#include <sys/systm.h>
#include <sys/domain.h>
#include <sys/kernel.h>
-#include <sys/proc.h>
+#include <sys/proc_internal.h>
+#include <sys/kauth.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/protosw.h>
#include <sys/signalvar.h>
#include <sys/sysctl.h>
#include <sys/ev.h>
-
+#include <kern/locks.h>
+#include <net/route.h>
+#include <netinet/in.h>
+#include <netinet/in_pcb.h>
#include <sys/kdebug.h>
#define DBG_FNC_SBDROP NETDBG_CODE(DBG_NETSOCK, 4)
* the kernel, the wakeups done here will sometimes
* cause software-interrupt process scheduling.
*/
-
void
soisconnecting(so)
register struct socket *so;
so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
so->so_state |= SS_ISCONNECTING;
+
+ sflt_notify(so, sock_evt_connecting, NULL);
}
void
struct socket *so;
{
struct socket *head = so->so_head;
- struct kextcb *kp;
-
- kp = sotokextcb(so);
- while (kp) {
- if (kp->e_soif && kp->e_soif->sf_soisconnected) {
- if ((*kp->e_soif->sf_soisconnected)(so, kp))
- return;
- }
- kp = kp->e_next;
- }
so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
so->so_state |= SS_ISCONNECTED;
+
+ sflt_notify(so, sock_evt_connected, NULL);
+
if (head && (so->so_state & SS_INCOMP)) {
- postevent(head,0,EV_RCONN);
+ so->so_state &= ~SS_INCOMP;
+ so->so_state |= SS_COMP;
+ if (head->so_proto->pr_getlock != NULL) {
+ socket_unlock(so, 0);
+ socket_lock(head, 1);
+ }
+ postevent(head, 0, EV_RCONN);
TAILQ_REMOVE(&head->so_incomp, so, so_list);
head->so_incqlen--;
- so->so_state &= ~SS_INCOMP;
TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
- so->so_state |= SS_COMP;
sorwakeup(head);
- wakeup_one(&head->so_timeo);
+ wakeup_one((caddr_t)&head->so_timeo);
+ if (head->so_proto->pr_getlock != NULL) {
+ socket_unlock(head, 1);
+ socket_lock(so, 0);
+ }
} else {
- postevent(so,0,EV_WCONN);
+ postevent(so, 0, EV_WCONN);
wakeup((caddr_t)&so->so_timeo);
sorwakeup(so);
sowwakeup(so);
soisdisconnecting(so)
register struct socket *so;
{
- register struct kextcb *kp;
-
- kp = sotokextcb(so);
- while (kp) {
- if (kp->e_soif && kp->e_soif->sf_soisdisconnecting) {
- if ((*kp->e_soif->sf_soisdisconnecting)(so, kp))
- return;
- }
- kp = kp->e_next;
- }
-
so->so_state &= ~SS_ISCONNECTING;
so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
+ sflt_notify(so, sock_evt_disconnecting, NULL);
wakeup((caddr_t)&so->so_timeo);
sowwakeup(so);
sorwakeup(so);
soisdisconnected(so)
register struct socket *so;
{
- register struct kextcb *kp;
-
- kp = sotokextcb(so);
- while (kp) {
- if (kp->e_soif && kp->e_soif->sf_soisdisconnected) {
- if ((*kp->e_soif->sf_soisdisconnected)(so, kp))
- return;
- }
- kp = kp->e_next;
- }
-
so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
+ sflt_notify(so, sock_evt_disconnected, NULL);
wakeup((caddr_t)&so->so_timeo);
sowwakeup(so);
sorwakeup(so);
sodropablereq(head)
register struct socket *head;
{
- register struct socket *so;
+ struct socket *so, *sonext = NULL;
unsigned int i, j, qlen;
static int rnd;
static struct timeval old_runtime;
so = TAILQ_FIRST(&head->so_incomp);
if (!so)
- return (so);
+ return (NULL);
qlen = head->so_incqlen;
if (++cur_cnt > qlen || old_cnt > qlen) {
rnd = (314159 * rnd + 66329) & 0xffff;
j = ((qlen + 1) * rnd) >> 16;
-
- while (j-- && so)
- so = TAILQ_NEXT(so, so_list);
+//###LD To clean up
+ while (j-- && so) {
+// if (in_pcb_checkstate(so->so_pcb, WNT_ACQUIRE, 0) != WNT_STOPUSING) {
+ socket_lock(so, 1);
+ sonext = TAILQ_NEXT(so, so_list);
+// in_pcb_check_state(so->so_pcb, WNT_RELEASE, 0);
+ socket_unlock(so, 1);
+ so = sonext;
+ }
}
- return (so);
+// if (in_pcb_checkstate(so->so_pcb, WNT_ACQUIRE, 0) == WNT_STOPUSING)
+// return (NULL);
+// else
+ return (so);
}
/*
* data structure of the original socket, and return this.
* Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
*/
-struct socket *
-sonewconn(head, connstatus)
+static struct socket *
+sonewconn_internal(head, connstatus)
register struct socket *head;
int connstatus;
{
int error = 0;
register struct socket *so;
- register struct kextcb *kp;
+ lck_mtx_t *mutex_held;
+
+ if (head->so_proto->pr_getlock != NULL)
+ mutex_held = (*head->so_proto->pr_getlock)(head, 0);
+ else
+ mutex_held = head->so_proto->pr_domain->dom_mtx;
+ lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
if (head->so_qlen > 3 * head->so_qlimit / 2)
return ((struct socket *)0);
so->so_timeo = head->so_timeo;
so->so_pgid = head->so_pgid;
so->so_uid = head->so_uid;
+ so->so_usecount = 1;
- /* Attach socket filters for this protocol */
- if (so->so_proto->pr_sfilter.tqh_first)
- error = sfilter_init(so);
- if (error != 0) {
+ if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat)) {
+ sflt_termsock(so);
sodealloc(so);
return ((struct socket *)0);
}
- /* Call socket filters' sonewconn1 function if set */
- kp = sotokextcb(so);
- while (kp) {
- if (kp->e_soif && kp->e_soif->sf_sonewconn) {
- error = (int)(*kp->e_soif->sf_sonewconn)(so, connstatus, kp);
- if (error == EJUSTRETURN) {
- return so;
- } else if (error != 0) {
- sodealloc(so);
- return NULL;
- }
- }
- kp = kp->e_next;
- }
-
- if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat) ||
- (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
- sfilter_term(so);
+ /*
+ * Must be done with head unlocked to avoid deadlock for protocol with per socket mutexes.
+ */
+ if (head->so_proto->pr_unlock)
+ socket_unlock(head, 0);
+ if (((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL) != 0) || error) {
+ sflt_termsock(so);
sodealloc(so);
+ if (head->so_proto->pr_unlock)
+ socket_lock(head, 0);
return ((struct socket *)0);
}
+ if (head->so_proto->pr_unlock)
+ socket_lock(head, 0);
#ifdef __APPLE__
so->so_proto->pr_domain->dom_refs++;
#endif
head->so_incqlen++;
}
head->so_qlen++;
- if (connstatus) {
- sorwakeup(head);
- wakeup((caddr_t)&head->so_timeo);
- so->so_state |= connstatus;
- }
#ifdef __APPLE__
so->so_rcv.sb_so = so->so_snd.sb_so = so;
TAILQ_INIT(&so->so_evlist);
+
+ /* Attach socket filters for this protocol */
+ sflt_initsock(so);
#endif
+ if (connstatus) {
+ so->so_state |= connstatus;
+ sorwakeup(head);
+ wakeup((caddr_t)&head->so_timeo);
+ }
return (so);
}
+
+struct socket *
+sonewconn(
+ struct socket *head,
+ int connstatus,
+ const struct sockaddr *from)
+{
+ int error = 0;
+ struct socket_filter_entry *filter;
+ int filtered = 0;
+
+ error = 0;
+ for (filter = head->so_filt; filter && (error == 0);
+ filter = filter->sfe_next_onsocket) {
+ if (filter->sfe_filter->sf_filter.sf_connect_in) {
+ if (filtered == 0) {
+ filtered = 1;
+ sflt_use(head);
+ socket_unlock(head, 0);
+ }
+ error = filter->sfe_filter->sf_filter.sf_connect_in(
+ filter->sfe_cookie, head, from);
+ }
+ }
+ if (filtered != 0) {
+ socket_lock(head, 0);
+ sflt_unuse(head);
+ }
+
+ if (error) {
+ return NULL;
+ }
+
+ return sonewconn_internal(head, connstatus);
+}
+
/*
* Socantsendmore indicates that no more data will be sent on the
* socket; it would normally be applied to a socket when the user
socantsendmore(so)
struct socket *so;
{
- register struct kextcb *kp;
-
- kp = sotokextcb(so);
- while (kp) {
- if (kp->e_soif && kp->e_soif->sf_socantsendmore) {
- if ((*kp->e_soif->sf_socantsendmore)(so, kp))
- return;
- }
- kp = kp->e_next;
- }
-
-
so->so_state |= SS_CANTSENDMORE;
+ sflt_notify(so, sock_evt_cantsendmore, NULL);
sowwakeup(so);
}
socantrcvmore(so)
struct socket *so;
{
- register struct kextcb *kp;
-
- kp = sotokextcb(so);
- while (kp) {
- if (kp->e_soif && kp->e_soif->sf_socantrcvmore) {
- if ((*kp->e_soif->sf_socantrcvmore)(so, kp))
- return;
- }
- kp = kp->e_next;
- }
-
-
so->so_state |= SS_CANTRCVMORE;
+ sflt_notify(so, sock_evt_cantrecvmore, NULL);
sorwakeup(so);
}
sbwait(sb)
struct sockbuf *sb;
{
+ int error = 0, lr, lr_saved;
+ struct socket *so = sb->sb_so;
+ lck_mtx_t *mutex_held;
+ struct timespec ts;
+
+#ifdef __ppc__
+ __asm__ volatile("mflr %0" : "=r" (lr));
+ lr_saved = lr;
+#endif
+
+
+ if (so->so_proto->pr_getlock != NULL)
+ mutex_held = (*so->so_proto->pr_getlock)(so, 0);
+ else
+ mutex_held = so->so_proto->pr_domain->dom_mtx;
sb->sb_flags |= SB_WAIT;
- return (tsleep((caddr_t)&sb->sb_cc,
- (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
- sb->sb_timeo));
+
+ if (so->so_usecount < 1)
+ panic("sbwait: so=%x refcount=%d\n", so, so->so_usecount);
+ ts.tv_sec = sb->sb_timeo.tv_sec;
+ ts.tv_nsec = sb->sb_timeo.tv_usec * 1000;
+ error = msleep((caddr_t)&sb->sb_cc, mutex_held,
+ (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
+ &ts);
+
+ lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
+
+ if (so->so_usecount < 1)
+ panic("sbwait: so=%x refcount=%d\n", so, so->so_usecount);
+
+ if ((so->so_state & SS_DRAINING)) {
+ error = EBADF;
+ }
+
+ return (error);
}
/*
sb_lock(sb)
register struct sockbuf *sb;
{
- int error;
+ 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;
- error = tsleep((caddr_t)&sb->sb_flags,
- (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
- "sblock", 0);
- if (error)
+ if (so->so_proto->pr_getlock != NULL)
+ mutex_held = (*so->so_proto->pr_getlock)(so, 0);
+ else
+ mutex_held = so->so_proto->pr_domain->dom_mtx;
+ if (so->so_usecount < 1)
+ panic("sb_lock: so=%x refcount=%d\n", so, so->so_usecount);
+ error = msleep((caddr_t)&sb->sb_flags, mutex_held,
+ (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sblock", 0);
+ if (so->so_usecount < 1)
+ panic("sb_lock: 2 so=%x refcount=%d\n", so, so->so_usecount);
+ if (error)
return (error);
}
sb->sb_flags |= SB_LOCK;
register struct sockbuf *sb;
{
struct proc *p = current_proc();
- /* We clear the flag before calling selwakeup. */
- /* BSD calls selwakeup then sets the flag */
sb->sb_flags &= ~SB_SEL;
selwakeup(&sb->sb_sel);
if (sb->sb_flags & SB_WAIT) {
else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
psignal(p, SIGIO);
}
- if (sb->sb_flags & SB_UPCALL)
+ if (sb->sb_flags & SB_KNOTE) {
+ KNOTE(&sb->sb_sel.si_note, SO_FILT_HINT_LOCKED);
+ }
+ if (sb->sb_flags & SB_UPCALL) {
+ socket_unlock(so, 0);
(*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
+ socket_lock(so, 0);
+ }
}
/*
register struct socket *so;
u_long sndcc, rcvcc;
{
- register struct kextcb *kp;
-
- kp = sotokextcb(so);
- while (kp) {
- if (kp->e_soif && kp->e_soif->sf_soreserve) {
- if ((*kp->e_soif->sf_soreserve)(so, sndcc, rcvcc, kp))
- return;
- }
- kp = kp->e_next;
- }
if (sbreserve(&so->so_snd, sndcc) == 0)
goto bad;
* the mbuf chain is recorded in sb. Empty mbufs are
* discarded and mbufs are compacted where possible.
*/
-void
+int
sbappend(sb, m)
struct sockbuf *sb;
struct mbuf *m;
{
- struct kextcb *kp;
- register struct mbuf *n;
+ register struct mbuf *n, *sb_first;
+ int result = 0;
+ int error = 0;
+ int filtered = 0;
KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_START), sb, m->m_len, 0, 0, 0);
if (m == 0)
- return;
- kp = sotokextcb(sbtoso(sb));
- while (kp) {
- if (kp->e_sout && kp->e_sout->su_sbappend) {
- if ((*kp->e_sout->su_sbappend)(sb, m, kp))
- return;
- }
- kp = kp->e_next;
- }
- n = sb->sb_mb;
+ return 0;
+
+again:
+ sb_first = n = sb->sb_mb;
if (n) {
while (n->m_nextpkt)
n = n->m_nextpkt;
do {
if (n->m_flags & M_EOR) {
- sbappendrecord(sb, m); /* XXXXXX!!!! */
- return;
+ result = sbappendrecord(sb, m); /* XXXXXX!!!! */
+ KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_END), sb, sb->sb_cc, 0, 0, 0);
+ return result;
}
} while (n->m_next && (n = n->m_next));
}
- sbcompress(sb, m, n);
+
+ if (!filtered && (sb->sb_flags & SB_RECV) != 0) {
+ error = sflt_data_in(sb->sb_so, NULL, &m, NULL, 0, &filtered);
+ if (error) {
+ /* no data was appended, caller should not call sowakeup */
+ return 0;
+ }
+
+ /*
+ If we any filters, the socket lock was dropped. n and sb_first
+ cached data from the socket buffer. This cache is not valid
+ since we dropped the lock. We must start over. Since filtered
+ is set we won't run through the filters a second time. We just
+ set n and sb_start again.
+ */
+ if (filtered)
+ goto again;
+ }
+
+ result = sbcompress(sb, m, n);
KERNEL_DEBUG((DBG_FNC_SBAPPEND | DBG_FUNC_END), sb, sb->sb_cc, 0, 0, 0);
+
+ return result;
}
#ifdef SOCKBUF_DEBUG
register struct mbuf *m;
register struct mbuf *n = 0;
register u_long len = 0, mbcnt = 0;
+ lck_mtx_t *mutex_held;
+
+ if (sb->sb_so->so_proto->pr_getlock != NULL)
+ mutex_held = (*sb->sb_so->so_proto->pr_getlock)(sb->sb_so, 0);
+ else
+ mutex_held = sb->sb_so->so_proto->pr_domain->dom_mtx;
+
+ lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
+
+ if (sbchecking == 0)
+ return;
for (m = sb->sb_mb; m; m = n) {
n = m->m_nextpkt;
mbcnt += m->m_ext.ext_size;
}
}
-#ifndef __APPLE__
if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
- printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
+ panic("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
mbcnt, sb->sb_mbcnt);
- panic("sbcheck");
}
-#else
- if (len != sb->sb_cc)
- printf("sbcheck len %ld != sb_cc %ld\n", len, sb->sb_cc);
- if (mbcnt != sb->sb_mbcnt)
- printf("sbcheck mbcnt %ld != sb_mbcnt %ld\n", mbcnt, sb->sb_mbcnt);
-#endif
}
#endif
* As above, except the mbuf chain
* begins a new record.
*/
-void
+int
sbappendrecord(sb, m0)
register struct sockbuf *sb;
register struct mbuf *m0;
{
register struct mbuf *m;
- register struct kextcb *kp;
+ int result = 0;
if (m0 == 0)
- return;
-
- kp = sotokextcb(sbtoso(sb));
- while (kp)
- { if (kp->e_sout && kp->e_sout->su_sbappendrecord)
- { if ((*kp->e_sout->su_sbappendrecord)(sb, m0, kp))
- return;
+ return 0;
+
+ if ((sb->sb_flags & SB_RECV) != 0) {
+ int error = sflt_data_in(sb->sb_so, NULL, &m0, NULL, sock_data_filt_flag_record, NULL);
+ if (error != 0) {
+ if (error != EJUSTRETURN)
+ m_freem(m0);
+ return 0;
}
- kp = kp->e_next;
}
m = sb->sb_mb;
m0->m_flags &= ~M_EOR;
m->m_flags |= M_EOR;
}
- sbcompress(sb, m, m0);
+ return sbcompress(sb, m, m0);
}
/*
* is inserted at the beginning of the sockbuf,
* but after any other OOB data.
*/
-void
+int
sbinsertoob(sb, m0)
- register struct sockbuf *sb;
- register struct mbuf *m0;
+ struct sockbuf *sb;
+ struct mbuf *m0;
{
- register struct mbuf *m;
- register struct mbuf **mp;
- register struct kextcb *kp;
+ struct mbuf *m;
+ struct mbuf **mp;
if (m0 == 0)
- return;
-
- kp = sotokextcb(sbtoso(sb));
- while (kp)
- { if (kp->e_sout && kp->e_sout->su_sbinsertoob)
- { if ((*kp->e_sout->su_sbinsertoob)(sb, m0, kp))
- return;
+ return 0;
+
+ if ((sb->sb_flags & SB_RECV) != 0) {
+ int error = sflt_data_in(sb->sb_so, NULL, &m0, NULL,
+ sock_data_filt_flag_oob, NULL);
+
+ if (error) {
+ if (error != EJUSTRETURN) {
+ m_freem(m0);
+ }
+ return 0;
}
- kp = kp->e_next;
}
for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
m0->m_flags &= ~M_EOR;
m->m_flags |= M_EOR;
}
- sbcompress(sb, m, m0);
+ return sbcompress(sb, m, m0);
}
/*
* m0 must include a packet header with total length.
* Returns 0 if no space in sockbuf or insufficient mbufs.
*/
-int
-sbappendaddr(sb, asa, m0, control)
+static int
+sbappendaddr_internal(sb, asa, m0, control)
register struct sockbuf *sb;
struct sockaddr *asa;
struct mbuf *m0, *control;
{
register struct mbuf *m, *n;
int space = asa->sa_len;
- register struct kextcb *kp;
if (m0 && (m0->m_flags & M_PKTHDR) == 0)
panic("sbappendaddr");
- kp = sotokextcb(sbtoso(sb));
- while (kp)
- { if (kp->e_sout && kp->e_sout->su_sbappendaddr)
- { if ((*kp->e_sout->su_sbappendaddr)(sb, asa, m0, control, kp))
- return 0;
- }
- kp = kp->e_next;
- }
-
if (m0)
space += m0->m_pkthdr.len;
for (n = control; n; n = n->m_next) {
}
int
-sbappendcontrol(sb, m0, control)
+sbappendaddr(
+ struct sockbuf* sb,
+ struct sockaddr* asa,
+ struct mbuf *m0,
+ struct mbuf *control,
+ int *error_out)
+{
+ int result = 0;
+
+ if (error_out) *error_out = 0;
+
+ if (m0 && (m0->m_flags & M_PKTHDR) == 0)
+ panic("sbappendaddrorfree");
+
+ /* Call socket data in filters */
+ if ((sb->sb_flags & SB_RECV) != 0) {
+ int error;
+ error = sflt_data_in(sb->sb_so, asa, &m0, &control, 0, NULL);
+ if (error) {
+ if (error != EJUSTRETURN) {
+ if (m0) m_freem(m0);
+ if (control) m_freem(control);
+ if (error_out) *error_out = error;
+ }
+ return 0;
+ }
+ }
+
+ result = sbappendaddr_internal(sb, asa, m0, control);
+ if (result == 0) {
+ if (m0) m_freem(m0);
+ if (control) m_freem(control);
+ if (error_out) *error_out = ENOBUFS;
+ }
+
+ return result;
+}
+
+static int
+sbappendcontrol_internal(sb, m0, control)
struct sockbuf *sb;
struct mbuf *control, *m0;
{
register struct mbuf *m, *n;
int space = 0;
- register struct kextcb *kp;
if (control == 0)
panic("sbappendcontrol");
- kp = sotokextcb(sbtoso(sb));
- while (kp)
- { if (kp->e_sout && kp->e_sout->su_sbappendcontrol)
- { if ((*kp->e_sout->su_sbappendcontrol)(sb, m0, control, kp))
- return 0;
- }
- kp = kp->e_next;
- }
-
for (m = control; ; m = m->m_next) {
space += m->m_len;
if (m->m_next == 0)
return (1);
}
+int
+sbappendcontrol(
+ struct sockbuf *sb,
+ struct mbuf *m0,
+ struct mbuf *control,
+ int *error_out)
+{
+ int result = 0;
+
+ if (error_out) *error_out = 0;
+
+ if (sb->sb_flags & SB_RECV) {
+ int error;
+ error = sflt_data_in(sb->sb_so, NULL, &m0, &control, 0, NULL);
+ if (error) {
+ if (error != EJUSTRETURN) {
+ if (m0) m_freem(m0);
+ if (control) m_freem(control);
+ if (error_out) *error_out = error;
+ }
+ return 0;
+ }
+ }
+
+ result = sbappendcontrol_internal(sb, m0, control);
+ if (result == 0) {
+ if (m0) m_freem(m0);
+ if (control) m_freem(control);
+ if (error_out) *error_out = ENOBUFS;
+ }
+
+ return result;
+}
+
/*
* Compress mbuf chain m into the socket
* buffer sb following mbuf n. If n
* is null, the buffer is presumed empty.
*/
-void
+static int
sbcompress(sb, m, n)
register struct sockbuf *sb;
register struct mbuf *m, *n;
printf("semi-panic: sbcompress\n");
}
postevent(0,sb, EV_RWBYTES);
+ return 1;
}
/*
sbflush(sb)
register struct sockbuf *sb;
{
- register struct kextcb *kp;
-
- kp = sotokextcb(sbtoso(sb));
- while (kp) {
- if (kp->e_sout && kp->e_sout->su_sbflush) {
- if ((*kp->e_sout->su_sbflush)(sb, kp))
- return;
- }
- kp = kp->e_next;
- }
-
- if (sb->sb_flags & SB_LOCK)
- sb_lock(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) {
/*
* Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
break;
sbdrop(sb, (int)sb->sb_cc);
}
- if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
- panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
+ 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);
+
postevent(0, sb, EV_RWBYTES);
+ sbunlock(sb, 1); /* keep socket locked */
+
}
/*
{
register struct mbuf *m, *free_list, *ml;
struct mbuf *next, *last;
- register struct kextcb *kp;
KERNEL_DEBUG((DBG_FNC_SBDROP | DBG_FUNC_START), sb, len, 0, 0, 0);
- kp = sotokextcb(sbtoso(sb));
- while (kp) {
- if (kp->e_sout && kp->e_sout->su_sbdrop) {
- if ((*kp->e_sout->su_sbdrop)(sb, len, kp))
- return;
- }
- kp = kp->e_next;
- }
next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
free_list = last = m;
ml = (struct mbuf *)0;
register struct sockbuf *sb;
{
register struct mbuf *m, *mn;
- register struct kextcb *kp;
-
- kp = sotokextcb(sbtoso(sb));
- while (kp) {
- if (kp->e_sout && kp->e_sout->su_sbdroprecord) {
- if ((*kp->e_sout->su_sbdroprecord)(sb, kp))
- return;
- }
- kp = kp->e_next;
- }
m = sb->sb_mb;
if (m) {
}
-int pru_sopoll_notsupp(struct socket *so, int events,
- struct ucred *cred)
+int
+pru_sopoll_notsupp(__unused struct socket *so, __unused int events,
+ __unused kauth_cred_t cred, __unused void *wql)
{
return EOPNOTSUPP;
}
int
sb_notify(struct sockbuf *sb)
{
- return ((sb->sb_flags & (SB_WAIT|SB_SEL|SB_ASYNC|SB_UPCALL)) != 0);
+ return ((sb->sb_flags & (SB_WAIT|SB_SEL|SB_ASYNC|SB_UPCALL|SB_KNOTE)) != 0);
}
/*
/* release lock on sockbuf sb */
void
-sbunlock(struct sockbuf *sb)
+sbunlock(struct sockbuf *sb, int keeplocked)
{
+ struct socket *so = sb->sb_so;
+ int lr, lr_saved;
+ lck_mtx_t *mutex_held;
+
+#ifdef __ppc__
+ __asm__ volatile("mflr %0" : "=r" (lr));
+ lr_saved = lr;
+#endif
sb->sb_flags &= ~SB_LOCK;
+
+ if (so->so_proto->pr_getlock != NULL)
+ mutex_held = (*so->so_proto->pr_getlock)(so, 0);
+ else
+ mutex_held = so->so_proto->pr_domain->dom_mtx;
+
+ if (keeplocked == 0)
+ lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
+
if (sb->sb_flags & SB_WANT) {
sb->sb_flags &= ~SB_WANT;
+ if (so->so_usecount < 0)
+ panic("sbunlock: b4 wakeup so=%x ref=%d lr=%x sb_flags=%x\n", sb->sb_so, so->so_usecount, lr_saved, sb->sb_flags);
+
wakeup((caddr_t)&(sb)->sb_flags);
}
+ if (keeplocked == 0) { /* unlock on exit */
+ so->so_usecount--;
+ if (so->so_usecount < 0)
+ panic("sbunlock: unlock on exit so=%x lr=%x sb_flags=%x\n", so, so->so_usecount,lr_saved, sb->sb_flags);
+ so->reserved4= lr_saved;
+ lck_mtx_unlock(mutex_held);
+ }
}
void
xso->so_linger = so->so_linger;
xso->so_state = so->so_state;
xso->so_pcb = so->so_pcb;
- xso->xso_protocol = so->so_proto->pr_protocol;
- xso->xso_family = so->so_proto->pr_domain->dom_family;
+ if (so->so_proto) {
+ xso->xso_protocol = so->so_proto->pr_protocol;
+ xso->xso_family = so->so_proto->pr_domain->dom_family;
+ }
+ else
+ xso->xso_protocol = xso->xso_family = 0;
xso->so_qlen = so->so_qlen;
xso->so_incqlen = so->so_incqlen;
xso->so_qlimit = so->so_qlimit;
xsb->sb_mbmax = sb->sb_mbmax;
xsb->sb_lowat = sb->sb_lowat;
xsb->sb_flags = sb->sb_flags;
- xsb->sb_timeo = sb->sb_timeo;
+ xsb->sb_timeo = (u_long)(sb->sb_timeo.tv_sec * hz) + sb->sb_timeo.tv_usec / tick;
+ if (xsb->sb_timeo == 0 && sb->sb_timeo.tv_usec != 0)
+ xsb->sb_timeo = 1;
}
/*