/*
* Partition an mbuf chain in two pieces, returning the tail --
* all but the first len0 bytes. In case of failure, it returns NULL and
* attempts to restore the chain to its original state.
*/
struct mbuf *
m_split(struct mbuf *m0, int len0, int wait)
{
struct mbuf *m, *n;
unsigned len = len0, remain, olen;
for (m = m0; m && len > m->m_len; m = m->m_next)
len -= m->m_len;
if (m == NULL)
return (NULL);
remain = m->m_len - len;
if (m0->m_flags & M_PKTHDR) {
MGETHDR(n, wait, m0->m_type);
if (n == NULL)
return (NULL);
if (m_dup_pkthdr(n, m0, wait)) {
m_freem(n);
return (NULL);
}
n->m_pkthdr.len -= len0;
olen = m0->m_pkthdr.len;
m0->m_pkthdr.len = len0;
if (m->m_flags & M_EXT)
goto extpacket;
if (remain > MHLEN) {
/* m can't be the lead packet */
MH_ALIGN(n, 0);
n->m_next = m_split(m, len, wait);
if (n->m_next == NULL) {
(void) m_free(n);
m0->m_pkthdr.len = olen;
return (NULL);
} else
return (n);
} else
MH_ALIGN(n, remain);
} else if (remain == 0) {
n = m->m_next;
m->m_next = NULL;
return (n);
} else {
MGET(n, wait, m->m_type);
if (n == NULL)
return (NULL);
M_ALIGN(n, remain);
}
extpacket:
if (m->m_flags & M_EXT) {
n->m_ext = m->m_ext;
MCLADDREFERENCE(m, n);
n->m_data = m->m_data + len;
} else {
memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + len, remain);
}
n->m_len = remain;
m->m_len = len;
n->m_next = m->m_next;
m->m_next = NULL;
return (n);
}
/*
* Builds and sends a single ARP request to locate the server
*
* Return value:
* 0 on success
* errno on error
*/
static int
netdump_send_arp(in_addr_t dst)
{
struct ether_addr bcast;
struct mbuf *m;
struct arphdr *ah;
int pktlen;
MPASS(nd_ifp != NULL);
/* Fill-up a broadcast address. */
memset(&bcast, 0xFF, ETHER_ADDR_LEN);
m = m_gethdr(M_NOWAIT, MT_DATA);
if (m == NULL) {
printf("netdump_send_arp: Out of mbufs\n");
return (ENOBUFS);
}
pktlen = arphdr_len2(ETHER_ADDR_LEN, sizeof(struct in_addr));
m->m_len = pktlen;
m->m_pkthdr.len = pktlen;
MH_ALIGN(m, pktlen);
ah = mtod(m, struct arphdr *);
ah->ar_hrd = htons(ARPHRD_ETHER);
ah->ar_pro = htons(ETHERTYPE_IP);
ah->ar_hln = ETHER_ADDR_LEN;
ah->ar_pln = sizeof(struct in_addr);
ah->ar_op = htons(ARPOP_REQUEST);
memcpy(ar_sha(ah), IF_LLADDR(nd_ifp), ETHER_ADDR_LEN);
((struct in_addr *)ar_spa(ah))->s_addr = nd_client.s_addr;
bzero(ar_tha(ah), ETHER_ADDR_LEN);
((struct in_addr *)ar_tpa(ah))->s_addr = dst;
return (netdump_ether_output(m, nd_ifp, bcast, ETHERTYPE_ARP));
}
/*
* Allocate and setup a management frame of the specified
* size. We return the mbuf and a pointer to the start
* of the contiguous data area that's been reserved based
* on the packet length. The data area is forced to 32-bit
* alignment and the buffer length to a multiple of 4 bytes.
* This is done mainly so beacon frames (that require this)
* can use this interface too.
*/
struct mbuf *
ieee80211_getmgtframe(uint8_t **frm, int headroom, int pktlen)
{
struct mbuf *m;
u_int len;
/*
* NB: we know the mbuf routines will align the data area
* so we don't need to do anything special.
*/
len = roundup2(headroom + pktlen, 4);
KASSERT(len <= MCLBYTES, ("802.11 mgt frame too large: %u", len));
if (len < MINCLSIZE) {
m = m_gethdr(M_NOWAIT, MT_DATA);
/*
* Align the data in case additional headers are added.
* This should only happen when a WEP header is added
* which only happens for shared key authentication mgt
* frames which all fit in MHLEN.
*/
if (m != NULL)
MH_ALIGN(m, len);
} else {
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
if (m != NULL)
MC_ALIGN(m, len);
}
if (m != NULL) {
m->m_data += headroom;
*frm = m->m_data;
}
return m;
}
/*
* Ensure len bytes of contiguous space at the beginning of the mbuf chain
*/
struct mbuf *
m_prepend(struct mbuf *m, int len, int how)
{
struct mbuf *mn;
if (len > MHLEN)
panic("mbuf prepend length too big");
if (M_LEADINGSPACE(m) >= len) {
m->m_data -= len;
m->m_len += len;
} else {
MGET(mn, how, m->m_type);
if (mn == NULL) {
m_freem(m);
return (NULL);
}
if (m->m_flags & M_PKTHDR)
M_MOVE_PKTHDR(mn, m);
mn->m_next = m;
m = mn;
MH_ALIGN(m, len);
m->m_len = len;
}
if (m->m_flags & M_PKTHDR)
m->m_pkthdr.len += len;
return (m);
}
static void
igmp_sendpkt(struct in_multi *inm, int type, unsigned long addr)
{
struct mbuf *m;
struct igmp *igmp;
struct ip *ip;
struct ip_moptions imo;
MGETHDR(m, M_DONTWAIT, MT_HEADER);
if (m == NULL)
return;
m->m_pkthdr.rcvif = loif;
m->m_pkthdr.len = sizeof(struct ip) + IGMP_MINLEN;
MH_ALIGN(m, IGMP_MINLEN + sizeof(struct ip));
m->m_data += sizeof(struct ip);
m->m_len = IGMP_MINLEN;
igmp = mtod(m, struct igmp *);
igmp->igmp_type = type;
igmp->igmp_code = 0;
igmp->igmp_group = inm->inm_addr;
igmp->igmp_cksum = 0;
igmp->igmp_cksum = in_cksum(m, IGMP_MINLEN);
m->m_data -= sizeof(struct ip);
m->m_len += sizeof(struct ip);
ip = mtod(m, struct ip *);
ip->ip_tos = 0;
ip->ip_len = sizeof(struct ip) + IGMP_MINLEN;
ip->ip_off = 0;
ip->ip_p = IPPROTO_IGMP;
ip->ip_src.s_addr = INADDR_ANY;
ip->ip_dst.s_addr = addr ? addr : igmp->igmp_group.s_addr;
imo.imo_multicast_ifp = inm->inm_ifp;
imo.imo_multicast_ttl = 1;
imo.imo_multicast_vif = -1;
/*
* Request loopback of the report if we are acting as a multicast
* router, so that the process-level routing demon can hear it.
*/
imo.imo_multicast_loop = (ip_mrouter != NULL);
/*
* XXX
* Do we have to worry about reentrancy here? Don't think so.
*/
ip_output(m, router_alert, &igmprt, 0, &imo);
++igmpstat.igps_snd_reports;
}
/*
* Lesser-used path for M_PREPEND:
* allocate new mbuf to prepend to chain,
* copy junk along.
*/
struct mbuf *
m_prepend(struct mbuf *m, int len, int how)
{
struct mbuf *mn;
if (len > MHLEN)
panic("mbuf prepend length too big");
MGET(mn, how, m->m_type);
if (mn == NULL) {
m_freem(m);
return (NULL);
}
if (m->m_flags & M_PKTHDR)
M_MOVE_PKTHDR(mn, m);
mn->m_next = m;
m = mn;
MH_ALIGN(m, len);
m->m_len = len;
return (m);
}
struct mbuf*
ieee80211_getmgtframe(uint8_t** frm, int headroom, int pktlen)
{
struct mbuf* m;
u_int len;
len = roundup2(headroom + pktlen, 4);
KASSERT(len <= MCLBYTES, ("802.11 mgt frame too large: %u", len));
if (len < MINCLSIZE) {
m = m_gethdr(M_NOWAIT, MT_DATA);
if (m != NULL)
MH_ALIGN(m, len);
} else {
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
if (m != NULL)
MC_ALIGN(m, len);
}
if (m != NULL) {
m->m_data += headroom;
*frm = (uint8_t*)m->m_data;
}
return m;
}
int
isc_sendPDU(isc_session_t *sp, pduq_t *pq)
{
struct mbuf *mh, **mp;
pdu_t *pp = &pq->pdu;
int len, error;
debug_called(8);
/*
| mbuf for the iSCSI header
*/
MGETHDR(mh, MB_TRYWAIT, MT_DATA);
mh->m_len = mh->m_pkthdr.len = sizeof(union ipdu_u);
mh->m_pkthdr.rcvif = NULL;
MH_ALIGN(mh, sizeof(union ipdu_u));
bcopy(&pp->ipdu, mh->m_data, sizeof(union ipdu_u));
mh->m_next = NULL;
if(sp->hdrDigest)
pq->pdu.hdr_dig = sp->hdrDigest(&pp->ipdu, sizeof(union ipdu_u), 0);
if(pp->ahs_len) {
/*
| Add any AHS to the iSCSI hdr mbuf
| XXX Assert: (mh->m_pkthdr.len + pp->ahs_len) < MHLEN
*/
bcopy(pp->ahs, (mh->m_data + mh->m_len), pp->ahs_len);
mh->m_len += pp->ahs_len;
mh->m_pkthdr.len += pp->ahs_len;
if(sp->hdrDigest)
pq->pdu.hdr_dig = sp->hdrDigest(&pp->ahs, pp->ahs_len, pq->pdu.hdr_dig);
}
if(sp->hdrDigest) {
debug(2, "hdr_dig=%x", pq->pdu.hdr_dig);
/*
| Add header digest to the iSCSI hdr mbuf
| XXX Assert: (mh->m_pkthdr.len + 4) < MHLEN
*/
bcopy(&pp->hdr_dig, (mh->m_data + mh->m_len), sizeof(int));
mh->m_len += sizeof(int);
mh->m_pkthdr.len += sizeof(int);
}
mp = &mh->m_next;
if(pq->pdu.ds) {
struct mbuf *md;
int off = 0;
len = pp->ds_len;
while(len & 03) // the specs say it must be int alligned
len++;
while(len > 0) {
int l;
MGET(md, MB_TRYWAIT, MT_DATA);
pq->refcnt++;
l = min(MCLBYTES, len);
debug(5, "setting ext_free(arg=%p len/l=%d/%d)", pq->buf, len, l);
md->m_ext.ext_buf = pq->buf;
md->m_ext.ext_free = ext_free;
md->m_ext.ext_ref = ext_ref;
md->m_ext.ext_arg = pq;
md->m_ext.ext_size = l;
md->m_flags |= M_EXT;
md->m_data = pp->ds + off;
md->m_len = l;
md->m_next = NULL;
mh->m_pkthdr.len += l;
*mp = md;
mp = &md->m_next;
len -= l;
off += l;
}
}
if(sp->dataDigest) {
struct mbuf *me;
pp->ds_dig = sp->dataDigest(pp->ds, pp->ds_len, 0);
MGET(me, MB_TRYWAIT, MT_DATA);
me->m_len = sizeof(int);
MH_ALIGN(mh, sizeof(int));
bcopy(&pp->ds_dig, me->m_data, sizeof(int));
me->m_next = NULL;
mh->m_pkthdr.len += sizeof(int);
*mp = me;
}
if((error = sosend(sp->soc, NULL, NULL, mh, 0, 0, curthread)) != 0) {
sdebug(3, "error=%d", error);
return error;
}
sp->stats.nsent++;
getmicrouptime(&sp->stats.t_sent);
return 0;
}
/*
* Construct and reliably send a netdump packet. May fail from a resource
* shortage or extreme number of unacknowledged retransmissions. Wait for
* an acknowledgement before returning. Splits packets into chunks small
* enough to be sent without fragmentation (looks up the interface MTU)
*
* Parameters:
* type netdump packet type (HERALD, FINISHED, or VMCORE)
* offset vmcore data offset (bytes)
* data vmcore data
* datalen vmcore data size (bytes)
*
* Returns:
* int see errno.h, 0 for success
*/
static int
netdump_send(uint32_t type, off_t offset, unsigned char *data, uint32_t datalen)
{
struct netdump_msg_hdr *nd_msg_hdr;
struct mbuf *m, *m2;
uint64_t want_acks;
uint32_t i, pktlen, sent_so_far;
int retries, polls, error;
want_acks = 0;
rcvd_acks = 0;
retries = 0;
MPASS(nd_ifp != NULL);
retransmit:
/* Chunks can be too big to fit in packets. */
for (i = sent_so_far = 0; sent_so_far < datalen ||
(i == 0 && datalen == 0); i++) {
pktlen = datalen - sent_so_far;
/* First bound: the packet structure. */
pktlen = min(pktlen, NETDUMP_DATASIZE);
/* Second bound: the interface MTU (assume no IP options). */
pktlen = min(pktlen, nd_ifp->if_mtu - sizeof(struct udpiphdr) -
sizeof(struct netdump_msg_hdr));
/*
* Check if it is retransmitting and this has been ACKed
* already.
*/
if ((rcvd_acks & (1 << i)) != 0) {
sent_so_far += pktlen;
continue;
}
/*
* Get and fill a header mbuf, then chain data as an extended
* mbuf.
*/
m = m_gethdr(M_NOWAIT, MT_DATA);
if (m == NULL) {
printf("netdump_send: Out of mbufs\n");
return (ENOBUFS);
}
m->m_len = sizeof(struct netdump_msg_hdr);
m->m_pkthdr.len = sizeof(struct netdump_msg_hdr);
MH_ALIGN(m, sizeof(struct netdump_msg_hdr));
nd_msg_hdr = mtod(m, struct netdump_msg_hdr *);
nd_msg_hdr->mh_seqno = htonl(nd_seqno + i);
nd_msg_hdr->mh_type = htonl(type);
nd_msg_hdr->mh_offset = htobe64(offset + sent_so_far);
nd_msg_hdr->mh_len = htonl(pktlen);
nd_msg_hdr->mh__pad = 0;
if (pktlen != 0) {
m2 = m_get(M_NOWAIT, MT_DATA);
if (m2 == NULL) {
m_freem(m);
printf("netdump_send: Out of mbufs\n");
return (ENOBUFS);
}
MEXTADD(m2, data + sent_so_far, pktlen,
netdump_mbuf_free, NULL, NULL, 0, EXT_DISPOSABLE);
m2->m_len = pktlen;
m_cat(m, m2);
m->m_pkthdr.len += pktlen;
}
error = netdump_udp_output(m);
if (error != 0)
return (error);
/* Note that we're waiting for this packet in the bitfield. */
want_acks |= (1 << i);
sent_so_far += pktlen;
}
if (i >= NETDUMP_MAX_IN_FLIGHT)
printf("Warning: Sent more than %d packets (%d). "
"Acknowledgements will fail unless the size of "
"rcvd_acks/want_acks is increased.\n",
NETDUMP_MAX_IN_FLIGHT, i);
/*
* Wait for acks. A *real* window would speed things up considerably.
*/
polls = 0;
while (rcvd_acks != want_acks) {
if (polls++ > nd_polls) {
if (retries++ > nd_retries)
return (ETIMEDOUT);
printf(". ");
goto retransmit;
}
netdump_network_poll();
DELAY(500);
}
nd_seqno += i;
return (0);
}
static void
aarpwhohas( struct arpcom *ac, struct sockaddr_at *sat )
{
struct mbuf *m;
struct ether_header *eh;
struct ether_aarp *ea;
struct at_ifaddr *aa;
struct llc *llc;
struct sockaddr sa;
if (( m = m_gethdr( M_DONTWAIT, MT_DATA )) == NULL ) {
return;
}
#ifdef MAC
mac_create_mbuf_linklayer(&ac->ac_if, m);
#endif
m->m_len = sizeof( *ea );
m->m_pkthdr.len = sizeof( *ea );
MH_ALIGN( m, sizeof( *ea ));
ea = mtod( m, struct ether_aarp *);
bzero((caddr_t)ea, sizeof( *ea ));
ea->aarp_hrd = htons( AARPHRD_ETHER );
ea->aarp_pro = htons( ETHERTYPE_AT );
ea->aarp_hln = sizeof( ea->aarp_sha );
ea->aarp_pln = sizeof( ea->aarp_spu );
ea->aarp_op = htons( AARPOP_REQUEST );
bcopy((caddr_t)ac->ac_enaddr, (caddr_t)ea->aarp_sha,
sizeof( ea->aarp_sha ));
/*
* We need to check whether the output ethernet type should
* be phase 1 or 2. We have the interface that we'll be sending
* the aarp out. We need to find an AppleTalk network on that
* interface with the same address as we're looking for. If the
* net is phase 2, generate an 802.2 and SNAP header.
*/
if ((aa = at_ifawithnet( sat )) == NULL) {
m_freem( m );
return;
}
eh = (struct ether_header *)sa.sa_data;
if ( aa->aa_flags & AFA_PHASE2 ) {
bcopy((caddr_t)atmulticastaddr, (caddr_t)eh->ether_dhost,
sizeof( eh->ether_dhost ));
eh->ether_type = htons(sizeof(struct llc) + sizeof(struct ether_aarp));
M_PREPEND( m, sizeof( struct llc ), M_TRYWAIT );
llc = mtod( m, struct llc *);
llc->llc_dsap = llc->llc_ssap = LLC_SNAP_LSAP;
llc->llc_control = LLC_UI;
bcopy( aarp_org_code, llc->llc_org_code, sizeof( aarp_org_code ));
llc->llc_ether_type = htons( ETHERTYPE_AARP );
bcopy( &AA_SAT( aa )->sat_addr.s_net, ea->aarp_spnet,
sizeof( ea->aarp_spnet ));
bcopy( &sat->sat_addr.s_net, ea->aarp_tpnet,
sizeof( ea->aarp_tpnet ));
ea->aarp_spnode = AA_SAT( aa )->sat_addr.s_node;
ea->aarp_tpnode = sat->sat_addr.s_node;
} else {
/*
* Slightly changed version of sosend()
*/
static int
kttcp_sosend(struct socket *so, unsigned long long slen,
unsigned long long *done, struct lwp *l, int flags)
{
struct mbuf **mp, *m, *top;
long space, len, mlen;
int error, dontroute, atomic;
long long resid;
atomic = sosendallatonce(so);
resid = slen;
top = NULL;
/*
* In theory resid should be unsigned.
* However, space must be signed, as it might be less than 0
* if we over-committed, and we must use a signed comparison
* of space and resid. On the other hand, a negative resid
* causes us to loop sending 0-length segments to the protocol.
*/
if (resid < 0) {
error = EINVAL;
goto out;
}
dontroute =
(flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
(so->so_proto->pr_flags & PR_ATOMIC);
l->l_ru.ru_msgsnd++;
#define snderr(errno) { error = errno; goto release; }
solock(so);
restart:
if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
goto out;
do {
if (so->so_state & SS_CANTSENDMORE)
snderr(EPIPE);
if (so->so_error) {
error = so->so_error;
so->so_error = 0;
goto release;
}
if ((so->so_state & SS_ISCONNECTED) == 0) {
if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
snderr(ENOTCONN);
} else {
snderr(EDESTADDRREQ);
}
}
space = sbspace(&so->so_snd);
if (flags & MSG_OOB)
space += 1024;
if ((atomic && resid > so->so_snd.sb_hiwat))
snderr(EMSGSIZE);
if (space < resid && (atomic || space < so->so_snd.sb_lowat)) {
if (so->so_state & SS_NBIO)
snderr(EWOULDBLOCK);
SBLASTRECORDCHK(&so->so_rcv,
"kttcp_soreceive sbwait 1");
SBLASTMBUFCHK(&so->so_rcv,
"kttcp_soreceive sbwait 1");
sbunlock(&so->so_snd);
error = sbwait(&so->so_snd);
if (error)
goto out;
goto restart;
}
mp = ⊤
do {
sounlock(so);
do {
if (top == 0) {
m = m_gethdr(M_WAIT, MT_DATA);
mlen = MHLEN;
m->m_pkthdr.len = 0;
m->m_pkthdr.rcvif = NULL;
} else {
m = m_get(M_WAIT, MT_DATA);
mlen = MLEN;
}
if (resid >= MINCLSIZE && space >= MCLBYTES) {
m_clget(m, M_WAIT);
if ((m->m_flags & M_EXT) == 0)
goto nopages;
mlen = MCLBYTES;
#ifdef MAPPED_MBUFS
len = lmin(MCLBYTES, resid);
#else
if (atomic && top == 0) {
len = lmin(MCLBYTES - max_hdr,
resid);
m->m_data += max_hdr;
} else
len = lmin(MCLBYTES, resid);
#endif
space -= len;
} else {
nopages:
len = lmin(lmin(mlen, resid), space);
space -= len;
/*
* For datagram protocols, leave room
* for protocol headers in first mbuf.
*/
if (atomic && top == 0 && len < mlen)
MH_ALIGN(m, len);
}
resid -= len;
m->m_len = len;
*mp = m;
top->m_pkthdr.len += len;
if (error)
goto release;
mp = &m->m_next;
if (resid <= 0) {
if (flags & MSG_EOR)
top->m_flags |= M_EOR;
break;
}
} while (space > 0 && atomic);
solock(so);
if (so->so_state & SS_CANTSENDMORE)
snderr(EPIPE);
if (dontroute)
so->so_options |= SO_DONTROUTE;
if (resid > 0)
so->so_state |= SS_MORETOCOME;
if (flags & MSG_OOB)
error = (*so->so_proto->pr_usrreqs->pr_sendoob)(so,
top, NULL);
else
error = (*so->so_proto->pr_usrreqs->pr_send)(so,
top, NULL, NULL, l);
if (dontroute)
so->so_options &= ~SO_DONTROUTE;
if (resid > 0)
so->so_state &= ~SS_MORETOCOME;
top = 0;
mp = ⊤
if (error)
goto release;
} while (resid && space > 0);
} while (resid);
release:
sbunlock(&so->so_snd);
out:
sounlock(so);
if (top)
m_freem(top);
*done = slen - resid;
#if 0
printf("sosend: error %d slen %llu resid %lld\n", error, slen, resid);
#endif
return (error);
}
/*
* Send on a socket.
* If send must go all at once and message is larger than
* send buffering, then hard error.
* Lock against other senders.
* If must go all at once and not enough room now, then
* inform user that this would block and do nothing.
* Otherwise, if nonblocking, send as much as possible.
* The data to be sent is described by "uio" if nonzero,
* otherwise by the mbuf chain "top" (which must be null
* if uio is not). Data provided in mbuf chain must be small
* enough to send all at once.
*
* Returns nonzero on error, timeout or signal; callers
* must check for short counts if EINTR/ERESTART are returned.
* Data and control buffers are freed on return.
*/
int
sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
struct mbuf *top, struct mbuf *control, int flags,
struct thread *td)
{
struct mbuf **mp;
struct mbuf *m;
size_t resid;
int space, len;
int clen = 0, error, dontroute, mlen;
int atomic = sosendallatonce(so) || top;
int pru_flags;
if (uio) {
resid = uio->uio_resid;
} else {
resid = (size_t)top->m_pkthdr.len;
#ifdef INVARIANTS
len = 0;
for (m = top; m; m = m->m_next)
len += m->m_len;
KKASSERT(top->m_pkthdr.len == len);
#endif
}
/*
* WARNING! resid is unsigned, space and len are signed. space
* can wind up negative if the sockbuf is overcommitted.
*
* Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
* type sockets since that's an error.
*/
if (so->so_type == SOCK_STREAM && (flags & MSG_EOR)) {
error = EINVAL;
goto out;
}
dontroute =
(flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
(so->so_proto->pr_flags & PR_ATOMIC);
if (td->td_lwp != NULL)
td->td_lwp->lwp_ru.ru_msgsnd++;
if (control)
clen = control->m_len;
#define gotoerr(errcode) { error = errcode; goto release; }
restart:
error = ssb_lock(&so->so_snd, SBLOCKWAIT(flags));
if (error)
goto out;
do {
if (so->so_state & SS_CANTSENDMORE)
gotoerr(EPIPE);
if (so->so_error) {
error = so->so_error;
so->so_error = 0;
goto release;
}
if ((so->so_state & SS_ISCONNECTED) == 0) {
/*
* `sendto' and `sendmsg' is allowed on a connection-
* based socket if it supports implied connect.
* Return ENOTCONN if not connected and no address is
* supplied.
*/
if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
(so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
if ((so->so_state & SS_ISCONFIRMING) == 0 &&
!(resid == 0 && clen != 0))
gotoerr(ENOTCONN);
} else if (addr == 0)
gotoerr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
ENOTCONN : EDESTADDRREQ);
}
if ((atomic && resid > so->so_snd.ssb_hiwat) ||
clen > so->so_snd.ssb_hiwat) {
gotoerr(EMSGSIZE);
}
space = ssb_space(&so->so_snd);
if (flags & MSG_OOB)
space += 1024;
if ((space < 0 || (size_t)space < resid + clen) && uio &&
(atomic || space < so->so_snd.ssb_lowat || space < clen)) {
if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT))
gotoerr(EWOULDBLOCK);
ssb_unlock(&so->so_snd);
error = ssb_wait(&so->so_snd);
if (error)
goto out;
goto restart;
}
mp = ⊤
space -= clen;
do {
if (uio == NULL) {
/*
* Data is prepackaged in "top".
*/
resid = 0;
if (flags & MSG_EOR)
top->m_flags |= M_EOR;
} else do {
if (resid > INT_MAX)
resid = INT_MAX;
m = m_getl((int)resid, MB_WAIT, MT_DATA,
top == NULL ? M_PKTHDR : 0, &mlen);
if (top == NULL) {
m->m_pkthdr.len = 0;
m->m_pkthdr.rcvif = NULL;
}
len = imin((int)szmin(mlen, resid), space);
if (resid < MINCLSIZE) {
/*
* For datagram protocols, leave room
* for protocol headers in first mbuf.
*/
if (atomic && top == 0 && len < mlen)
MH_ALIGN(m, len);
}
space -= len;
error = uiomove(mtod(m, caddr_t), (size_t)len, uio);
resid = uio->uio_resid;
m->m_len = len;
*mp = m;
top->m_pkthdr.len += len;
if (error)
goto release;
mp = &m->m_next;
if (resid == 0) {
if (flags & MSG_EOR)
top->m_flags |= M_EOR;
break;
}
} while (space > 0 && atomic);
if (dontroute)
so->so_options |= SO_DONTROUTE;
if (flags & MSG_OOB) {
pru_flags = PRUS_OOB;
} else if ((flags & MSG_EOF) &&
(so->so_proto->pr_flags & PR_IMPLOPCL) &&
(resid == 0)) {
/*
* If the user set MSG_EOF, the protocol
* understands this flag and nothing left to
* send then use PRU_SEND_EOF instead of PRU_SEND.
*/
pru_flags = PRUS_EOF;
} else if (resid > 0 && space > 0) {
/* If there is more to send, set PRUS_MORETOCOME */
pru_flags = PRUS_MORETOCOME;
} else {
pru_flags = 0;
}
/*
* XXX all the SS_CANTSENDMORE checks previously
* done could be out of date. We could have recieved
* a reset packet in an interrupt or maybe we slept
* while doing page faults in uiomove() etc. We could
* probably recheck again inside the splnet() protection
* here, but there are probably other places that this
* also happens. We must rethink this.
*/
error = so_pru_send(so, pru_flags, top, addr, control, td);
if (dontroute)
so->so_options &= ~SO_DONTROUTE;
clen = 0;
control = 0;
top = NULL;
mp = ⊤
if (error)
goto release;
} while (resid && space > 0);
} while (resid);
release:
ssb_unlock(&so->so_snd);
out:
if (top)
m_freem(top);
if (control)
m_freem(control);
return (error);
}