) +{ +#pragma unused(so, events, cred, wql) + return (EOPNOTSUPP); +} int -pru_sopoll_notsupp(__unused struct socket *so, __unused int events, - __unused kauth_cred_t cred, __unused void *wql) +pru_socheckopt_null(struct socket *so, struct sockopt *sopt) { - return EOPNOTSUPP; +#pragma unused(so, sopt) + /* + * Allow all options for set/get by default. + */ + return (0); } +void +pru_sanitize(struct pr_usrreqs *pru) +{ +#define DEFAULT(foo, bar) if ((foo) == NULL) (foo) = (bar) + DEFAULT(pru->pru_abort, pru_abort_notsupp); + DEFAULT(pru->pru_accept, pru_accept_notsupp); + DEFAULT(pru->pru_attach, pru_attach_notsupp); + DEFAULT(pru->pru_bind, pru_bind_notsupp); + DEFAULT(pru->pru_connect, pru_connect_notsupp); + DEFAULT(pru->pru_connect2, pru_connect2_notsupp); + DEFAULT(pru->pru_connectx, pru_connectx_notsupp); + DEFAULT(pru->pru_control, pru_control_notsupp); + DEFAULT(pru->pru_detach, pru_detach_notsupp); + DEFAULT(pru->pru_disconnect, pru_disconnect_notsupp); + DEFAULT(pru->pru_disconnectx, pru_disconnectx_notsupp); + DEFAULT(pru->pru_listen, pru_listen_notsupp); + DEFAULT(pru->pru_peeloff, pru_peeloff_notsupp); + DEFAULT(pru->pru_peeraddr, pru_peeraddr_notsupp); + DEFAULT(pru->pru_rcvd, pru_rcvd_notsupp); + DEFAULT(pru->pru_rcvoob, pru_rcvoob_notsupp); + DEFAULT(pru->pru_send, pru_send_notsupp); + DEFAULT(pru->pru_send_list, pru_send_list_notsupp); + DEFAULT(pru->pru_sense, pru_sense_null); + DEFAULT(pru->pru_shutdown, pru_shutdown_notsupp); + DEFAULT(pru->pru_sockaddr, pru_sockaddr_notsupp); + DEFAULT(pru->pru_sopoll, pru_sopoll_notsupp); + DEFAULT(pru->pru_soreceive, pru_soreceive_notsupp); + DEFAULT(pru->pru_soreceive_list, pru_soreceive_list_notsupp); + DEFAULT(pru->pru_sosend, pru_sosend_notsupp); + DEFAULT(pru->pru_sosend_list, pru_sosend_list_notsupp); + DEFAULT(pru->pru_socheckopt, pru_socheckopt_null); +#undef DEFAULT +} -#ifdef __APPLE__ /* * The following are macros on BSD and functions on Darwin */ @@ -1386,39 +2248,83 @@ pru_sopoll_notsupp(__unused struct socket *so, __unused int events, * Do we need to notify the other side when I/O is possible? */ -int +int sb_notify(struct sockbuf *sb) { - return ((sb->sb_flags & (SB_WAIT|SB_SEL|SB_ASYNC|SB_UPCALL|SB_KNOTE)) != 0); + return (sb->sb_waiters > 0 || + (sb->sb_flags & (SB_SEL|SB_ASYNC|SB_UPCALL|SB_KNOTE))); } /* * 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. + * overflow and return 0. */ -long +int sbspace(struct sockbuf *sb) { - return ((long) imin((int)(sb->sb_hiwat - sb->sb_cc), - (int)(sb->sb_mbmax - sb->sb_mbcnt))); + int pending = 0; + int space = imin((int)(sb->sb_hiwat - sb->sb_cc), + (int)(sb->sb_mbmax - sb->sb_mbcnt)); + if (space < 0) + space = 0; + + /* Compensate for data being processed by content filters */ +#if CONTENT_FILTER + pending = cfil_sock_data_space(sb); +#endif /* CONTENT_FILTER */ + if (pending > space) + space = 0; + else + space -= pending; + + return (space); +} + +/* + * If this socket has priority queues, check if there is enough + * space in the priority queue for this msg. + */ +int +msgq_sbspace(struct socket *so, struct mbuf *control) +{ + int space = 0, error; + u_int32_t msgpri; + VERIFY(so->so_type == SOCK_STREAM && + SOCK_PROTO(so) == IPPROTO_TCP); + if (control != NULL) { + error = tcp_get_msg_priority(control, &msgpri); + if (error) + return (0); + } else { + msgpri = MSG_PRI_0; + } + space = (so->so_snd.sb_idealsize / MSG_PRI_COUNT) - + so->so_msg_state->msg_priq[msgpri].msgq_bytes; + if (space < 0) + space = 0; + return (space); } /* do we have to send all at once on a socket? */ int sosendallatonce(struct socket *so) { - return (so->so_proto->pr_flags & PR_ATOMIC); + 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); + return (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat || + ((so->so_state & SS_CANTRCVMORE) +#if CONTENT_FILTER + && cfil_sock_data_pending(&so->so_rcv) == 0 +#endif /* CONTENT_FILTER */ + ) || + so->so_comp.tqh_first || so->so_error); } /* can we write something to so? */ @@ -1426,11 +2332,34 @@ soreadable(struct socket *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); + if ((so->so_state & SS_CANTSENDMORE) || + so->so_error > 0) + return (1); + + if (!so_wait_for_if_feedback(so) && + sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat && + ((so->so_state & SS_ISCONNECTED) || + !(so->so_proto->pr_flags & PR_CONNREQUIRED))) { + if (so->so_flags & SOF_NOTSENT_LOWAT) { + if ((SOCK_DOM(so) == PF_INET6 + || SOCK_DOM(so) == PF_INET) + && so->so_type == SOCK_STREAM) { + return (tcp_notsent_lowat_check(so)); + } +#if MPTCP + else if ((SOCK_DOM(so) == PF_MULTIPATH) && + (SOCK_PROTO(so) == IPPROTO_TCP)) { + return (mptcp_notsent_lowat_check(so)); + } +#endif + else { + return (1); + } + } else { + return (1); + } + } + return (0); } /* adjust counters in sb reflecting allocation of m */ @@ -1438,20 +2367,40 @@ sowriteable(struct socket *so) 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; + u_int32_t cnt = 1; + sb->sb_cc += m->m_len; + if (m->m_type != MT_DATA && m->m_type != MT_HEADER && + m->m_type != MT_OOBDATA) + sb->sb_ctl += m->m_len; + sb->sb_mbcnt += MSIZE; + + if (m->m_flags & M_EXT) { + sb->sb_mbcnt += m->m_ext.ext_size; + cnt += (m->m_ext.ext_size >> MSIZESHIFT); + } + OSAddAtomic(cnt, &total_sbmb_cnt); + VERIFY(total_sbmb_cnt > 0); + if (total_sbmb_cnt > total_sbmb_cnt_peak) + total_sbmb_cnt_peak = total_sbmb_cnt; } /* 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; + int cnt = -1; + + sb->sb_cc -= m->m_len; + if (m->m_type != MT_DATA && m->m_type != MT_HEADER && + m->m_type != MT_OOBDATA) + sb->sb_ctl -= m->m_len; + sb->sb_mbcnt -= MSIZE; + if (m->m_flags & M_EXT) { + sb->sb_mbcnt -= m->m_ext.ext_size; + cnt -= (m->m_ext.ext_size >> MSIZESHIFT); + } + OSAddAtomic(cnt, &total_sbmb_cnt); + VERIFY(total_sbmb_cnt >= 0); } /* @@ -1460,81 +2409,296 @@ sbfree(struct sockbuf *sb, struct mbuf *m) * Returns error without lock if sleep is interrupted. */ int -sblock(struct sockbuf *sb, int wf) +sblock(struct sockbuf *sb, uint32_t flags) { - return(sb->sb_flags & SB_LOCK ? - ((wf == M_WAIT) ? sb_lock(sb) : EWOULDBLOCK) : - (sb->sb_flags |= SB_LOCK), 0); + boolean_t nointr = ((sb->sb_flags & SB_NOINTR) || (flags & SBL_NOINTR)); + void *lr_saved = __builtin_return_address(0); + struct socket *so = sb->sb_so; + void * wchan; + int error = 0; + thread_t tp = current_thread(); + + VERIFY((flags & SBL_VALID) == flags); + + /* so_usecount may be 0 if we get here from sofreelastref() */ + if (so == NULL) { + panic("%s: null so, sb=%p sb_flags=0x%x lr=%p\n", + __func__, sb, sb->sb_flags, lr_saved); + /* NOTREACHED */ + } else if (so->so_usecount < 0) { + panic("%s: sb=%p sb_flags=0x%x sb_so=%p usecount=%d lr=%p " + "lrh= %s\n", __func__, sb, sb->sb_flags, so, + so->so_usecount, lr_saved, solockhistory_nr(so)); + /* NOTREACHED */ + } + + /* + * The content filter thread must hold the sockbuf lock + */ + if ((so->so_flags & SOF_CONTENT_FILTER) && sb->sb_cfil_thread == tp) { + /* + * Don't panic if we are defunct because SB_LOCK has + * been cleared by sodefunct() + */ + if (!(so->so_flags & SOF_DEFUNCT) && !(sb->sb_flags & SB_LOCK)) + panic("%s: SB_LOCK not held for %p\n", + __func__, sb); + + /* Keep the sockbuf locked */ + return (0); + } + + if ((sb->sb_flags & SB_LOCK) && !(flags & SBL_WAIT)) + return (EWOULDBLOCK); + /* + * We may get here from sorflush(), in which case "sb" may not + * point to the real socket buffer. Use the actual socket buffer + * address from the socket instead. + */ + wchan = (sb->sb_flags & SB_RECV) ? + &so->so_rcv.sb_flags : &so->so_snd.sb_flags; + + /* + * A content filter thread has exclusive access to the sockbuf + * until it clears the + */ + while ((sb->sb_flags & SB_LOCK) || + ((so->so_flags & SOF_CONTENT_FILTER) && + sb->sb_cfil_thread != NULL)) { + lck_mtx_t *mutex_held; + + /* + * XXX: This code should be moved up above outside of this loop; + * however, we may get here as part of sofreelastref(), and + * at that time pr_getlock() may no longer be able to return + * us the lock. This will be fixed in future. + */ + 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; + + lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); + + sb->sb_wantlock++; + VERIFY(sb->sb_wantlock != 0); + + error = msleep(wchan, mutex_held, + nointr ? PSOCK : PSOCK | PCATCH, + nointr ? "sb_lock_nointr" : "sb_lock", NULL); + + VERIFY(sb->sb_wantlock != 0); + sb->sb_wantlock--; + + if (error == 0 && (so->so_flags & SOF_DEFUNCT) && + !(flags & SBL_IGNDEFUNCT)) { + error = EBADF; + SODEFUNCTLOG(("%s[%d]: defunct so 0x%llx [%d,%d] " + "(%d)\n", __func__, proc_selfpid(), + (uint64_t)VM_KERNEL_ADDRPERM(so), + SOCK_DOM(so), SOCK_TYPE(so), error)); + } + + if (error != 0) + return (error); + } + sb->sb_flags |= SB_LOCK; + return (0); } -/* release lock on sockbuf sb */ +/* + * Release lock on sockbuf sb + */ void -sbunlock(struct sockbuf *sb, int keeplocked) +sbunlock(struct sockbuf *sb, boolean_t keeplocked) { + void *lr_saved = __builtin_return_address(0); struct socket *so = sb->sb_so; - int lr, lr_saved; - lck_mtx_t *mutex_held; + thread_t tp = current_thread(); + + /* so_usecount may be 0 if we get here from sofreelastref() */ + if (so == NULL) { + panic("%s: null so, sb=%p sb_flags=0x%x lr=%p\n", + __func__, sb, sb->sb_flags, lr_saved); + /* NOTREACHED */ + } else if (so->so_usecount < 0) { + panic("%s: sb=%p sb_flags=0x%x sb_so=%p usecount=%d lr=%p " + "lrh= %s\n", __func__, sb, sb->sb_flags, so, + so->so_usecount, lr_saved, solockhistory_nr(so)); + /* NOTREACHED */ + } -#ifdef __ppc__ - __asm__ volatile("mflr %0" : "=r" (lr)); - lr_saved = lr; -#endif - sb->sb_flags &= ~SB_LOCK; + /* + * The content filter thread must hold the sockbuf lock + */ + if ((so->so_flags & SOF_CONTENT_FILTER) && sb->sb_cfil_thread == tp) { + /* + * Don't panic if we are defunct because SB_LOCK has + * been cleared by sodefunct() + */ + if (!(so->so_flags & SOF_DEFUNCT) && + !(sb->sb_flags & SB_LOCK) && + !(so->so_state & SS_DEFUNCT) && + !(so->so_flags1 & SOF1_DEFUNCTINPROG)) { + panic("%s: SB_LOCK not held for %p\n", + __func__, sb); + } + /* Keep the sockbuf locked and proceed*/ + } else { + VERIFY((sb->sb_flags & SB_LOCK) || + (so->so_state & SS_DEFUNCT) || + (so->so_flags1 & SOF1_DEFUNCTINPROG)); + + sb->sb_flags &= ~SB_LOCK; + + if (sb->sb_wantlock > 0) { + /* + * We may get here from sorflush(), in which case "sb" may not + * point to the real socket buffer. Use the actual socket + * buffer address from the socket instead. + */ + wakeup((sb->sb_flags & SB_RECV) ? &so->so_rcv.sb_flags : + &so->so_snd.sb_flags); + } + } - 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) { /* unlock on exit */ + lck_mtx_t *mutex_held; - if (keeplocked == 0) - lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); + 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 (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); + lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED); - wakeup((caddr_t)&(sb)->sb_flags); - } - if (keeplocked == 0) { /* unlock on exit */ + VERIFY(so->so_usecount != 0); 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->reserved4= lr_saved; + so->unlock_lr[so->next_unlock_lr] = lr_saved; + so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX; lck_mtx_unlock(mutex_held); } } void -sorwakeup(struct socket * so) +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); +} + +void +soevent(struct socket *so, long hint) { - if (sb_notify(&so->so_rcv)) - sowakeup(so, &so->so_rcv); + if (so->so_flags & SOF_KNOTE) + KNOTE(&so->so_klist, hint); + + soevupcall(so, hint); + + /* + * Don't post an event if this a subflow socket or + * the app has opted out of using cellular interface + */ + if ((hint & SO_FILT_HINT_IFDENIED) && + !(so->so_flags & SOF_MP_SUBFLOW) && + !(so->so_restrictions & SO_RESTRICT_DENY_CELLULAR) && + !(so->so_restrictions & SO_RESTRICT_DENY_EXPENSIVE)) + soevent_ifdenied(so); } void -sowwakeup(struct socket * so) +soevupcall(struct socket *so, u_int32_t hint) { - if (sb_notify(&so->so_snd)) - sowakeup(so, &so->so_snd); + if (so->so_event != NULL) { + caddr_t so_eventarg = so->so_eventarg; + + hint &= so->so_eventmask; + if (hint != 0) { + socket_unlock(so, 0); + so->so_event(so, so_eventarg, hint); + socket_lock(so, 0); + } + } +} + +static void +soevent_ifdenied(struct socket *so) +{ + struct kev_netpolicy_ifdenied ev_ifdenied; + + bzero(&ev_ifdenied, sizeof (ev_ifdenied)); + /* + * The event consumer is interested about the effective {upid,pid,uuid} + * info which can be different than the those related to the process + * that recently performed a system call on the socket, i.e. when the + * socket is delegated. + */ + if (so->so_flags & SOF_DELEGATED) { + ev_ifdenied.ev_data.eupid = so->e_upid; + ev_ifdenied.ev_data.epid = so->e_pid; + uuid_copy(ev_ifdenied.ev_data.euuid, so->e_uuid); + } else { + ev_ifdenied.ev_data.eupid = so->last_upid; + ev_ifdenied.ev_data.epid = so->last_pid; + uuid_copy(ev_ifdenied.ev_data.euuid, so->last_uuid); + } + + if (++so->so_ifdenied_notifies > 1) { + /* + * Allow for at most one kernel event to be generated per + * socket; so_ifdenied_notifies is reset upon changes in + * the UUID policy. See comments in inp_update_policy. + */ + if (net_io_policy_log) { + uuid_string_t buf; + + uuid_unparse(ev_ifdenied.ev_data.euuid, buf); + log(LOG_DEBUG, "%s[%d]: so 0x%llx [%d,%d] epid %d " + "euuid %s%s has %d redundant events supressed\n", + __func__, so->last_pid, + (uint64_t)VM_KERNEL_ADDRPERM(so), SOCK_DOM(so), + SOCK_TYPE(so), ev_ifdenied.ev_data.epid, buf, + ((so->so_flags & SOF_DELEGATED) ? + " [delegated]" : ""), so->so_ifdenied_notifies); + } + } else { + if (net_io_policy_log) { + uuid_string_t buf; + + uuid_unparse(ev_ifdenied.ev_data.euuid, buf); + log(LOG_DEBUG, "%s[%d]: so 0x%llx [%d,%d] epid %d " + "euuid %s%s event posted\n", __func__, + so->last_pid, (uint64_t)VM_KERNEL_ADDRPERM(so), + SOCK_DOM(so), SOCK_TYPE(so), + ev_ifdenied.ev_data.epid, buf, + ((so->so_flags & SOF_DELEGATED) ? + " [delegated]" : "")); + } + netpolicy_post_msg(KEV_NETPOLICY_IFDENIED, &ev_ifdenied.ev_data, + sizeof (ev_ifdenied)); + } } -#endif __APPLE__ /* * Make a copy of a sockaddr in a malloced buffer of type M_SONAME. */ struct sockaddr * -dup_sockaddr(sa, canwait) - struct sockaddr *sa; - int canwait; +dup_sockaddr(struct sockaddr *sa, int canwait) { struct sockaddr *sa2; - MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME, - canwait ? M_WAITOK : M_NOWAIT); + MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME, + canwait ? M_WAITOK : M_NOWAIT); if (sa2) bcopy(sa, sa2, sa->sa_len); - return sa2; + return (sa2); } /* @@ -1548,19 +2712,49 @@ dup_sockaddr(sa, canwait) void sotoxsocket(struct socket *so, struct xsocket *xso) { - xso->xso_len = sizeof *xso; - xso->xso_so = so; + xso->xso_len = sizeof (*xso); + xso->xso_so = (_XSOCKET_PTR(struct socket *))VM_KERNEL_ADDRPERM(so); xso->so_type = so->so_type; - xso->so_options = so->so_options; + xso->so_options = (short)(so->so_options & 0xffff); xso->so_linger = so->so_linger; xso->so_state = so->so_state; - xso->so_pcb = so->so_pcb; + xso->so_pcb = (_XSOCKET_PTR(caddr_t))VM_KERNEL_ADDRPERM(so->so_pcb); if (so->so_proto) { - xso->xso_protocol = so->so_proto->pr_protocol; - xso->xso_family = so->so_proto->pr_domain->dom_family; + xso->xso_protocol = SOCK_PROTO(so); + xso->xso_family = SOCK_DOM(so); + } else { + xso->xso_protocol = xso->xso_family = 0; } - else + xso->so_qlen = so->so_qlen; + xso->so_incqlen = so->so_incqlen; + xso->so_qlimit = so->so_qlimit; + xso->so_timeo = so->so_timeo; + xso->so_error = so->so_error; + xso->so_pgid = so->so_pgid; + xso->so_oobmark = so->so_oobmark; + sbtoxsockbuf(&so->so_snd, &xso->so_snd); + sbtoxsockbuf(&so->so_rcv, &xso->so_rcv); + xso->so_uid = kauth_cred_getuid(so->so_cred); +} + + + +void +sotoxsocket64(struct socket *so, struct xsocket64 *xso) +{ + xso->xso_len = sizeof (*xso); + xso->xso_so = (u_int64_t)VM_KERNEL_ADDRPERM(so); + xso->so_type = so->so_type; + xso->so_options = (short)(so->so_options & 0xffff); + xso->so_linger = so->so_linger; + xso->so_state = so->so_state; + xso->so_pcb = (u_int64_t)VM_KERNEL_ADDRPERM(so->so_pcb); + if (so->so_proto) { + xso->xso_protocol = SOCK_PROTO(so); + xso->xso_family = SOCK_DOM(so); + } 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; @@ -1570,9 +2764,10 @@ sotoxsocket(struct socket *so, struct xsocket *xso) xso->so_oobmark = so->so_oobmark; sbtoxsockbuf(&so->so_snd, &xso->so_snd); sbtoxsockbuf(&so->so_rcv, &xso->so_rcv); - xso->so_uid = so->so_uid; + xso->so_uid = kauth_cred_getuid(so->so_cred); } + /* * This does the same for sockbufs. Note that the xsockbuf structure, * since it is always embedded in a socket, does not include a self @@ -1588,26 +2783,133 @@ 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 = (u_long)(sb->sb_timeo.tv_sec * hz) + sb->sb_timeo.tv_usec / tick; + xsb->sb_timeo = (short) + (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; } +/* + * Based on the policy set by an all knowing decison maker, throttle sockets + * that either have been marked as belonging to "background" process. + */ +int +soisthrottled(struct socket *so) +{ + /* + * On non-embedded, we rely on implicit throttling by the + * application, as we're missing the system wide "decision maker" + */ + return ( + (so->so_traffic_mgt_flags & TRAFFIC_MGT_SO_BACKGROUND)); +} + +int +soisprivilegedtraffic(struct socket *so) +{ + return ((so->so_flags & SOF_PRIVILEGED_TRAFFIC_CLASS) ? 1 : 0); +} + +int +soissrcbackground(struct socket *so) +{ + return ((so->so_traffic_mgt_flags & TRAFFIC_MGT_SO_BACKGROUND) || + IS_SO_TC_BACKGROUND(so->so_traffic_class)); +} + +int +soissrcrealtime(struct socket *so) +{ + return (so->so_traffic_class >= SO_TC_AV); +} + +void +sonullevent(struct socket *so, void *arg, uint32_t hint) +{ +#pragma unused(so, arg, hint) +} + /* * 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"); +SYSCTL_NODE(_kern, KERN_IPC, ipc, + CTLFLAG_RW|CTLFLAG_LOCKED|CTLFLAG_ANYBODY, 0, "IPC"); + +/* Check that the maximum socket buffer size is within a range */ + +static int +sysctl_sb_max SYSCTL_HANDLER_ARGS +{ +#pragma unused(oidp, arg1, arg2) + u_int32_t new_value; + int changed = 0; + int error = sysctl_io_number(req, sb_max, sizeof (u_int32_t), + &new_value, &changed); + if (!error && changed) { + if (new_value > LOW_SB_MAX && new_value <= high_sb_max) { + sb_max = new_value; + } else { + error = ERANGE; + } + } + return (error); +} + +static int +sysctl_io_policy_throttled SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2) + int i, err; + + i = net_io_policy_throttled; + + err = sysctl_handle_int(oidp, &i, 0, req); + if (err != 0 || req->newptr == USER_ADDR_NULL) + return (err); + + if (i != net_io_policy_throttled) + SOTHROTTLELOG(("throttle: network IO policy throttling is " + "now %s\n", i ? "ON" : "OFF")); + + net_io_policy_throttled = i; + + return (err); +} + +SYSCTL_PROC(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, + CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, + &sb_max, 0, &sysctl_sb_max, "IU", "Maximum socket buffer size"); + +SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, + CTLFLAG_RW | CTLFLAG_LOCKED, &sb_efficiency, 0, ""); + +SYSCTL_INT(_kern_ipc, KIPC_NMBCLUSTERS, nmbclusters, + CTLFLAG_RD | CTLFLAG_LOCKED, &nmbclusters, 0, ""); + +SYSCTL_INT(_kern_ipc, OID_AUTO, njcl, + CTLFLAG_RD | CTLFLAG_LOCKED, &njcl, 0, ""); + +SYSCTL_INT(_kern_ipc, OID_AUTO, njclbytes, + CTLFLAG_RD | CTLFLAG_LOCKED, &njclbytes, 0, ""); + +SYSCTL_INT(_kern_ipc, KIPC_SOQLIMITCOMPAT, soqlimitcompat, + CTLFLAG_RW | CTLFLAG_LOCKED, &soqlimitcompat, 1, + "Enable socket queue limit compatibility"); + +SYSCTL_INT(_kern_ipc, OID_AUTO, soqlencomp, CTLFLAG_RW | CTLFLAG_LOCKED, + &soqlencomp, 0, "Listen backlog represents only complete queue"); + +SYSCTL_NODE(_kern_ipc, OID_AUTO, io_policy, CTLFLAG_RW, 0, "network IO policy"); -/* 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_PROC(_kern_ipc_io_policy, OID_AUTO, throttled, + CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &net_io_policy_throttled, 0, + sysctl_io_policy_throttled, "I", ""); -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, ""); +SYSCTL_INT(_kern_ipc_io_policy, OID_AUTO, log, CTLFLAG_RW | CTLFLAG_LOCKED, + &net_io_policy_log, 0, ""); +#if CONFIG_PROC_UUID_POLICY +SYSCTL_INT(_kern_ipc_io_policy, OID_AUTO, uuid, CTLFLAG_RW | CTLFLAG_LOCKED, + &net_io_policy_uuid, 0, ""); +#endif /* CONFIG_PROC_UUID_POLICY */