/**
* Called to deliver a frame to either the host, the wire or both.
*/
int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
{
NOREF(pvIfData);
void (*input_f)(struct ifnet *, struct mbuf *);
struct ifnet *ifp;
struct mbuf *m;
struct m_tag *mtag;
bool fActive;
int error;
ifp = ASMAtomicUoReadPtrT(&pThis->u.s.ifp, struct ifnet *);
VBOXCURVNET_SET(ifp->if_vnet);
if (fDst & INTNETTRUNKDIR_WIRE)
{
m = vboxNetFltFreeBSDSGMBufFromSG(pThis, pSG);
if (m == NULL)
return VERR_NO_MEMORY;
m = m_pullup(m, ETHER_HDR_LEN);
if (m == NULL)
return VERR_NO_MEMORY;
m->m_flags |= M_PKTHDR;
ether_output_frame(ifp, m);
}
if (fDst & INTNETTRUNKDIR_HOST)
{
m = vboxNetFltFreeBSDSGMBufFromSG(pThis, pSG);
if (m == NULL)
return VERR_NO_MEMORY;
m = m_pullup(m, ETHER_HDR_LEN);
if (m == NULL)
return VERR_NO_MEMORY;
/*
* Delivering packets to the host will be captured by the
* input hook. Tag the packet with a mbuf tag so that we
* can skip re-delivery of the packet to the guest during
* input hook processing.
*/
mtag = m_tag_alloc(MTAG_VBOX, PACKET_TAG_VBOX, 0, M_NOWAIT);
if (mtag == NULL)
{
m_freem(m);
return VERR_NO_MEMORY;
}
m_tag_init(m);
m_tag_prepend(m, mtag);
m->m_flags |= M_PKTHDR;
m->m_pkthdr.rcvif = ifp;
ifp->if_input(ifp, m);
}
VBOXCURVNET_RESTORE();
return VINF_SUCCESS;
}
/**
* Output processing task, handles outgoing frames
*/
static void vboxNetFltFreeBSDoutput(void *arg, int pending)
{
PVBOXNETFLTINS pThis = (PVBOXNETFLTINS)arg;
struct mbuf *m, *m0;
struct ifnet *ifp = pThis->u.s.ifp;
unsigned int cSegs = 0;
bool fDropIt = false, fActive;
PINTNETSG pSG;
VBOXCURVNET_SET(ifp->if_vnet);
vboxNetFltRetain(pThis, true /* fBusy */);
for (;;)
{
mtx_lock_spin(&pThis->u.s.outq.ifq_mtx);
_IF_DEQUEUE(&pThis->u.s.outq, m);
mtx_unlock_spin(&pThis->u.s.outq.ifq_mtx);
if (m == NULL)
break;
for (m0 = m; m0 != NULL; m0 = m0->m_next)
if (m0->m_len > 0)
cSegs++;
#ifdef PADD_RUNT_FRAMES_FROM_HOST
if (m_length(m, NULL) < 60)
cSegs++;
#endif
/* Create a copy and deliver to the virtual switch */
pSG = RTMemTmpAlloc(RT_OFFSETOF(INTNETSG, aSegs[cSegs]));
vboxNetFltFreeBSDMBufToSG(pThis, m, pSG, cSegs, 0);
fDropIt = pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, INTNETTRUNKDIR_HOST);
RTMemTmpFree(pSG);
if (fDropIt)
m_freem(m);
else
ether_output_frame(ifp, m);
}
vboxNetFltRelease(pThis, true /* fBusy */);
VBOXCURVNET_RESTORE();
}
/*
* Ethernet output routine.
* Encapsulate a packet of type family for the local net.
* Use trailer local net encapsulation if enough data in first
* packet leaves a multiple of 512 bytes of data in remainder.
*/
int
ether_output(struct ifnet *ifp, struct mbuf *m,
const struct sockaddr *dst, struct route *ro)
{
short type;
int error = 0, hdrcmplt = 0;
u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
struct llentry *lle = NULL;
struct rtentry *rt0 = NULL;
struct ether_header *eh;
struct pf_mtag *t;
int loop_copy = 1;
int hlen; /* link layer header length */
if (ro != NULL) {
if (!(m->m_flags & (M_BCAST | M_MCAST)))
lle = ro->ro_lle;
rt0 = ro->ro_rt;
}
#ifdef MAC
error = mac_ifnet_check_transmit(ifp, m);
if (error)
senderr(error);
#endif
M_PROFILE(m);
if (ifp->if_flags & IFF_MONITOR)
senderr(ENETDOWN);
if (!((ifp->if_flags & IFF_UP) &&
(ifp->if_drv_flags & IFF_DRV_RUNNING)))
senderr(ENETDOWN);
hlen = ETHER_HDR_LEN;
switch (dst->sa_family) {
#ifdef INET
case AF_INET:
if (lle != NULL && (lle->la_flags & LLE_VALID))
memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
else
error = arpresolve(ifp, rt0, m, dst, edst, &lle);
if (error)
return (error == EWOULDBLOCK ? 0 : error);
type = htons(ETHERTYPE_IP);
break;
case AF_ARP:
{
struct arphdr *ah;
ah = mtod(m, struct arphdr *);
ah->ar_hrd = htons(ARPHRD_ETHER);
loop_copy = 0; /* if this is for us, don't do it */
switch(ntohs(ah->ar_op)) {
case ARPOP_REVREQUEST:
case ARPOP_REVREPLY:
type = htons(ETHERTYPE_REVARP);
break;
case ARPOP_REQUEST:
case ARPOP_REPLY:
default:
type = htons(ETHERTYPE_ARP);
break;
}
if (m->m_flags & M_BCAST)
bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
else
bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
}
break;
#endif
#ifdef INET6
case AF_INET6:
if (lle != NULL && (lle->la_flags & LLE_VALID))
memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
else
error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle);
if (error)
return error;
type = htons(ETHERTYPE_IPV6);
break;
#endif
#ifdef IPX
case AF_IPX:
if (ef_outputp) {
error = ef_outputp(ifp, &m, dst, &type, &hlen);
if (error)
goto bad;
} else
type = htons(ETHERTYPE_IPX);
bcopy(&((const struct sockaddr_ipx *)dst)->sipx_addr.x_host,
edst, sizeof (edst));
break;
#endif
#ifdef NETATALK
case AF_APPLETALK:
{
struct at_ifaddr *aa;
if ((aa = at_ifawithnet((const struct sockaddr_at *)dst)) == NULL)
senderr(EHOSTUNREACH); /* XXX */
if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
ifa_free(&aa->aa_ifa);
return (0);
}
/*
* In the phase 2 case, need to prepend an mbuf for the llc header.
*/
if ( aa->aa_flags & AFA_PHASE2 ) {
struct llc llc;
ifa_free(&aa->aa_ifa);
M_PREPEND(m, LLC_SNAPFRAMELEN, M_NOWAIT);
if (m == NULL)
senderr(ENOBUFS);
llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
llc.llc_control = LLC_UI;
bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
type = htons(m->m_pkthdr.len);
hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
} else {
ifa_free(&aa->aa_ifa);
type = htons(ETHERTYPE_AT);
}
break;
}
#endif /* NETATALK */
case pseudo_AF_HDRCMPLT:
{
const struct ether_header *eh;
hdrcmplt = 1;
eh = (const struct ether_header *)dst->sa_data;
(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
/* FALLTHROUGH */
case AF_UNSPEC:
loop_copy = 0; /* if this is for us, don't do it */
eh = (const struct ether_header *)dst->sa_data;
(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
type = eh->ether_type;
break;
}
#ifdef MPLS
case AF_MPLS:
if (lle != NULL && (lle->la_flags & LLE_VALID))
bcopy(&lle->ll_addr.mac16 , edst, sizeof(edst));
else
error = mpls_arpresolve(ifp, rt0, m, dst, edst, &lle);
if (error)
return (error == EWOULDBLOCK ? 0 : error);
if (m->m_flags & (M_BCAST | M_MCAST))
type = htons(ETHERTYPE_MPLS_MCAST);
else
type = htons(ETHERTYPE_MPLS);
break;
#endif /* MPLS */
default:
if_printf(ifp, "can't handle af%d\n", dst->sa_family);
senderr(EAFNOSUPPORT);
}
if (lle != NULL && (lle->la_flags & LLE_IFADDR)) {
update_mbuf_csumflags(m, m);
return (if_simloop(ifp, m, dst->sa_family, 0));
}
/*
* Add local net header. If no space in first mbuf,
* allocate another.
*/
M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
if (m == NULL)
senderr(ENOBUFS);
eh = mtod(m, struct ether_header *);
(void)memcpy(&eh->ether_type, &type,
sizeof(eh->ether_type));
(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
if (hdrcmplt)
(void)memcpy(eh->ether_shost, esrc,
sizeof(eh->ether_shost));
else
(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
sizeof(eh->ether_shost));
/*
* If a simplex interface, and the packet is being sent to our
* Ethernet address or a broadcast address, loopback a copy.
* XXX To make a simplex device behave exactly like a duplex
* device, we should copy in the case of sending to our own
* ethernet address (thus letting the original actually appear
* on the wire). However, we don't do that here for security
* reasons and compatibility with the original behavior.
*/
if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
((t = pf_find_mtag(m)) == NULL || !t->routed)) {
if (m->m_flags & M_BCAST) {
struct mbuf *n;
/*
* Because if_simloop() modifies the packet, we need a
* writable copy through m_dup() instead of a readonly
* one as m_copy[m] would give us. The alternative would
* be to modify if_simloop() to handle the readonly mbuf,
* but performancewise it is mostly equivalent (trading
* extra data copying vs. extra locking).
*
* XXX This is a local workaround. A number of less
* often used kernel parts suffer from the same bug.
* See PR kern/105943 for a proposed general solution.
*/
if ((n = m_dup(m, M_NOWAIT)) != NULL) {
update_mbuf_csumflags(m, n);
(void)if_simloop(ifp, n, dst->sa_family, hlen);
} else
ifp->if_iqdrops++;
} else if (bcmp(eh->ether_dhost, eh->ether_shost,
ETHER_ADDR_LEN) == 0) {
update_mbuf_csumflags(m, m);
(void) if_simloop(ifp, m, dst->sa_family, hlen);
return (0); /* XXX */
}
}
/*
* Bridges require special output handling.
*/
if (ifp->if_bridge) {
BRIDGE_OUTPUT(ifp, m, error);
return (error);
}
#if defined(INET) || defined(INET6)
if (ifp->if_carp &&
(error = (*carp_output_p)(ifp, m, dst)))
goto bad;
#endif
/* Handle ng_ether(4) processing, if any */
if (IFP2AC(ifp)->ac_netgraph != NULL) {
KASSERT(ng_ether_output_p != NULL,
("ng_ether_output_p is NULL"));
if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
bad: if (m != NULL)
m_freem(m);
return (error);
}
if (m == NULL)
return (0);
}
/* Continue with link-layer output */
return ether_output_frame(ifp, m);
}
/*
* Ethernet output routine.
* Encapsulate a packet of type family for the local net.
* Use trailer local net encapsulation if enough data in first
* packet leaves a multiple of 512 bytes of data in remainder.
*/
int
ether_output(struct ifnet *ifp, struct mbuf *m,
const struct sockaddr *dst, struct route *ro)
{
short type;
int error = 0, hdrcmplt = 0;
u_char edst[ETHER_ADDR_LEN];
struct llentry *lle = NULL;
struct rtentry *rt0 = NULL;
struct ether_header *eh;
struct pf_mtag *t;
int loop_copy = 1;
int hlen; /* link layer header length */
int is_gw = 0;
uint32_t pflags = 0;
if (ro != NULL) {
if (!(m->m_flags & (M_BCAST | M_MCAST))) {
lle = ro->ro_lle;
if (lle != NULL)
pflags = lle->la_flags;
}
rt0 = ro->ro_rt;
if (rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) != 0)
is_gw = 1;
}
#ifdef MAC
error = mac_ifnet_check_transmit(ifp, m);
if (error)
senderr(error);
#endif
M_PROFILE(m);
if (ifp->if_flags & IFF_MONITOR)
senderr(ENETDOWN);
if (!((ifp->if_flags & IFF_UP) &&
(ifp->if_drv_flags & IFF_DRV_RUNNING)))
senderr(ENETDOWN);
hlen = ETHER_HDR_LEN;
switch (dst->sa_family) {
#ifdef INET
case AF_INET:
if (lle != NULL && (pflags & LLE_VALID) != 0)
memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
else
error = arpresolve(ifp, is_gw, m, dst, edst, &pflags);
if (error)
return (error == EWOULDBLOCK ? 0 : error);
type = htons(ETHERTYPE_IP);
break;
case AF_ARP:
{
struct arphdr *ah;
ah = mtod(m, struct arphdr *);
ah->ar_hrd = htons(ARPHRD_ETHER);
loop_copy = 0; /* if this is for us, don't do it */
switch(ntohs(ah->ar_op)) {
case ARPOP_REVREQUEST:
case ARPOP_REVREPLY:
type = htons(ETHERTYPE_REVARP);
break;
case ARPOP_REQUEST:
case ARPOP_REPLY:
default:
type = htons(ETHERTYPE_ARP);
break;
}
if (m->m_flags & M_BCAST)
bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
else
bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
}
break;
#endif
#ifdef INET6
case AF_INET6:
if (lle != NULL && (pflags & LLE_VALID))
memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
else
error = nd6_resolve(ifp, is_gw, m, dst, (u_char *)edst,
&pflags);
if (error)
return (error == EWOULDBLOCK ? 0 : error);
type = htons(ETHERTYPE_IPV6);
break;
#endif
case pseudo_AF_HDRCMPLT:
{
const struct ether_header *eh;
hdrcmplt = 1;
/* FALLTHROUGH */
case AF_UNSPEC:
loop_copy = 0; /* if this is for us, don't do it */
eh = (const struct ether_header *)dst->sa_data;
(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
type = eh->ether_type;
break;
}
default:
if_printf(ifp, "can't handle af%d\n", dst->sa_family);
senderr(EAFNOSUPPORT);
}
if ((pflags & LLE_IFADDR) != 0) {
update_mbuf_csumflags(m, m);
return (if_simloop(ifp, m, dst->sa_family, 0));
}
/*
* Add local net header. If no space in first mbuf,
* allocate another.
*/
M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
if (m == NULL)
senderr(ENOBUFS);
eh = mtod(m, struct ether_header *);
if (hdrcmplt == 0) {
memcpy(&eh->ether_type, &type, sizeof(eh->ether_type));
memcpy(eh->ether_dhost, edst, sizeof (edst));
memcpy(eh->ether_shost, IF_LLADDR(ifp),sizeof(eh->ether_shost));
}
/*
* If a simplex interface, and the packet is being sent to our
* Ethernet address or a broadcast address, loopback a copy.
* XXX To make a simplex device behave exactly like a duplex
* device, we should copy in the case of sending to our own
* ethernet address (thus letting the original actually appear
* on the wire). However, we don't do that here for security
* reasons and compatibility with the original behavior.
*/
if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
((t = pf_find_mtag(m)) == NULL || !t->routed)) {
if (m->m_flags & M_BCAST) {
struct mbuf *n;
/*
* Because if_simloop() modifies the packet, we need a
* writable copy through m_dup() instead of a readonly
* one as m_copy[m] would give us. The alternative would
* be to modify if_simloop() to handle the readonly mbuf,
* but performancewise it is mostly equivalent (trading
* extra data copying vs. extra locking).
*
* XXX This is a local workaround. A number of less
* often used kernel parts suffer from the same bug.
* See PR kern/105943 for a proposed general solution.
*/
if ((n = m_dup(m, M_NOWAIT)) != NULL) {
update_mbuf_csumflags(m, n);
(void)if_simloop(ifp, n, dst->sa_family, hlen);
} else
if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
} else if (bcmp(eh->ether_dhost, eh->ether_shost,
ETHER_ADDR_LEN) == 0) {
update_mbuf_csumflags(m, m);
(void) if_simloop(ifp, m, dst->sa_family, hlen);
return (0); /* XXX */
}
}
/*
* Bridges require special output handling.
*/
if (ifp->if_bridge) {
BRIDGE_OUTPUT(ifp, m, error);
return (error);
}
#if defined(INET) || defined(INET6)
if (ifp->if_carp &&
(error = (*carp_output_p)(ifp, m, dst)))
goto bad;
#endif
/* Handle ng_ether(4) processing, if any */
if (ifp->if_l2com != NULL) {
KASSERT(ng_ether_output_p != NULL,
("ng_ether_output_p is NULL"));
if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
bad: if (m != NULL)
m_freem(m);
return (error);
}
if (m == NULL)
return (0);
}
/* Continue with link-layer output */
return ether_output_frame(ifp, m);
}
/**
* Handle data on netgraph hooks.
* Frames processing is deferred to a taskqueue because this might
* be called with non-sleepable locks held and code paths inside
* the virtual switch might sleep.
*/
static int ng_vboxnetflt_rcvdata(hook_p hook, item_p item)
{
const node_p node = NG_HOOK_NODE(hook);
PVBOXNETFLTINS pThis = NG_NODE_PRIVATE(node);
struct ifnet *ifp = pThis->u.s.ifp;
struct mbuf *m;
struct m_tag *mtag;
bool fActive;
VBOXCURVNET_SET(ifp->if_vnet);
fActive = vboxNetFltTryRetainBusyActive(pThis);
NGI_GET_M(item, m);
NG_FREE_ITEM(item);
/* Locate tag to see if processing should be skipped for this frame */
mtag = m_tag_locate(m, MTAG_VBOX, PACKET_TAG_VBOX, NULL);
if (mtag != NULL)
{
m_tag_unlink(m, mtag);
m_tag_free(mtag);
}
/*
* Handle incoming hook. This is connected to the
* input path of the interface, thus handling incoming frames.
*/
if (pThis->u.s.input == hook)
{
if (mtag != NULL || !fActive)
{
ether_demux(ifp, m);
if (fActive)
vboxNetFltRelease(pThis, true /*fBusy*/);
VBOXCURVNET_RESTORE();
return (0);
}
mtx_lock_spin(&pThis->u.s.inq.ifq_mtx);
_IF_ENQUEUE(&pThis->u.s.inq, m);
mtx_unlock_spin(&pThis->u.s.inq.ifq_mtx);
taskqueue_enqueue_fast(taskqueue_fast, &pThis->u.s.tskin);
}
/*
* Handle mbufs on the outgoing hook, frames going to the interface
*/
else if (pThis->u.s.output == hook)
{
if (mtag != NULL || !fActive)
{
int rc = ether_output_frame(ifp, m);
if (fActive)
vboxNetFltRelease(pThis, true /*fBusy*/);
VBOXCURVNET_RESTORE();
return rc;
}
mtx_lock_spin(&pThis->u.s.outq.ifq_mtx);
_IF_ENQUEUE(&pThis->u.s.outq, m);
mtx_unlock_spin(&pThis->u.s.outq.ifq_mtx);
taskqueue_enqueue_fast(taskqueue_fast, &pThis->u.s.tskout);
}
else
{
m_freem(m);
}
if (fActive)
vboxNetFltRelease(pThis, true /*fBusy*/);
VBOXCURVNET_RESTORE();
return (0);
}
/*
* forward a chain of packets to the proper destination.
* This runs outside the dummynet lock.
*/
static void
dummynet_send(struct mbuf *m)
{
struct mbuf *n;
for (; m != NULL; m = n) {
struct ifnet *ifp = NULL; /* gcc 3.4.6 complains */
struct m_tag *tag;
int dst;
n = m->m_nextpkt;
m->m_nextpkt = NULL;
tag = m_tag_first(m);
if (tag == NULL) { /* should not happen */
dst = DIR_DROP;
} else {
struct dn_pkt_tag *pkt = dn_tag_get(m);
/* extract the dummynet info, rename the tag
* to carry reinject info.
*/
dst = pkt->dn_dir;
ifp = pkt->ifp;
tag->m_tag_cookie = MTAG_IPFW_RULE;
tag->m_tag_id = 0;
}
switch (dst) {
case DIR_OUT:
ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
break ;
case DIR_IN :
netisr_dispatch(NETISR_IP, m);
break;
#ifdef INET6
case DIR_IN | PROTO_IPV6:
netisr_dispatch(NETISR_IPV6, m);
break;
case DIR_OUT | PROTO_IPV6:
ip6_output(m, NULL, NULL, IPV6_FORWARDING, NULL, NULL, NULL);
break;
#endif
case DIR_FWD | PROTO_IFB: /* DN_TO_IFB_FWD: */
if (bridge_dn_p != NULL)
((*bridge_dn_p)(m, ifp));
else
printf("dummynet: if_bridge not loaded\n");
break;
case DIR_IN | PROTO_LAYER2: /* DN_TO_ETH_DEMUX: */
/*
* The Ethernet code assumes the Ethernet header is
* contiguous in the first mbuf header.
* Insure this is true.
*/
if (m->m_len < ETHER_HDR_LEN &&
(m = m_pullup(m, ETHER_HDR_LEN)) == NULL) {
printf("dummynet/ether: pullup failed, "
"dropping packet\n");
break;
}
ether_demux(m->m_pkthdr.rcvif, m);
break;
case DIR_OUT | PROTO_LAYER2: /* N_TO_ETH_OUT: */
ether_output_frame(ifp, m);
break;
case DIR_DROP:
/* drop the packet after some time */
FREE_PKT(m);
break;
default:
printf("dummynet: bad switch %d!\n", dst);
FREE_PKT(m);
break;
}
}
}
