/*
- * Copyright (c) 1998-2013 Apple Inc. All rights reserved.
+ * Copyright (c) 1998-2019 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
#include <sys/ev.h>
#include <kern/locks.h>
#include <net/route.h>
+#include <net/content_filter.h>
#include <netinet/in.h>
#include <netinet/in_pcb.h>
+#include <netinet/tcp_var.h>
#include <sys/kdebug.h>
#include <libkern/OSAtomic.h>
#include <mach/vm_param.h>
-/* TODO: this should be in a header file somewhere */
-extern void postevent(struct socket *, struct sockbuf *, int);
+#if MPTCP
+#include <netinet/mptcp_var.h>
+#endif
+
+#define DBG_FNC_SBDROP NETDBG_CODE(DBG_NETSOCK, 4)
+#define DBG_FNC_SBAPPEND NETDBG_CODE(DBG_NETSOCK, 5)
+
+SYSCTL_DECL(_kern_ipc);
-#define DBG_FNC_SBDROP NETDBG_CODE(DBG_NETSOCK, 4)
-#define DBG_FNC_SBAPPEND NETDBG_CODE(DBG_NETSOCK, 5)
+__private_extern__ u_int32_t net_io_policy_throttle_best_effort = 0;
+SYSCTL_INT(_kern_ipc, OID_AUTO, throttle_best_effort,
+ CTLFLAG_RW | CTLFLAG_LOCKED, &net_io_policy_throttle_best_effort, 0, "");
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 *);
* 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;
+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() */
-__private_extern__ int32_t total_sbmb_cnt = 0;
+static u_int32_t sb_efficiency = 8; /* parameter for sbreserve() */
+int32_t total_sbmb_cnt __attribute__((aligned(8))) = 0;
+int32_t total_sbmb_cnt_floor __attribute__((aligned(8))) = 0;
+int32_t total_sbmb_cnt_peak __attribute__((aligned(8))) = 0;
+int64_t sbmb_limreached __attribute__((aligned(8))) = 0;
-/* 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;
-
-u_int32_t net_io_policy_log = 0; /* log socket policy changes */
+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 */
+u_int32_t net_io_policy_uuid = 1; /* enable UUID socket policy */
#endif /* CONFIG_PROC_UUID_POLICY */
/*
void
soisconnecting(struct socket *so)
{
-
- so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
+ so->so_state &= ~(SS_ISCONNECTED | SS_ISDISCONNECTING);
so->so_state |= SS_ISCONNECTING;
sflt_notify(so, sock_evt_connecting, NULL);
void
soisconnected(struct socket *so)
{
- struct socket *head = so->so_head;
-
- so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
+ so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING);
so->so_state |= SS_ISCONNECTED;
+ soreserve_preconnect(so, 0);
+
sflt_notify(so, sock_evt_connected, NULL);
- if (head && (so->so_state & SS_INCOMP)) {
- so->so_state &= ~SS_INCOMP;
- so->so_state |= SS_COMP;
+ if (so->so_head != NULL && (so->so_state & SS_INCOMP)) {
+ struct socket *head = so->so_head;
+ int locked = 0;
+
+ /*
+ * Enforce lock order when the protocol has per socket locks
+ */
if (head->so_proto->pr_getlock != NULL) {
- socket_unlock(so, 0);
socket_lock(head, 1);
+ so_acquire_accept_list(head, so);
+ locked = 1;
}
- postevent(head, 0, EV_RCONN);
- TAILQ_REMOVE(&head->so_incomp, so, so_list);
- head->so_incqlen--;
- TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
- sorwakeup(head);
- wakeup_one((caddr_t)&head->so_timeo);
- if (head->so_proto->pr_getlock != NULL) {
+ if (so->so_head == head && (so->so_state & SS_INCOMP)) {
+ so->so_state &= ~SS_INCOMP;
+ so->so_state |= SS_COMP;
+ TAILQ_REMOVE(&head->so_incomp, so, so_list);
+ TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
+ head->so_incqlen--;
+
+ /*
+ * We have to release the accept list in
+ * case a socket callback calls sock_accept()
+ */
+ if (locked != 0) {
+ so_release_accept_list(head);
+ socket_unlock(so, 0);
+ }
+ postevent(head, 0, EV_RCONN);
+ sorwakeup(head);
+ wakeup_one((caddr_t)&head->so_timeo);
+
+ if (locked != 0) {
+ socket_unlock(head, 1);
+ socket_lock(so, 0);
+ }
+ } else if (locked != 0) {
+ so_release_accept_list(head);
socket_unlock(head, 1);
- socket_lock(so, 0);
}
} else {
postevent(so, 0, EV_WCONN);
}
}
+boolean_t
+socanwrite(struct socket *so)
+{
+ return (so->so_state & SS_ISCONNECTED) ||
+ !(so->so_proto->pr_flags & PR_CONNREQUIRED) ||
+ (so->so_flags1 & SOF1_PRECONNECT_DATA);
+}
+
void
soisdisconnecting(struct socket *so)
{
so->so_state &= ~SS_ISCONNECTING;
- so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
+ 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);
void
soisdisconnected(struct socket *so)
{
- so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
- so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
+ 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 */
}
/*
void
sodisconnectwakeup(struct socket *so)
{
- so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
- so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
+ 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 CONTENT_FILTER
+ /* Notify content filters as soon as we cannot send/receive data */
+ cfil_sock_notify_shutdown(so, SHUT_RDWR);
+#endif /* CONTENT_FILTER */
}
/*
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);
+ }
+ LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
if (!soqlencomp) {
/*
* as so_qlen so that we fail immediately below.
*/
so_qlen = head->so_qlen - head->so_incqlen;
- if (head->so_incqlen > somaxconn)
+ if (head->so_incqlen > somaxconn) {
so_qlen = somaxconn;
+ }
}
if (so_qlen >=
- (soqlimitcompat ? head->so_qlimit : (3 * head->so_qlimit / 2)))
- return ((struct socket *)0);
+ (soqlimitcompat ? head->so_qlimit : (3 * head->so_qlimit / 2))) {
+ return (struct socket *)0;
+ }
so = soalloc(1, SOCK_DOM(head), head->so_type);
- if (so == NULL)
- return ((struct socket *)0);
+ 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);
+ return (struct socket *)0;
}
so->so_type = head->so_type;
- so->so_options = head->so_options &~ SO_ACCEPTCONN;
+ so->so_options = head->so_options & ~SO_ACCEPTCONN;
so->so_linger = head->so_linger;
so->so_state = head->so_state | SS_NOFDREF;
so->so_proto = head->so_proto;
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));
+ 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));
+ 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_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;
- so->so_rcv.sb_flags |= SB_RECV; /* XXX */
+ so->so_rcv.sb_flags |= SB_RECV; /* XXX */
so->so_rcv.sb_so = so->so_snd.sb_so = so;
TAILQ_INIT(&so->so_evlist);
#endif
/* inherit traffic management properties of listener */
- so->so_traffic_mgt_flags =
- head->so_traffic_mgt_flags & (TRAFFIC_MGT_SO_BACKGROUND);
+ so->so_flags1 |=
+ head->so_flags1 & (SOF1_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)) {
sodealloc(so);
- return ((struct socket *)0);
+ 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.
*/
- if (head->so_proto->pr_unlock)
+ if (head->so_proto->pr_unlock) {
socket_unlock(head, 0);
+ }
if (((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL) != 0) ||
error) {
sodealloc(so);
- if (head->so_proto->pr_unlock)
+ if (head->so_proto->pr_unlock) {
socket_lock(head, 0);
- return ((struct socket *)0);
+ }
+ return (struct socket *)0;
}
if (head->so_proto->pr_unlock) {
socket_lock(head, 0);
if ((head->so_options & SO_ACCEPTCONN) == 0) {
so->so_state &= ~SS_NOFDREF;
soclose(so);
- return ((struct socket *)0);
+ return (struct socket *)0;
}
}
atomic_add_32(&so->so_proto->pr_domain->dom_refs, 1);
/* Insert in head appropriate lists */
+ so_acquire_accept_list(head, NULL);
+
so->so_head = head;
/*
}
head->so_qlen++;
+ so_release_accept_list(head);
+
/* Attach socket filters for this protocol */
sflt_initsock(so);
sorwakeup(head);
wakeup((caddr_t)&head->so_timeo);
}
- return (so);
+ return so;
}
{
int error = sflt_connectin(head, from);
if (error) {
- return (NULL);
+ return NULL;
}
- return (sonewconn_internal(head, connstatus));
+ return sonewconn_internal(head, connstatus);
}
/*
/* NOTREACHED */
}
- if (so->so_proto->pr_getlock != NULL)
- mutex_held = (*so->so_proto->pr_getlock)(so, 0);
- else
+ if ((so->so_state & SS_DRAINING) || (so->so_flags & SOF_DEFUNCT)) {
+ error = EBADF;
+ if (so->so_flags & SOF_DEFUNCT) {
+ SODEFUNCTLOG("%s[%d, %s]: defunct so 0x%llx [%d,%d] "
+ "(%d)\n", __func__, proc_selfpid(),
+ proc_best_name(current_proc()),
+ (uint64_t)VM_KERNEL_ADDRPERM(so),
+ SOCK_DOM(so), SOCK_TYPE(so), error);
+ }
+ return error;
+ }
+
+ if (so->so_proto->pr_getlock != NULL) {
+ mutex_held = (*so->so_proto->pr_getlock)(so, PR_F_WILLUNLOCK);
+ } else {
mutex_held = so->so_proto->pr_domain->dom_mtx;
+ }
- lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
+ LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
ts.tv_sec = sb->sb_timeo.tv_sec;
ts.tv_nsec = sb->sb_timeo.tv_usec * 1000;
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] "
+ SODEFUNCTLOG("%s[%d, %s]: defunct so 0x%llx [%d,%d] "
"(%d)\n", __func__, proc_selfpid(),
+ proc_best_name(current_proc()),
(uint64_t)VM_KERNEL_ADDRPERM(so),
- SOCK_DOM(so), SOCK_TYPE(so), error));
+ SOCK_DOM(so), SOCK_TYPE(so), error);
}
}
- return (error);
+ return error;
}
void
sbwakeup(struct sockbuf *sb)
{
- if (sb->sb_waiters > 0)
+ if (sb->sb_waiters > 0) {
wakeup((caddr_t)&sb->sb_cc);
+ }
}
/*
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, "
+ SODEFUNCTLOG("%s[%d, %s]: defunct so 0x%llx [%d,%d] si 0x%x, "
"fl 0x%x [%s]\n", __func__, proc_selfpid(),
+ proc_best_name(current_proc()),
(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_RECV) ? "rcv" : "snd");
}
sb->sb_flags &= ~SB_SEL;
selwakeup(&sb->sb_sel);
sbwakeup(sb);
if (so->so_state & SS_ASYNC) {
- if (so->so_pgid < 0)
+ if (so->so_pgid < 0) {
gsignal(-so->so_pgid, SIGIO);
- else if (so->so_pgid > 0)
+ } 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;
+ int lock = !(sb->sb_flags & SB_UPCALL_LOCK);
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);
+ if (lock) {
+ socket_unlock(so, 0);
+ }
(*sb_upcall)(so, sb_upcallarg, M_DONTWAIT);
- socket_lock(so, 0);
+ if (lock) {
+ 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)
+ 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 */
}
/*
int
soreserve(struct socket *so, u_int32_t sndcc, u_int32_t rcvcc)
{
-
- if (sbreserve(&so->so_snd, sndcc) == 0)
+ /*
+ * We do not want to fail the creation of a socket
+ * when kern.ipc.maxsockbuf is less than the
+ * default socket buffer socket size of the protocol
+ * so force the buffer sizes to be at most the
+ * limit enforced by sbreserve()
+ */
+ uint64_t maxcc = (uint64_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
+ if (sndcc > maxcc) {
+ sndcc = maxcc;
+ }
+ if (rcvcc > maxcc) {
+ rcvcc = maxcc;
+ }
+ if (sbreserve(&so->so_snd, sndcc) == 0) {
goto bad;
- else
+ } else {
so->so_snd.sb_idealsize = sndcc;
+ }
- if (sbreserve(&so->so_rcv, rcvcc) == 0)
+ if (sbreserve(&so->so_rcv, rcvcc) == 0) {
goto bad2;
- else
+ } else {
so->so_rcv.sb_idealsize = rcvcc;
+ }
- if (so->so_rcv.sb_lowat == 0)
+ if (so->so_rcv.sb_lowat == 0) {
so->so_rcv.sb_lowat = 1;
- if (so->so_snd.sb_lowat == 0)
+ }
+ if (so->so_snd.sb_lowat == 0) {
so->so_snd.sb_lowat = MCLBYTES;
- if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
+ }
+ if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat) {
so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
- return (0);
+ }
+ return 0;
bad2:
so->so_snd.sb_flags &= ~SB_SEL;
selthreadclear(&so->so_snd.sb_sel);
sbrelease(&so->so_snd);
bad:
- return (ENOBUFS);
+ return ENOBUFS;
+}
+
+void
+soreserve_preconnect(struct socket *so, unsigned int pre_cc)
+{
+ /* As of now, same bytes for both preconnect read and write */
+ so->so_snd.sb_preconn_hiwat = pre_cc;
+ so->so_rcv.sb_preconn_hiwat = pre_cc;
}
/*
int
sbreserve(struct sockbuf *sb, u_int32_t cc)
{
- if ((u_quad_t)cc > (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES))
- return (0);
+ if ((u_quad_t)cc > (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES)) {
+ return 0;
+ }
sb->sb_hiwat = cc;
sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
- if (sb->sb_lowat > sb->sb_hiwat)
+ if (sb->sb_lowat > sb->sb_hiwat) {
sb->sb_lowat = sb->sb_hiwat;
- return (1);
+ }
+ return 1;
}
/*
struct socket *so = sb->sb_so;
if (m == NULL || (sb->sb_flags & SB_DROP)) {
- if (m != NULL)
+ if (m != NULL) {
m_freem(m);
- return (0);
+ }
+ return 0;
}
SBLASTRECORDCHK(sb, "sbappend 1");
- if (sb->sb_lastrecord != NULL && (sb->sb_mbtail->m_flags & M_EOR))
- return (sbappendrecord(sb, m));
+ if (sb->sb_lastrecord != NULL && (sb->sb_mbtail->m_flags & M_EOR)) {
+ return sbappendrecord(sb, m);
+ }
- if (sb->sb_flags & SB_RECV) {
+ 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)
+ if (error != EJUSTRETURN) {
m_freem(m);
- return (0);
+ }
+ return 0;
}
+ } 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)
+ if (sb->sb_lastrecord == NULL) {
sb->sb_lastrecord = m;
+ }
sbcompress(sb, m, sb->sb_mbtail);
SBLASTRECORDCHK(sb, "sbappend 3");
- return (1);
+ return 1;
}
/*
struct socket *so = sb->sb_so;
if (m == NULL || (sb->sb_flags & SB_DROP)) {
- if (m != NULL)
+ if (m != NULL) {
m_freem(m);
- return (0);
+ }
+ return 0;
}
if (m->m_nextpkt != NULL || (sb->sb_mb != sb->sb_lastrecord)) {
SBLASTMBUFCHK(sb, __func__);
- if (sb->sb_flags & SB_RECV) {
+ 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)
+ if (error != EJUSTRETURN) {
m_freem(m);
- return (0);
+ }
+ return 0;
}
+ } else if (m) {
+ m->m_flags &= ~M_SKIPCFIL;
}
sbcompress(sb, m, sb->sb_mbtail);
sb->sb_lastrecord = sb->sb_mb;
SBLASTRECORDCHK(sb, "sbappendstream 2");
- return (1);
+ return 1;
}
#ifdef SOCKBUF_DEBUG
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);
+ LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
- if (sbchecking == 0)
+ if (sbchecking == 0) {
return;
+ }
for (m = sb->sb_mb; m; m = n) {
n = m->m_nextpkt;
len += m->m_len;
mbcnt += MSIZE;
/* XXX pretty sure this is bogus */
- if (m->m_flags & M_EXT)
+ if (m->m_flags & M_EXT) {
mbcnt += m->m_ext.ext_size;
+ }
}
}
if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
{
struct mbuf *m = sb->sb_mb;
- while (m && m->m_nextpkt)
+ while (m && m->m_nextpkt) {
m = m->m_nextpkt;
+ }
if (m != sb->sb_lastrecord) {
- printf("sblastrecordchk: mb %p lastrecord %p last %p\n",
- sb->sb_mb, sb->sb_lastrecord, m);
+ 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("\t%p\n", m);
+ 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);
}
}
struct mbuf *m = sb->sb_mb;
struct mbuf *n;
- while (m && m->m_nextpkt)
+ while (m && m->m_nextpkt) {
m = m->m_nextpkt;
+ }
- while (m && m->m_next)
+ while (m && m->m_next) {
m = m->m_next;
+ }
if (m != sb->sb_mbtail) {
- printf("sblastmbufchk: mb %p mbtail %p last %p\n",
- sb->sb_mb, sb->sb_mbtail, m);
+ 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("%p ", n);
+ 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);
int space = 0;
if (m0 == NULL || (sb->sb_flags & SB_DROP)) {
- if (m0 != NULL)
+ if (m0 != NULL) {
m_freem(m0);
- return (0);
+ }
+ return 0;
}
- for (m = m0; m != NULL; m = m->m_next)
+ 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);
+ return 0;
}
- if (sb->sb_flags & SB_RECV) {
+ 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 (error != 0) {
SBLASTRECORDCHK(sb, "sbappendrecord 1");
- if (error != EJUSTRETURN)
+ if (error != EJUSTRETURN) {
m_freem(m0);
- return (0);
+ }
+ return 0;
}
+ } else if (m0) {
+ m0->m_flags &= ~M_SKIPCFIL;
}
/*
}
sbcompress(sb, m, m0);
SBLASTRECORDCHK(sb, "sbappendrecord 3");
- return (1);
+ return 1;
}
/*
struct mbuf *m;
struct mbuf **mp;
- if (m0 == 0)
- return (0);
+ if (m0 == 0) {
+ return 0;
+ }
SBLASTRECORDCHK(sb, "sbinsertoob 1");
- if ((sb->sb_flags & SB_RECV) != 0) {
+ 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);
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:
switch (m->m_type) {
-
case MT_OOBDATA:
- continue; /* WANT next train */
+ continue; /* WANT next train */
case MT_CONTROL:
m = m->m_next;
- if (m)
- goto again; /* inspect THIS train further */
+ if (m) {
+ goto again; /* inspect THIS train further */
+ }
}
break;
}
}
sbcompress(sb, m, m0);
SBLASTRECORDCHK(sb, "sbinsertoob 3");
- return (1);
+ return 1;
}
/*
- * Append address and data, and optionally, control (ancillary) data
- * 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.
+ * Concatenate address (optional), control (optional) and data into one
+ * single mbuf chain. If sockbuf *sb is passed in, space check will be
+ * performed.
*
- * Returns: 0 No space/out of mbufs
- * 1 Success
+ * Returns: mbuf chain pointer if succeeded, NULL if failed
*/
-static int
-sbappendaddr_internal(struct sockbuf *sb, struct sockaddr *asa,
- struct mbuf *m0, struct mbuf *control)
+struct mbuf *
+sbconcat_mbufs(struct sockbuf *sb, struct sockaddr *asa, struct mbuf *m0, struct mbuf *control)
{
- struct mbuf *m, *n, *nlast;
- int space = asa->sa_len;
+ struct mbuf *m = NULL, *n = NULL;
+ int space = 0;
- if (m0 && (m0->m_flags & M_PKTHDR) == 0)
- panic("sbappendaddr");
+ if (m0 && (m0->m_flags & M_PKTHDR) == 0) {
+ panic("sbconcat_mbufs");
+ }
- if (m0)
+ if (m0) {
space += m0->m_pkthdr.len;
+ }
for (n = control; n; n = n->m_next) {
space += n->m_len;
- if (n->m_next == 0) /* keep pointer to last control buf */
+ if (n->m_next == 0) { /* keep pointer to last control buf */
break;
+ }
+ }
+
+ if (asa != NULL) {
+ if (asa->sa_len > MLEN) {
+ return NULL;
+ }
+ space += asa->sa_len;
}
- if (space > sbspace(sb))
- return (0);
- if (asa->sa_len > MLEN)
- return (0);
- MGET(m, M_DONTWAIT, MT_SONAME);
- if (m == 0)
- return (0);
- m->m_len = asa->sa_len;
- bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
- if (n)
- n->m_next = m0; /* concatenate data to control */
- else
+
+ if (sb != NULL && space > sbspace(sb)) {
+ return NULL;
+ }
+
+ if (n) {
+ n->m_next = m0; /* concatenate data to control */
+ } else {
control = m0;
- m->m_next = control;
+ }
- SBLASTRECORDCHK(sb, "sbappendadddr 1");
+ if (asa != NULL) {
+ MGET(m, M_DONTWAIT, MT_SONAME);
+ if (m == 0) {
+ if (n) {
+ /* unchain control and data if necessary */
+ n->m_next = NULL;
+ }
+ return NULL;
+ }
+ m->m_len = asa->sa_len;
+ bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
+
+ m->m_next = control;
+ } else {
+ m = control;
+ }
- for (n = m; n->m_next != NULL; n = n->m_next)
+ return m;
+}
+
+/*
+ * Queue mbuf chain to the receive queue of a socket.
+ * Parameter space is the total len of the mbuf chain.
+ * If passed in, sockbuf space will be checked.
+ *
+ * Returns: 0 Invalid mbuf chain
+ * 1 Success
+ */
+int
+sbappendchain(struct sockbuf *sb, struct mbuf *m, int space)
+{
+ struct mbuf *n, *nlast;
+
+ if (m == NULL) {
+ return 0;
+ }
+
+ if (space != 0 && space > sbspace(sb)) {
+ return 0;
+ }
+
+ for (n = m; n->m_next != NULL; n = n->m_next) {
sballoc(sb, n);
+ }
sballoc(sb, n);
nlast = n;
SBLASTRECORDCHK(sb, "sbappendadddr 2");
postevent(0, sb, EV_RWBYTES);
- return (1);
+ return 1;
}
/*
{
int result = 0;
boolean_t sb_unix = (sb->sb_flags & SB_UNIX);
+ struct mbuf *mbuf_chain = NULL;
- if (error_out)
+ if (error_out) {
*error_out = 0;
+ }
- if (m0 && (m0->m_flags & M_PKTHDR) == 0)
+ if (m0 && (m0->m_flags & M_PKTHDR) == 0) {
panic("sbappendaddrorfree");
+ }
if (sb->sb_flags & SB_DROP) {
- if (m0 != NULL)
+ if (m0 != NULL) {
m_freem(m0);
- if (control != NULL && !sb_unix)
+ }
+ if (control != NULL && !sb_unix) {
m_freem(control);
- if (error_out != NULL)
+ }
+ if (error_out != NULL) {
*error_out = EINVAL;
- return (0);
+ }
+ 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);
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)
+ if (m0) {
m_freem(m0);
- if (control != NULL && !sb_unix)
+ }
+ if (control != NULL && !sb_unix) {
m_freem(control);
- if (error_out)
+ }
+ 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);
+ mbuf_chain = sbconcat_mbufs(sb, asa, m0, control);
+ SBLASTRECORDCHK(sb, "sbappendadddr 1");
+ result = sbappendchain(sb, mbuf_chain, 0);
if (result == 0) {
- if (m0)
+ if (m0) {
m_freem(m0);
- if (control != NULL && !sb_unix)
+ }
+ if (control != NULL && !sb_unix) {
m_freem(control);
- if (error_out)
+ }
+ if (error_out) {
*error_out = ENOBUFS;
+ }
+ }
+
+ return result;
+}
+
+inline boolean_t
+is_cmsg_valid(struct mbuf *control, struct cmsghdr *cmsg)
+{
+ if (cmsg == NULL) {
+ return FALSE;
+ }
+
+ if (cmsg->cmsg_len < sizeof(struct cmsghdr)) {
+ return FALSE;
}
- return (result);
+ if ((uint8_t *)control->m_data >= (uint8_t *)cmsg + cmsg->cmsg_len) {
+ return FALSE;
+ }
+
+ if ((uint8_t *)control->m_data + control->m_len <
+ (uint8_t *)cmsg + cmsg->cmsg_len) {
+ return FALSE;
+ }
+
+ return TRUE;
}
static int
struct mbuf *m, *mlast, *n;
int space = 0;
- if (control == 0)
+ if (control == 0) {
panic("sbappendcontrol");
+ }
- for (m = control; ; m = m->m_next) {
+ for (m = control;; m = m->m_next) {
space += m->m_len;
- if (m->m_next == 0)
+ if (m->m_next == 0) {
break;
+ }
}
- n = m; /* save pointer to last control buffer */
- for (m = m0; m; m = m->m_next)
+ n = m; /* save pointer to last control buffer */
+ for (m = m0; m; m = m->m_next) {
space += m->m_len;
- if (space > sbspace(sb) && !(sb->sb_flags & SB_UNIX))
- return (0);
- n->m_next = m0; /* concatenate data to control */
+ }
+ if (space > sbspace(sb) && !(sb->sb_flags & SB_UNIX)) {
+ return 0;
+ }
+ n->m_next = m0; /* concatenate data to control */
SBLASTRECORDCHK(sb, "sbappendcontrol 1");
- for (m = control; m->m_next != NULL; m = m->m_next)
+ for (m = control; m->m_next != NULL; m = m->m_next) {
sballoc(sb, m);
+ }
sballoc(sb, m);
mlast = m;
SBLASTRECORDCHK(sb, "sbappendcontrol 2");
postevent(0, sb, EV_RWBYTES);
- return (1);
+ return 1;
}
int
-sbappendcontrol(struct sockbuf *sb, struct mbuf *m0, struct mbuf *control,
+sbappendcontrol(struct sockbuf *sb, struct mbuf *m0, struct mbuf *control,
int *error_out)
{
int result = 0;
boolean_t sb_unix = (sb->sb_flags & SB_UNIX);
- if (error_out)
+ if (error_out) {
*error_out = 0;
+ }
if (sb->sb_flags & SB_DROP) {
- if (m0 != NULL)
+ if (m0 != NULL) {
m_freem(m0);
- if (control != NULL && !sb_unix)
+ }
+ if (control != NULL && !sb_unix) {
m_freem(control);
- if (error_out != NULL)
+ }
+ if (error_out != NULL) {
*error_out = EINVAL;
- return (0);
+ }
+ return 0;
}
- if (sb->sb_flags & SB_RECV) {
+ 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);
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)
+ if (m0) {
m_freem(m0);
- if (control != NULL && !sb_unix)
+ }
+ if (control != NULL && !sb_unix) {
m_freem(control);
- if (error_out)
+ }
+ 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)
+ if (m0) {
m_freem(m0);
- if (control != NULL && !sb_unix)
+ }
+ if (control != NULL && !sb_unix) {
m_freem(control);
- if (error_out)
+ }
+ if (error_out) {
*error_out = ENOBUFS;
+ }
}
- return (result);
+ return result;
}
/*
int ret = 0;
struct socket *so = sb->sb_so;
+ if (m == NULL) {
+ return 0;
+ }
+
VERIFY((m->m_flags & M_PKTHDR) && m_pktlen(m) > 0);
VERIFY(so->so_msg_state != NULL);
VERIFY(sb->sb_flags & SB_RECV);
m->m_pkthdr.msg_seq = seqnum;
/* find last mbuf and set M_EOR */
- for (m_eor = m; ; m_eor = m_eor->m_next) {
+ 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,
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 {
+ } else {
m_eor->m_flags &= ~M_UNORDERED_DATA;
}
-
- if (m_eor->m_next == NULL)
+ if (m_eor->m_next == NULL) {
break;
+ }
}
/* set EOR flag at end of byte blob */
__func__);
}
- ret = sbappendrecord(sb, m);
- return (ret);
+ 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;
}
/*
*/
int
sbappendstream_rcvdemux(struct socket *so, struct mbuf *m, uint32_t seqnum,
- int unordered)
+ int unordered)
{
int ret = 0;
- if ((m != NULL) && (m_pktlen(m) <= 0)) {
+ if ((m != NULL) &&
+ m_pktlen(m) <= 0 &&
+ !((so->so_flags & SOF_MP_SUBFLOW) &&
+ (m->m_flags & M_PKTHDR) &&
+ (m->m_pkthdr.pkt_flags & PKTF_MPTCP_DFIN))) {
m_freem(m);
- return (ret);
+ 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) {
+ else if (so->so_flags & SOF_MP_SUBFLOW) {
ret = sbappendmptcpstream_rcv(&so->so_rcv, m);
}
#endif /* MPTCP */
else {
ret = sbappendstream(&so->so_rcv, m);
}
- return (ret);
+ return ret;
}
#if MPTCP
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));
+ 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)
+ if (m && (m->m_flags & M_PKTHDR) &&
+ m_pktlen(m) == 0 &&
+ (m->m_pkthdr.pkt_flags & PKTF_MPTCP_DFIN)) {
+ mptcp_input(tptomptp(sototcpcb(so))->mpt_mpte, m);
+ return 1;
+ } else if (m != NULL) {
m_freem(m);
- return (0);
+ }
+ return 0;
}
/* the socket is not closed, so SOF_MP_SUBFLOW must be set */
VERIFY(so->so_flags & SOF_MP_SUBFLOW);
SBLASTMBUFCHK(sb, __func__);
- mptcp_adj_rmap(so, m);
-
/* 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);
+ return 1;
}
#endif /* MPTCP */
VERIFY(so->so_msg_state != NULL);
- if (m->m_nextpkt != NULL || (sb->sb_mb != sb->sb_lastrecord))
+ 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)
+ if (m != NULL) {
m_freem(m);
- return (0);
+ }
+ return 0;
}
priq = &so->so_msg_state->msg_priq[m->m_pkthdr.msg_pri];
SBLASTRECORDCHK(sb, "sbappendstream 2");
postevent(0, sb, EV_RWBYTES);
- return (1);
+ return 1;
}
/*
* the mbuf that is being dequeued, update
* it to point to the new head.
*/
- if (priq->msgq_lastmsg == m)
+ if (priq->msgq_lastmsg == m) {
priq->msgq_lastmsg = priq->msgq_head;
+ }
m->m_nextpkt = NULL;
mend->m_next = NULL;
topull =
(off + len) - so->so_msg_state->msg_serial_bytes;
- if (priq->msgq_flags & MSGQ_MSG_NOTDONE)
+ if (priq->msgq_flags & MSGQ_MSG_NOTDONE) {
break;
+ }
} else {
--i;
}
if (m == NULL) {
/* There is nothing to compress; just update the tail */
- for (; n->m_next != NULL; n = n->m_next)
+ for (; n->m_next != NULL; n = n->m_next) {
;
+ }
sb->sb_mbtail = n;
goto done;
}
eor |= m->m_flags & M_EOR;
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)
+ if (sb->sb_lastrecord == m) {
sb->sb_lastrecord = m->m_next;
+ }
m = m_free(m);
continue;
}
/* XXX: Probably don't need */
sb->sb_ctl += m->m_len;
}
+
+ /* update send byte count */
+ if (sb->sb_flags & SB_SNDBYTE_CNT) {
+ inp_incr_sndbytes_total(sb->sb_so,
+ m->m_len);
+ inp_incr_sndbytes_unsent(sb->sb_so,
+ m->m_len);
+ }
m = m_free(m);
continue;
}
- if (n != NULL)
+ if (n != NULL) {
n->m_next = m;
- else
+ } else {
sb->sb_mb = m;
+ }
sb->sb_mbtail = m;
sballoc(sb, m);
n = m;
n->m_next = NULL;
}
if (eor != 0) {
- if (n != NULL)
+ if (n != NULL) {
n->m_flags |= eor;
- else
+ } else {
printf("semi-panic: sbcompress\n");
+ }
}
done:
SBLASTMBUFCHK(sb, __func__);
{
struct mbuf *m;
m = priq->msgq_head;
- if (m != NULL)
+ if (m != NULL) {
m_freem_list(m);
+ }
priq->msgq_head = priq->msgq_tail = priq->msgq_lastmsg = NULL;
priq->msgq_bytes = priq->msgq_flags = 0;
}
{
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() */
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
* 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))
+ if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len)) {
break;
+ }
sbdrop(sb, (int)sb->sb_cc);
}
sb_empty_assert(sb, __func__);
postevent(0, sb, EV_RWBYTES);
- sbunlock(sb, TRUE); /* keep socket locked */
+ sbunlock(sb, TRUE); /* keep socket locked */
}
/*
next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
#if MPTCP
- if ((m != NULL) && (len > 0) &&
- (!(sb->sb_flags & SB_RECV)) &&
+ 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))))) {
- mptcp_preproc_sbdrop(m, (unsigned int)len);
+ (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(sb->sb_so, m, (unsigned int)len);
+ }
+ if (m != NULL && len > 0 && !(sb->sb_flags & SB_RECV) &&
+ (sb->sb_so->so_flags & SOF_MP_SUBFLOW) &&
+ (sb->sb_so->so_flags1 & SOF1_POST_FALLBACK_SYNC)) {
+ mptcp_fallback_sbdrop(sb->sb_so, m, len);
}
#endif /* MPTCP */
KERNEL_DEBUG((DBG_FNC_SBDROP | DBG_FUNC_START), sb, len, 0, 0, 0);
ml = (struct mbuf *)0;
while (len > 0) {
- if (m == 0) {
- if (next == 0) {
+ if (m == NULL) {
+ if (next == NULL) {
/*
* temporarily replacing this panic with printf
* because it occurs occasionally when closing
if (!(sb->sb_flags & SB_RECV) &&
(sb->sb_so->so_flags & SOF_ENABLE_MSGS)) {
sb->sb_so->so_msg_state->
- msg_serial_bytes = 0;
+ msg_serial_bytes = 0;
}
break;
}
m->m_len -= len;
m->m_data += len;
sb->sb_cc -= len;
+ /* update the send byte count */
+ if (sb->sb_flags & SB_SNDBYTE_CNT) {
+ inp_decr_sndbytes_total(sb->sb_so, len);
+ }
if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
- m->m_type != MT_OOBDATA)
+ m->m_type != MT_OOBDATA) {
sb->sb_ctl -= len;
+ }
break;
}
len -= m->m_len;
sb->sb_lastrecord = m;
}
+#if CONTENT_FILTER
+ cfil_sock_buf_update(sb);
+#endif /* CONTENT_FILTER */
+
postevent(0, sb, EV_RWBYTES);
KERNEL_DEBUG((DBG_FNC_SBDROP | DBG_FUNC_END), sb, 0, 0, 0, 0);
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);
+ 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 *);
- VERIFY(IS_P2ALIGNED(cp, sizeof (u_int32_t)));
+ VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
/* XXX check size? */
(void) memcpy(CMSG_DATA(cp), p, size);
m->m_len = CMSG_SPACE(size);
cp->cmsg_len = CMSG_LEN(size);
cp->cmsg_level = level;
cp->cmsg_type = type;
- return (m);
+ return m;
}
struct mbuf **
if (*mp == NULL) {
*mp = sbcreatecontrol(p, size, type, level);
- return (mp);
+ return mp;
}
if (CMSG_SPACE((u_int)size) + (*mp)->m_len > MLEN) {
mp = &(*mp)->m_next;
*mp = sbcreatecontrol(p, size, type, level);
- return (mp);
+ 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)));
+ VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
m->m_len += CMSG_SPACE(size);
/* XXX check size? */
cp->cmsg_level = level;
cp->cmsg_type = type;
- return (mp);
+ return mp;
}
pru_abort_notsupp(struct socket *so)
{
#pragma unused(so)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
{
#pragma unused(so, nam)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_attach_notsupp(struct socket *so, int proto, struct proc *p)
{
#pragma unused(so, proto, p)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_bind_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p)
{
#pragma unused(so, nam, p)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p)
{
#pragma unused(so, nam, p)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_connect2_notsupp(struct socket *so1, struct socket *so2)
{
#pragma unused(so1, so2)
- return (EOPNOTSUPP);
+ 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)
+pru_connectx_notsupp(struct socket *so, struct sockaddr *src,
+ struct sockaddr *dst, struct proc *p, uint32_t ifscope,
+ sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
+ uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
{
-#pragma unused(so, src_sl, dst_sl, p, ifscope, aid, pcid, flags, arg, arglen)
- return (EOPNOTSUPP);
+#pragma unused(so, src, dst, p, ifscope, aid, pcid, flags, arg, arglen, uio, bytes_written)
+ return EOPNOTSUPP;
}
int
struct ifnet *ifp, struct proc *p)
{
#pragma unused(so, cmd, data, ifp, p)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_detach_notsupp(struct socket *so)
{
#pragma unused(so)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_disconnect_notsupp(struct socket *so)
{
#pragma unused(so)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
-pru_disconnectx_notsupp(struct socket *so, associd_t aid, connid_t cid)
+pru_disconnectx_notsupp(struct socket *so, sae_associd_t aid, sae_connid_t cid)
{
#pragma unused(so, aid, cid)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_listen_notsupp(struct socket *so, struct proc *p)
{
#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);
+ return EOPNOTSUPP;
}
int
pru_peeraddr_notsupp(struct socket *so, struct sockaddr **nam)
{
#pragma unused(so, nam)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_rcvd_notsupp(struct socket *so, int flags)
{
#pragma unused(so, flags)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
{
#pragma unused(so, m, flags)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
struct sockaddr *addr, struct mbuf *control, struct proc *p)
{
#pragma unused(so, flags, m, addr, control, p)
- return (EOPNOTSUPP);
+ 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;
}
/*
sb->st_blksize = so->so_snd.sb_hiwat;
}
- return (0);
+ return 0;
}
struct mbuf *top, struct mbuf *control, int flags)
{
#pragma unused(so, addr, uio, top, control, flags)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
+}
+
+int
+pru_sosend_list_notsupp(struct socket *so, struct uio **uio,
+ u_int uiocnt, int flags)
+{
+#pragma unused(so, uio, uiocnt, flags)
+ return EOPNOTSUPP;
}
int
struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
{
#pragma unused(so, paddr, uio, mp0, controlp, flagsp)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
+}
+
+int
+pru_soreceive_list_notsupp(struct socket *so,
+ struct recv_msg_elem *recv_msg_array, u_int uiocnt, int *flagsp)
+{
+#pragma unused(so, recv_msg_array, uiocnt, flagsp)
+ return EOPNOTSUPP;
}
int
pru_shutdown_notsupp(struct socket *so)
{
#pragma unused(so)
- return (EOPNOTSUPP);
+ return EOPNOTSUPP;
}
int
pru_sockaddr_notsupp(struct socket *so, struct sockaddr **nam)
{
#pragma unused(so, nam)
- return (EOPNOTSUPP);
+ 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);
+ return EOPNOTSUPP;
}
int
/*
* Allow all options for set/get by default.
*/
- return (0);
+ return 0;
+}
+
+static int
+pru_preconnect_null(struct socket *so)
+{
+#pragma unused(so)
+ return 0;
}
void
pru_sanitize(struct pr_usrreqs *pru)
{
-#define DEFAULT(foo, bar) if ((foo) == NULL) (foo) = (bar)
+#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_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);
+ DEFAULT(pru->pru_preconnect, pru_preconnect_null);
#undef DEFAULT
}
int
sb_notify(struct sockbuf *sb)
{
- return (sb->sb_waiters > 0 ||
- (sb->sb_flags & (SB_SEL|SB_ASYNC|SB_UPCALL|SB_KNOTE)));
+ return sb->sb_waiters > 0 ||
+ (sb->sb_flags & (SB_SEL | SB_ASYNC | SB_UPCALL | SB_KNOTE));
}
/*
int
sbspace(struct sockbuf *sb)
{
+ int pending = 0;
int space = imin((int)(sb->sb_hiwat - sb->sb_cc),
(int)(sb->sb_mbmax - sb->sb_mbcnt));
- if (space < 0)
+
+ if (sb->sb_preconn_hiwat != 0) {
+ space = imin((int)(sb->sb_preconn_hiwat - sb->sb_cc), space);
+ }
+
+ 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);
+ return space;
}
/*
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 &&
- control != NULL);
- error = tcp_get_msg_priority(control, &msgpri);
- if (error)
- return (0);
+ u_int32_t msgpri = 0;
+ 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)
+ if (space < 0) {
space = 0;
- return (space);
+ }
+ 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? */
int
sowriteable(struct socket *so)
{
- return ((!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) == 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) || !socanwrite(so)) {
+ return 0;
+ }
+ if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
+ return 1;
+ }
+
+ if (sbspace(&(so)->so_snd) >= (so)->so_snd.sb_lowat) {
+ 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 */
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)
+ m->m_type != MT_OOBDATA) {
sb->sb_ctl += m->m_len;
+ }
sb->sb_mbcnt += MSIZE;
if (m->m_flags & M_EXT) {
}
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;
+ }
+
+ /*
+ * If data is being added to the send socket buffer,
+ * update the send byte count
+ */
+ if (sb->sb_flags & SB_SNDBYTE_CNT) {
+ inp_incr_sndbytes_total(sb->sb_so, m->m_len);
+ inp_incr_sndbytes_unsent(sb->sb_so, m->m_len);
+ }
}
/* adjust counters in sb reflecting freeing of m */
sb->sb_cc -= m->m_len;
if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
- m->m_type != MT_OOBDATA)
+ 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;
}
OSAddAtomic(cnt, &total_sbmb_cnt);
VERIFY(total_sbmb_cnt >= 0);
+ if (total_sbmb_cnt < total_sbmb_cnt_floor) {
+ total_sbmb_cnt_floor = total_sbmb_cnt;
+ }
+
+ /*
+ * If data is being removed from the send socket buffer,
+ * update the send byte count
+ */
+ if (sb->sb_flags & SB_SNDBYTE_CNT) {
+ inp_decr_sndbytes_total(sb->sb_so, m->m_len);
+ }
}
/*
struct socket *so = sb->sb_so;
void * wchan;
int error = 0;
+ thread_t tp = current_thread();
VERIFY((flags & SBL_VALID) == flags);
/* NOTREACHED */
}
- if ((sb->sb_flags & SB_LOCK) && !(flags & SBL_WAIT))
- return (EWOULDBLOCK);
+ /*
+ * 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
wchan = (sb->sb_flags & SB_RECV) ?
&so->so_rcv.sb_flags : &so->so_snd.sb_flags;
- while (sb->sb_flags & SB_LOCK) {
+ /*
+ * 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;
/*
* 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
+ if (so->so_proto->pr_getlock != NULL) {
+ mutex_held = (*so->so_proto->pr_getlock)(so, PR_F_WILLUNLOCK);
+ } else {
mutex_held = so->so_proto->pr_domain->dom_mtx;
+ }
- lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
+ LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
sb->sb_wantlock++;
VERIFY(sb->sb_wantlock != 0);
if (error == 0 && (so->so_flags & SOF_DEFUNCT) &&
!(flags & SBL_IGNDEFUNCT)) {
error = EBADF;
- SODEFUNCTLOG(("%s[%d]: defunct so 0x%llx [%d,%d] "
+ SODEFUNCTLOG("%s[%d, %s]: defunct so 0x%llx [%d,%d] "
"(%d)\n", __func__, proc_selfpid(),
+ proc_best_name(current_proc()),
(uint64_t)VM_KERNEL_ADDRPERM(so),
- SOCK_DOM(so), SOCK_TYPE(so), error));
+ SOCK_DOM(so), SOCK_TYPE(so), error);
}
- if (error != 0)
- return (error);
+ if (error != 0) {
+ return error;
+ }
}
sb->sb_flags |= SB_LOCK;
- return (0);
+ return 0;
}
/*
{
void *lr_saved = __builtin_return_address(0);
struct socket *so = sb->sb_so;
+ thread_t tp = current_thread();
/* so_usecount may be 0 if we get here from sofreelastref() */
if (so == NULL) {
/* NOTREACHED */
}
- VERIFY(sb->sb_flags & SB_LOCK);
- sb->sb_flags &= ~SB_LOCK;
-
- if (sb->sb_wantlock > 0) {
+ /*
+ * The content filter thread must hold the sockbuf lock
+ */
+ if ((so->so_flags & SOF_CONTENT_FILTER) && sb->sb_cfil_thread == tp) {
/*
- * 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.
+ * Don't panic if we are defunct because SB_LOCK has
+ * been cleared by sodefunct()
*/
- wakeup((sb->sb_flags & SB_RECV) ? &so->so_rcv.sb_flags :
- &so->so_snd.sb_flags);
+ 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 (!keeplocked) { /* unlock on exit */
+ if (!keeplocked) { /* unlock on exit */
lck_mtx_t *mutex_held;
- if (so->so_proto->pr_getlock != NULL)
- mutex_held = (*so->so_proto->pr_getlock)(so, 0);
- else
+ if (so->so_proto->pr_getlock != NULL) {
+ mutex_held = (*so->so_proto->pr_getlock)(so, PR_F_WILLUNLOCK);
+ } else {
mutex_held = so->so_proto->pr_domain->dom_mtx;
+ }
- lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
+ LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
- VERIFY(so->so_usecount != 0);
+ VERIFY(so->so_usecount > 0);
so->so_usecount--;
so->unlock_lr[so->next_unlock_lr] = lr_saved;
so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
void
sorwakeup(struct socket *so)
{
- if (sb_notify(&so->so_rcv))
+ if (sb_notify(&so->so_rcv)) {
sowakeup(so, &so->so_rcv);
+ }
}
void
sowwakeup(struct socket *so)
{
- if (sb_notify(&so->so_snd))
+ if (sb_notify(&so->so_snd)) {
sowakeup(so, &so->so_snd);
+ }
}
void
soevent(struct socket *so, long hint)
{
- if (so->so_flags & SOF_KNOTE)
+ if (so->so_flags & SOF_KNOTE) {
KNOTE(&so->so_klist, hint);
+ }
soevupcall(so, hint);
- /* Don't post an event if this a subflow socket */
- if ((hint & SO_FILT_HINT_IFDENIED) && !(so->so_flags & SOF_MP_SUBFLOW))
+ /*
+ * 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
soevupcall(struct socket *so, u_int32_t hint)
{
- void (*so_event)(struct socket *, void *, uint32_t);
-
- if ((so_event = so->so_event) != NULL) {
+ 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);
}
}
}
{
struct kev_netpolicy_ifdenied ev_ifdenied;
- bzero(&ev_ifdenied, sizeof (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
" [delegated]" : ""));
}
netpolicy_post_msg(KEV_NETPOLICY_IFDENIED, &ev_ifdenied.ev_data,
- sizeof (ev_ifdenied));
+ sizeof(ev_ifdenied));
}
}
MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME,
canwait ? M_WAITOK : M_NOWAIT);
- if (sa2)
+ if (sa2) {
bcopy(sa, sa2, sa->sa_len);
- return (sa2);
+ }
+ return sa2;
}
/*
void
sotoxsocket(struct socket *so, struct xsocket *xso)
{
- xso->xso_len = sizeof (*xso);
+ 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 = (short)(so->so_options & 0xffff);
}
+#if !CONFIG_EMBEDDED
void
sotoxsocket64(struct socket *so, struct xsocket64 *xso)
{
- xso->xso_len = sizeof (*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_uid = kauth_cred_getuid(so->so_cred);
}
+#endif /* !CONFIG_EMBEDDED */
/*
* This does the same for sockbufs. Note that the xsockbuf structure,
xsb->sb_flags = sb->sb_flags;
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)
+ 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
+inline 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));
+ return so->so_flags1 & SOF1_TRAFFIC_MGT_SO_BACKGROUND;
}
-int
+inline int
soisprivilegedtraffic(struct socket *so)
{
- return ((so->so_flags & SOF_PRIVILEGED_TRAFFIC_CLASS) ? 1 : 0);
+ return (so->so_flags & SOF_PRIVILEGED_TRAFFIC_CLASS) ? 1 : 0;
}
-int
+inline int
soissrcbackground(struct socket *so)
{
- return ((so->so_traffic_mgt_flags & TRAFFIC_MGT_SO_BACKGROUND) ||
- IS_SO_TC_BACKGROUND(so->so_traffic_class));
+ return (so->so_flags1 & SOF1_TRAFFIC_MGT_SO_BACKGROUND) ||
+ IS_SO_TC_BACKGROUND(so->so_traffic_class);
+}
+
+inline int
+soissrcrealtime(struct socket *so)
+{
+ return so->so_traffic_class >= SO_TC_AV &&
+ so->so_traffic_class <= SO_TC_VO;
+}
+
+inline int
+soissrcbesteffort(struct socket *so)
+{
+ return so->so_traffic_class == SO_TC_BE ||
+ so->so_traffic_class == SO_TC_RD ||
+ so->so_traffic_class == SO_TC_OAM;
+}
+
+void
+soclearfastopen(struct socket *so)
+{
+ if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
+ so->so_flags1 &= ~SOF1_PRECONNECT_DATA;
+ }
+
+ if (so->so_flags1 & SOF1_DATA_IDEMPOTENT) {
+ so->so_flags1 &= ~SOF1_DATA_IDEMPOTENT;
+ }
+}
+
+void
+sonullevent(struct socket *so, void *arg, uint32_t hint)
+{
+#pragma unused(so, arg, hint)
}
/*
* branch of the MIB.
*/
SYSCTL_NODE(_kern, KERN_IPC, ipc,
- CTLFLAG_RW|CTLFLAG_LOCKED|CTLFLAG_ANYBODY, 0, "IPC");
+ CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY, 0, "IPC");
/* Check that the maximum socket buffer size is within a range */
#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),
+ 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) {
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);
+ return error;
}
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, OID_AUTO, maxsockets,
- CTLFLAG_RD | CTLFLAG_LOCKED, &maxsockets, 0,
- "Maximum number of sockets avaliable");
+ 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, "");
+ CTLFLAG_RW | CTLFLAG_LOCKED, &sb_efficiency, 0, "");
SYSCTL_INT(_kern_ipc, KIPC_NMBCLUSTERS, nmbclusters,
- CTLFLAG_RD | CTLFLAG_LOCKED, &nmbclusters, 0, "");
+ CTLFLAG_RD | CTLFLAG_LOCKED, &nmbclusters, 0, "");
SYSCTL_INT(_kern_ipc, OID_AUTO, njcl,
- CTLFLAG_RD | CTLFLAG_LOCKED, &njcl, 0, "");
+ CTLFLAG_RD | CTLFLAG_LOCKED, &njcl, 0, "");
SYSCTL_INT(_kern_ipc, OID_AUTO, njclbytes,
- CTLFLAG_RD | CTLFLAG_LOCKED, &njclbytes, 0, "");
+ CTLFLAG_RD | CTLFLAG_LOCKED, &njclbytes, 0, "");
SYSCTL_INT(_kern_ipc, KIPC_SOQLIMITCOMPAT, soqlimitcompat,
- CTLFLAG_RW | CTLFLAG_LOCKED, &soqlimitcompat, 1,
- "Enable socket queue limit compatibility");
+ CTLFLAG_RW | CTLFLAG_LOCKED, &soqlimitcompat, 1,
+ "Enable socket queue limit compatibility");
+
+/*
+ * Hack alert -- rdar://33572856
+ * A loopback test we cannot change was failing because it sets
+ * SO_SENDTIMEO to 5 seconds and that's also the value
+ * of the minimum persist timer. Because of the persist timer,
+ * the connection was not idle for 5 seconds and SO_SNDTIMEO
+ * was not triggering at 5 seconds causing the test failure.
+ * As a workaround we check the sysctl soqlencomp the test is already
+ * setting to set disable auto tuning of the receive buffer.
+ */
-SYSCTL_INT(_kern_ipc, OID_AUTO, soqlencomp, CTLFLAG_RW | CTLFLAG_LOCKED,
- &soqlencomp, 0, "Listen backlog represents only complete queue");
+extern u_int32_t tcp_do_autorcvbuf;
-SYSCTL_NODE(_kern_ipc, OID_AUTO, io_policy, CTLFLAG_RW, 0, "network IO policy");
+static int
+sysctl_soqlencomp SYSCTL_HANDLER_ARGS
+{
+#pragma unused(oidp, arg1, arg2)
+ u_int32_t new_value;
+ int changed = 0;
+ int error = sysctl_io_number(req, soqlencomp, sizeof(u_int32_t),
+ &new_value, &changed);
+ if (!error && changed) {
+ soqlencomp = new_value;
+ if (new_value != 0) {
+ tcp_do_autorcvbuf = 0;
+ tcptv_persmin_val = 6 * TCP_RETRANSHZ;
+ }
+ }
+ return error;
+}
+SYSCTL_PROC(_kern_ipc, OID_AUTO, soqlencomp,
+ CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
+ &soqlencomp, 0, &sysctl_soqlencomp, "IU", "");
-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, OID_AUTO, sbmb_cnt, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &total_sbmb_cnt, 0, "");
+SYSCTL_INT(_kern_ipc, OID_AUTO, sbmb_cnt_peak, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &total_sbmb_cnt_peak, 0, "");
+SYSCTL_INT(_kern_ipc, OID_AUTO, sbmb_cnt_floor, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &total_sbmb_cnt_floor, 0, "");
+SYSCTL_QUAD(_kern_ipc, OID_AUTO, sbmb_limreached, CTLFLAG_RD | CTLFLAG_LOCKED,
+ &sbmb_limreached, "");
+
+
+SYSCTL_NODE(_kern_ipc, OID_AUTO, io_policy, CTLFLAG_RW, 0, "network IO policy");
SYSCTL_INT(_kern_ipc_io_policy, OID_AUTO, log, CTLFLAG_RW | CTLFLAG_LOCKED,
- &net_io_policy_log, 0, "");
+ &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, "");
+ &net_io_policy_uuid, 0, "");
#endif /* CONFIG_PROC_UUID_POLICY */
projects / apple / xnu.git / blobdiff