errno_t
mbuf_find_drvaux(mbuf_t mbuf, u_int32_t *family_p, u_int32_t *subfamily_p,
u_int32_t *length_p, void **data_p)
{
struct m_drvaux_tag *p;
struct m_tag *tag;
if (mbuf == NULL || !(mbuf->m_flags & M_PKTHDR) || data_p == NULL)
return (EINVAL);
*data_p = NULL;
if ((tag = m_tag_locate(mbuf, KERNEL_MODULE_TAG_ID,
KERNEL_TAG_TYPE_DRVAUX, NULL)) == NULL)
return (ENOENT);
/* Must be at least size of m_drvaux_tag */
VERIFY(tag->m_tag_len >= sizeof (*p));
p = (struct m_drvaux_tag *)(tag + 1);
VERIFY(p->da_length > 0 && p->da_length <= MBUF_DRVAUX_MAXLEN);
if (family_p != NULL)
*family_p = p->da_family;
if (subfamily_p != NULL)
*subfamily_p = p->da_subfamily;
if (length_p != NULL)
*length_p = p->da_length;
*data_p = (p + 1);
return (0);
}
/*
* Receive incoming data on our hook. Send it out the socket.
*/
static int
ng_ksocket_rcvdata(hook_p hook, item_p item)
{
struct thread *td = curthread; /* XXX broken */
const node_p node = NG_HOOK_NODE(hook);
const priv_p priv = NG_NODE_PRIVATE(node);
struct socket *const so = priv->so;
struct sockaddr *sa = NULL;
int error;
struct mbuf *m;
struct sa_tag *stag;
/* Extract data */
NGI_GET_M(item, m);
NG_FREE_ITEM(item);
/*
* Look if socket address is stored in packet tags.
* If sockaddr is ours, or provided by a third party (zero id),
* then we accept it.
*/
if (((stag = (struct sa_tag *)m_tag_locate(m, NGM_KSOCKET_COOKIE,
NG_KSOCKET_TAG_SOCKADDR, NULL)) != NULL) &&
(stag->id == NG_NODE_ID(node) || stag->id == 0))
sa = &stag->sa;
/* Reset specific mbuf flags to prevent addressing problems. */
m->m_flags &= ~(M_BCAST|M_MCAST);
/* Send packet */
error = sosend(so, sa, 0, m, 0, 0, td);
return (error);
}
/* Enqueue a packet 'm' to a queue 'q' and add timestamp to that packet.
* Return 1 when unable to add timestamp, otherwise return 0
*/
static int
codel_enqueue(struct fq_codel_flow *q, struct mbuf *m, struct fq_codel_si *si)
{
uint64_t len;
len = m->m_pkthdr.len;
/* finding maximum packet size */
if (len > q->cst.maxpkt_size)
q->cst.maxpkt_size = len;
/* Add timestamp to mbuf as MTAG */
struct m_tag *mtag;
mtag = m_tag_locate(m, MTAG_ABI_COMPAT, DN_AQM_MTAG_TS, NULL);
if (mtag == NULL)
mtag = m_tag_alloc(MTAG_ABI_COMPAT, DN_AQM_MTAG_TS, sizeof(aqm_time_t),
M_NOWAIT);
if (mtag == NULL) {
m_freem(m);
goto drop;
}
*(aqm_time_t *)(mtag + 1) = AQM_UNOW;
m_tag_prepend(m, mtag);
mq_append(&q->mq, m);
fq_update_stats(q, si, len, 0);
return 0;
drop:
fq_update_stats(q, si, len, 1);
m_freem(m);
return 1;
}
int
codel_addq(struct codel *c, class_queue_t *q, struct mbuf *m)
{
struct m_tag *mtag;
uint64_t *enqueue_time;
if (qlen(q) < qlimit(q)) {
mtag = m_tag_locate(m, MTAG_CODEL, 0, NULL);
if (mtag == NULL)
mtag = m_tag_alloc(MTAG_CODEL, 0, sizeof(uint64_t),
M_NOWAIT);
if (mtag == NULL) {
m_freem(m);
return (-1);
}
enqueue_time = (uint64_t *)(mtag + 1);
*enqueue_time = read_machclk();
m_tag_prepend(m, mtag);
_addq(q, m);
return (0);
}
c->drop_overlimit++;
m_freem(m);
return (-1);
}
void
ieee80211_process_callback(struct ieee80211_node* ni, struct mbuf* m,
int status)
{
struct m_tag* mtag;
mtag = m_tag_locate(m, MTAG_ABI_NET80211, NET80211_TAG_CALLBACK, NULL);
if (mtag != NULL) {
struct ieee80211_cb* cb = (struct ieee80211_cb*)(mtag+1);
cb->func(ni, cb->arg, status);
}
}
void
mbuf_del_drvaux(mbuf_t mbuf)
{
struct m_tag *tag;
if (mbuf == NULL || !(mbuf->m_flags & M_PKTHDR))
return;
if ((tag = m_tag_locate(mbuf, KERNEL_MODULE_TAG_ID,
KERNEL_TAG_TYPE_DRVAUX, NULL)) != NULL)
m_tag_delete(mbuf, tag);
}
void *
encap_getarg(struct mbuf *m)
{
struct m_tag *tag;
struct encaptabtag *et;
void *p = NULL;
tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_ENCAP, NULL);
if (tag) {
et = (struct encaptabtag*)(tag + 1);
p = et->arg;
m_tag_delete(m, tag);
}
return p;
}
static int
codel_should_drop(struct codel *c, class_queue_t *q, struct mbuf *m,
u_int64_t now)
{
struct m_tag *mtag;
uint64_t *enqueue_time;
if (m == NULL) {
c->vars.first_above_time = 0;
return (0);
}
mtag = m_tag_locate(m, MTAG_CODEL, 0, NULL);
if (mtag == NULL) {
/* Only one warning per second. */
if (ppsratecheck(&c->last_log, &c->last_pps, 1))
printf("%s: could not found the packet mtag!\n",
__func__);
c->vars.first_above_time = 0;
return (0);
}
enqueue_time = (uint64_t *)(mtag + 1);
c->vars.ldelay = now - *enqueue_time;
c->stats.maxpacket = MAX(c->stats.maxpacket, m_pktlen(m));
if (codel_time_before(c->vars.ldelay, c->params.target) ||
qsize(q) <= c->stats.maxpacket) {
/* went below - stay below for at least interval */
c->vars.first_above_time = 0;
return (0);
}
if (c->vars.first_above_time == 0) {
/* just went above from below. If we stay above
* for at least interval we'll say it's ok to drop
*/
c->vars.first_above_time = now + c->params.interval;
return (0);
}
if (codel_time_after(now, c->vars.first_above_time))
return (1);
return (0);
}
errno_t
mbuf_tag_allocate(
mbuf_t mbuf,
mbuf_tag_id_t id,
mbuf_tag_type_t type,
size_t length,
mbuf_how_t how,
void** data_p)
{
struct m_tag *tag;
u_int32_t mtag_id_first, mtag_id_last;
if (data_p != NULL)
*data_p = NULL;
/* Sanity check parameters */
(void) net_str_id_first_last(&mtag_id_first, &mtag_id_last,
NSI_MBUF_TAG);
if (mbuf == NULL || (mbuf->m_flags & M_PKTHDR) == 0 ||
id < mtag_id_first || id > mtag_id_last || length < 1 ||
(length & 0xffff0000) != 0 || data_p == NULL) {
return (EINVAL);
}
/* Make sure this mtag hasn't already been allocated */
tag = m_tag_locate(mbuf, id, type, NULL);
if (tag != NULL) {
return (EEXIST);
}
/* Allocate an mtag */
tag = m_tag_create(id, type, length, how, mbuf);
if (tag == NULL) {
return (how == M_WAITOK ? ENOMEM : EWOULDBLOCK);
}
/* Attach the mtag and set *data_p */
m_tag_prepend(mbuf, tag);
*data_p = tag + 1;
return (0);
}
void
mbuf_tag_free(
mbuf_t mbuf,
mbuf_tag_id_t id,
mbuf_tag_type_t type)
{
struct m_tag *tag;
u_int32_t mtag_id_first, mtag_id_last;
/* Sanity check parameters */
(void) net_str_id_first_last(&mtag_id_first, &mtag_id_last,
NSI_MBUF_TAG);
if (mbuf == NULL || (mbuf->m_flags & M_PKTHDR) == 0 ||
id < mtag_id_first || id > mtag_id_last)
return;
tag = m_tag_locate(mbuf, id, type, NULL);
if (tag == NULL) {
return;
}
m_tag_delete(mbuf, tag);
}
errno_t
mbuf_add_drvaux(mbuf_t mbuf, mbuf_how_t how, u_int32_t family,
u_int32_t subfamily, size_t length, void **data_p)
{
struct m_drvaux_tag *p;
struct m_tag *tag;
if (mbuf == NULL || !(mbuf->m_flags & M_PKTHDR) ||
length == 0 || length > MBUF_DRVAUX_MAXLEN)
return (EINVAL);
if (data_p != NULL)
*data_p = NULL;
/* Check if one is already associated */
if ((tag = m_tag_locate(mbuf, KERNEL_MODULE_TAG_ID,
KERNEL_TAG_TYPE_DRVAUX, NULL)) != NULL)
return (EEXIST);
/* Tag is (m_drvaux_tag + module specific data) */
if ((tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DRVAUX,
sizeof (*p) + length, how, mbuf)) == NULL)
return ((how == MBUF_WAITOK) ? ENOMEM : EWOULDBLOCK);
p = (struct m_drvaux_tag *)(tag + 1);
p->da_family = family;
p->da_subfamily = subfamily;
p->da_length = length;
/* Associate the tag */
m_tag_prepend(mbuf, tag);
if (data_p != NULL)
*data_p = (p + 1);
return (0);
}
errno_t
mbuf_tag_find(
mbuf_t mbuf,
mbuf_tag_id_t id,
mbuf_tag_type_t type,
size_t *length,
void **data_p)
{
struct m_tag *tag;
u_int32_t mtag_id_first, mtag_id_last;
if (length != NULL)
*length = 0;
if (data_p != NULL)
*data_p = NULL;
/* Sanity check parameters */
(void) net_str_id_first_last(&mtag_id_first, &mtag_id_last,
NSI_MBUF_TAG);
if (mbuf == NULL || (mbuf->m_flags & M_PKTHDR) == 0 ||
id < mtag_id_first || id > mtag_id_last || length == NULL ||
data_p == NULL) {
return (EINVAL);
}
/* Locate an mtag */
tag = m_tag_locate(mbuf, id, type, NULL);
if (tag == NULL) {
return (ENOENT);
}
/* Copy out the pointer to the data and the lenght value */
*length = tag->m_tag_len;
*data_p = tag + 1;
return (0);
}
/*
* IP output. The packet in mbuf chain m contains a skeletal IP
* header (with len, off, ttl, proto, tos, src, dst).
* ip_len and ip_off are in host format.
* The mbuf chain containing the packet will be freed.
* The mbuf opt, if present, will not be freed.
* In the IP forwarding case, the packet will arrive with options already
* inserted, so must have a NULL opt pointer.
*/
int
ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
struct ip_moptions *imo, struct inpcb *inp)
{
struct ip *ip = NULL;
struct ifnet *ifp = NULL; /* keep compiler happy */
struct mbuf *m0;
int hlen = sizeof (struct ip);
int mtu;
int n; /* scratchpad */
int error = 0;
int nortfree = 0;
struct sockaddr_in *dst;
struct in_ifaddr *ia = NULL;
int isbroadcast, sw_csum;
struct route iproute;
struct rtentry *rte; /* cache for ro->ro_rt */
struct in_addr odst;
#ifdef IPFIREWALL_FORWARD
struct m_tag *fwd_tag = NULL;
#endif
#ifdef IPSEC
int no_route_but_check_spd = 0;
#endif
#ifdef PROMISCUOUS_INET
struct ifl2info *l2i_tag = NULL;
int ispromisc = 0;
#endif
M_ASSERTPKTHDR(m);
if (inp != NULL) {
INP_LOCK_ASSERT(inp);
M_SETFIB(m, inp->inp_inc.inc_fibnum);
if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
m->m_pkthdr.flowid = inp->inp_flowid;
m->m_flags |= M_FLOWID;
}
}
#ifdef PROMISCUOUS_INET
l2i_tag = (struct ifl2info *)m_tag_locate(m,
MTAG_PROMISCINET,
MTAG_PROMISCINET_L2INFO,
NULL);
if ((inp && (inp->inp_flags2 & INP_PROMISC)) || l2i_tag) {
unsigned int fib;
if (l2i_tag) {
/*
* This is a packet that has been turned around
* after reception, such as a TCP SYN packet being
* recycled as a RST, so fib comes from the mbuf,
* not the (probably nonexistent) connection
* context.
*/
fib = M_GETFIB(m);
} else {
fib = inp->inp_fibnum;
if (0 != if_promiscinet_add_tag(m, inp->inp_l2info)) {
goto bad;
}
}
ifp = ifnet_byfib_ref(fib);
if (NULL == ifp) {
IPSTAT_INC(ips_noroute);
error = EHOSTUNREACH;
goto bad;
}
isbroadcast = 0;
ispromisc = 1;
}
#endif /* PROMISCUOUS_INET */
if (ro == NULL) {
ro = &iproute;
bzero(ro, sizeof (*ro));
#ifdef FLOWTABLE
{
struct flentry *fle;
/*
* The flow table returns route entries valid for up to 30
* seconds; we rely on the remainder of ip_output() taking no
* longer than that long for the stability of ro_rt. The
* flow ID assignment must have happened before this point.
*/
if ((fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET)) != NULL) {
flow_to_route(fle, ro);
nortfree = 1;
}
}
#endif
}
if (opt) {
int len = 0;
m = ip_insertoptions(m, opt, &len);
if (len != 0)
hlen = len; /* ip->ip_hl is updated above */
}
ip = mtod(m, struct ip *);
/*
* Fill in IP header. If we are not allowing fragmentation,
* then the ip_id field is meaningless, but we don't set it
* to zero. Doing so causes various problems when devices along
* the path (routers, load balancers, firewalls, etc.) illegally
* disable DF on our packet. Note that a 16-bit counter
* will wrap around in less than 10 seconds at 100 Mbit/s on a
* medium with MTU 1500. See Steven M. Bellovin, "A Technique
* for Counting NATted Hosts", Proc. IMW'02, available at
* <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
*/
if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
ip->ip_v = IPVERSION;
ip->ip_hl = hlen >> 2;
ip->ip_id = ip_newid();
IPSTAT_INC(ips_localout);
} else {
static int
firewire_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
struct route *ro)
{
struct fw_com *fc = IFP2FWC(ifp);
int error, type;
struct m_tag *mtag;
union fw_encap *enc;
struct fw_hwaddr *destfw;
uint8_t speed;
uint16_t psize, fsize, dsize;
struct mbuf *mtail;
int unicast, dgl, foff;
static int next_dgl;
#if defined(INET) || defined(INET6)
struct llentry *lle;
#endif
#ifdef MAC
error = mac_ifnet_check_transmit(ifp, m);
if (error)
goto bad;
#endif
if (!((ifp->if_flags & IFF_UP) &&
(ifp->if_drv_flags & IFF_DRV_RUNNING))) {
error = ENETDOWN;
goto bad;
}
/*
* For unicast, we make a tag to store the lladdr of the
* destination. This might not be the first time we have seen
* the packet (for instance, the arp code might be trying to
* re-send it after receiving an arp reply) so we only
* allocate a tag if there isn't one there already. For
* multicast, we will eventually use a different tag to store
* the channel number.
*/
unicast = !(m->m_flags & (M_BCAST | M_MCAST));
if (unicast) {
mtag = m_tag_locate(m, MTAG_FIREWIRE, MTAG_FIREWIRE_HWADDR, NULL);
if (!mtag) {
mtag = m_tag_alloc(MTAG_FIREWIRE, MTAG_FIREWIRE_HWADDR,
sizeof (struct fw_hwaddr), M_NOWAIT);
if (!mtag) {
error = ENOMEM;
goto bad;
}
m_tag_prepend(m, mtag);
}
destfw = (struct fw_hwaddr *)(mtag + 1);
} else {
destfw = 0;
}
switch (dst->sa_family) {
#ifdef INET
case AF_INET:
/*
* Only bother with arp for unicast. Allocation of
* channels etc. for firewire is quite different and
* doesn't fit into the arp model.
*/
if (unicast) {
error = arpresolve(ifp, ro ? ro->ro_rt : NULL, m, dst, (u_char *) destfw, &lle);
if (error)
return (error == EWOULDBLOCK ? 0 : error);
}
type = ETHERTYPE_IP;
break;
case AF_ARP:
{
struct arphdr *ah;
ah = mtod(m, struct arphdr *);
ah->ar_hrd = htons(ARPHRD_IEEE1394);
type = ETHERTYPE_ARP;
if (unicast)
*destfw = *(struct fw_hwaddr *) ar_tha(ah);
/*
* The standard arp code leaves a hole for the target
* hardware address which we need to close up.
*/
bcopy(ar_tpa(ah), ar_tha(ah), ah->ar_pln);
m_adj(m, -ah->ar_hln);
break;
}
#endif
#ifdef INET6
case AF_INET6:
if (unicast) {
error = nd6_storelladdr(fc->fc_ifp, m, dst,
(u_char *) destfw, &lle);
if (error)
return (error);
}
type = ETHERTYPE_IPV6;
break;
#endif
default:
if_printf(ifp, "can't handle af%d\n", dst->sa_family);
error = EAFNOSUPPORT;
goto bad;
}
/*
* Let BPF tap off a copy before we encapsulate.
*/
if (bpf_peers_present(ifp->if_bpf)) {
struct fw_bpfhdr h;
if (unicast)
bcopy(destfw, h.firewire_dhost, 8);
else
bcopy(&firewire_broadcastaddr, h.firewire_dhost, 8);
bcopy(&fc->fc_hwaddr, h.firewire_shost, 8);
h.firewire_type = htons(type);
bpf_mtap2(ifp->if_bpf, &h, sizeof(h), m);
}
/*
* Punt on MCAP for now and send all multicast packets on the
* broadcast channel.
*/
if (m->m_flags & M_MCAST)
m->m_flags |= M_BCAST;
/*
* Figure out what speed to use and what the largest supported
* packet size is. For unicast, this is the minimum of what we
* can speak and what they can hear. For broadcast, lets be
* conservative and use S100. We could possibly improve that
* by examining the bus manager's speed map or similar. We
* also reduce the packet size for broadcast to account for
* the GASP header.
*/
if (unicast) {
speed = min(fc->fc_speed, destfw->sspd);
psize = min(512 << speed, 2 << destfw->sender_max_rec);
} else {
speed = 0;
psize = 512 - 2*sizeof(uint32_t);
}
/*
* Next, we encapsulate, possibly fragmenting the original
* datagram if it won't fit into a single packet.
*/
if (m->m_pkthdr.len <= psize - sizeof(uint32_t)) {
/*
* No fragmentation is necessary.
*/
M_PREPEND(m, sizeof(uint32_t), M_NOWAIT);
if (!m) {
error = ENOBUFS;
goto bad;
}
enc = mtod(m, union fw_encap *);
enc->unfrag.ether_type = type;
enc->unfrag.lf = FW_ENCAP_UNFRAG;
enc->unfrag.reserved = 0;
/*
* Byte swap the encapsulation header manually.
*/
enc->ul[0] = htonl(enc->ul[0]);
error = (ifp->if_transmit)(ifp, m);
return (error);
} else {
/**
* 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);
}
