#if !INET
#error "IPDIVERT requires INET."
#endif
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/socket.h>
#include <sys/domain.h>
#include <sys/protosw.h>
#include <sys/socketvar.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <machine/endian.h>
#include <net/if.h>
#include <net/route.h>
#include <net/kpi_protocol.h>
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/in_pcb.h>
#include <netinet/in_var.h>
#include <netinet/ip_var.h>
#include <netinet/ip_fw.h>
#include <netinet/ip_divert.h>
#include <kern/zalloc.h>
#include <libkern/OSAtomic.h>
#define DIVSNDQ (65536 + 100)
#define DIVRCVQ (65536 + 100)
static struct inpcbhead divcb;
static struct inpcbinfo divcbinfo;
static u_int32_t div_sendspace = DIVSNDQ;
static u_int32_t div_recvspace = DIVRCVQ;
static struct sockaddr_in divsrc = {
.sin_len = sizeof(divsrc),
.sin_family = AF_INET,
.sin_port = 0,
.sin_addr = { .s_addr = 0 },
.sin_zero = { 0, 0, 0, 0, 0, 0, 0, 0 }
};
static int div_output(struct socket *so,
struct mbuf *m, struct sockaddr_in *addr, struct mbuf *control);
extern int load_ipfw(void);
void
div_init(struct protosw *pp, struct domain *dp)
{
#pragma unused(dp)
static int div_initialized = 0;
struct inpcbinfo *pcbinfo;
VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
if (div_initialized) {
return;
}
div_initialized = 1;
LIST_INIT(&divcb);
divcbinfo.ipi_listhead = &divcb;
divcbinfo.ipi_hashbase = hashinit(1, M_PCB, &divcbinfo.ipi_hashmask);
divcbinfo.ipi_porthashbase = hashinit(1, M_PCB, &divcbinfo.ipi_porthashmask);
divcbinfo.ipi_zone = zinit(sizeof(struct inpcb), (512 * sizeof(struct inpcb)),
4096, "divzone");
pcbinfo = &divcbinfo;
pcbinfo->ipi_lock_grp_attr = lck_grp_attr_alloc_init();
pcbinfo->ipi_lock_grp = lck_grp_alloc_init("divcb", pcbinfo->ipi_lock_grp_attr);
pcbinfo->ipi_lock_attr = lck_attr_alloc_init();
if ((pcbinfo->ipi_lock = lck_rw_alloc_init(pcbinfo->ipi_lock_grp,
pcbinfo->ipi_lock_attr)) == NULL) {
panic("%s: unable to allocate PCB lock\n", __func__);
}
in_pcbinfo_attach(&divcbinfo);
#if IPFIREWALL
if (!IPFW_LOADED) {
load_ipfw();
}
#endif
}
void
div_input(struct mbuf *m, __unused int off)
{
OSAddAtomic(1, &ipstat.ips_noproto);
m_freem(m);
}
void
divert_packet(struct mbuf *m, int incoming, int port, int rule)
{
struct ip *ip;
struct inpcb *inp;
struct socket *sa;
u_int16_t nport;
KASSERT(port != 0, ("%s: port=0", __FUNCTION__));
divsrc.sin_port = rule;
if (m->m_len < sizeof(struct ip) &&
(m = m_pullup(m, sizeof(struct ip))) == 0) {
return;
}
ip = mtod(m, struct ip *);
divsrc.sin_addr.s_addr = 0;
if (incoming) {
struct ifaddr *ifa;
KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__));
ifnet_lock_shared(m->m_pkthdr.rcvif);
TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
IFA_LOCK(ifa);
if (ifa->ifa_addr->sa_family != AF_INET) {
IFA_UNLOCK(ifa);
continue;
}
divsrc.sin_addr =
((struct sockaddr_in *)(void *) ifa->ifa_addr)->sin_addr;
IFA_UNLOCK(ifa);
break;
}
ifnet_lock_done(m->m_pkthdr.rcvif);
}
bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero));
if (m->m_pkthdr.rcvif) {
snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero),
"%s", if_name(m->m_pkthdr.rcvif));
}
sa = NULL;
nport = htons((u_int16_t)port);
lck_rw_lock_shared(divcbinfo.ipi_lock);
LIST_FOREACH(inp, &divcb, inp_list) {
if (inp->inp_lport == nport) {
sa = inp->inp_socket;
}
}
if (sa) {
int error = 0;
socket_lock(sa, 1);
if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc,
m, (struct mbuf *)0, &error) != 0) {
sorwakeup(sa);
}
socket_unlock(sa, 1);
} else {
m_freem(m);
OSAddAtomic(1, &ipstat.ips_noproto);
OSAddAtomic(-1, &ipstat.ips_delivered);
}
lck_rw_done(divcbinfo.ipi_lock);
}
static int
div_output(struct socket *so, struct mbuf *m, struct sockaddr_in *sin,
struct mbuf *control)
{
struct inpcb *const inp = sotoinpcb(so);
struct ip *const ip = mtod(m, struct ip *);
int error = 0;
int sotc = SO_TC_UNSPEC;
if (control != NULL) {
int ignored;
(void) so_tc_from_control(contro, &sotc, &ignored);
m_freem(control);
control = NULL;
}
if (sotc == SO_TC_UNSPEC) {
sotc = so->so_traffic_class;
}
if (sin) {
struct m_tag *mtag;
struct divert_tag *dt;
int len = 0;
char *c = sin->sin_zero;
mtag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DIVERT,
sizeof(struct divert_tag), M_NOWAIT, m);
if (mtag == NULL) {
error = ENOBUFS;
goto cantsend;
}
dt = (struct divert_tag *)(mtag + 1);
dt->info = 0;
dt->cookie = sin->sin_port;
m_tag_prepend(m, mtag);
while (*c++ && (len++ < sizeof(sin->sin_zero))) {
;
}
if ((len > 0) && (len < sizeof(sin->sin_zero))) {
m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
}
}
if (!sin || sin->sin_addr.s_addr == 0) {
struct ip_out_args ipoa;
struct route ro;
struct ip_moptions *imo;
bzero(&ipoa, sizeof(ipoa));
ipoa.ipoa_boundif = IFSCOPE_NONE;
ipoa.ipoa_flags = IPOAF_SELECT_SRCIF;
ipoa.ipoa_sotc = SO_TC_UNSPEC;
ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
if (((ip->ip_hl != (sizeof(*ip) >> 2)) && inp->inp_options) ||
((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
error = EINVAL;
goto cantsend;
}
#if BYTE_ORDER != BIG_ENDIAN
NTOHS(ip->ip_len);
NTOHS(ip->ip_off);
#endif
OSAddAtomic(1, &ipstat.ips_rawout);
inp_route_copyout(inp, &ro);
if (sotc != SO_TC_UNSPEC) {
ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
ipoa.ipoa_sotc = sotc;
}
set_packet_service_class(m, so, sotc, 0);
imo = inp->inp_moptions;
if (imo != NULL) {
IMO_ADDREF(imo);
}
socket_unlock(so, 0);
#if CONFIG_MACF_NET
mac_mbuf_label_associate_inpcb(inp, m);
#endif
error = ip_output(m, inp->inp_options, &ro,
(so->so_options & SO_DONTROUTE) |
IP_ALLOWBROADCAST | IP_RAWOUTPUT | IP_OUTARGS,
imo, &ipoa);
socket_lock(so, 0);
if (imo != NULL) {
IMO_REMREF(imo);
}
inp_route_copyin(inp, &ro);
} else {
struct ifaddr *ifa;
if (m->m_pkthdr.rcvif == NULL) {
struct sockaddr_in _sin;
bzero(&_sin, sizeof(_sin));
_sin.sin_family = AF_INET;
_sin.sin_len = sizeof(struct sockaddr_in);
_sin.sin_addr.s_addr = sin->sin_addr.s_addr;
if (!(ifa = ifa_ifwithaddr(SA(&_sin)))) {
error = EADDRNOTAVAIL;
goto cantsend;
}
m->m_pkthdr.rcvif = ifa->ifa_ifp;
IFA_REMREF(ifa);
}
#if CONFIG_MACF_NET
mac_mbuf_label_associate_socket(so, m);
#endif
proto_inject(PF_INET, m);
}
return error;
cantsend:
m_freem(m);
return error;
}
static int
div_attach(struct socket *so, int proto, struct proc *p)
{
struct inpcb *inp;
int error;
inp = sotoinpcb(so);
if (inp) {
panic("div_attach");
}
if ((error = proc_suser(p)) != 0) {
return error;
}
error = soreserve(so, div_sendspace, div_recvspace);
if (error) {
return error;
}
error = in_pcballoc(so, &divcbinfo, p);
if (error) {
return error;
}
inp = (struct inpcb *)so->so_pcb;
inp->inp_ip_p = proto;
inp->inp_vflag |= INP_IPV4;
inp->inp_flags |= INP_HDRINCL;
so->so_state |= SS_ISCONNECTED;
#ifdef MORE_DICVLOCK_DEBUG
printf("div_attach: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x\n",
(uint64_t)VM_KERNEL_ADDRPERM(so),
(uint64_t)VM_KERNEL_ADDRPERM(so->so_pcb),
(uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)),
so->so_usecount);
#endif
return 0;
}
static int
div_detach(struct socket *so)
{
struct inpcb *inp;
#ifdef MORE_DICVLOCK_DEBUG
printf("div_detach: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x\n",
(uint64_t)VM_KERNEL_ADDRPERM(so),
(uint64_t)VM_KERNEL_ADDRPERM(so->so_pcb),
(uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)),
so->so_usecount);
#endif
inp = sotoinpcb(so);
if (inp == 0) {
panic("div_detach: so=%p null inp\n", so);
}
in_pcbdetach(inp);
inp->inp_state = INPCB_STATE_DEAD;
return 0;
}
static int
div_abort(struct socket *so)
{
soisdisconnected(so);
return div_detach(so);
}
static int
div_disconnect(struct socket *so)
{
if ((so->so_state & SS_ISCONNECTED) == 0) {
return ENOTCONN;
}
return div_abort(so);
}
static int
div_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
{
struct inpcb *inp;
int error;
inp = sotoinpcb(so);
if (nam->sa_family != AF_INET) {
error = EAFNOSUPPORT;
} else {
((struct sockaddr_in *)(void *)nam)->sin_addr.s_addr = INADDR_ANY;
error = in_pcbbind(inp, nam, p);
}
return error;
}
static int
div_shutdown(struct socket *so)
{
socantsendmore(so);
return 0;
}
static int
div_send(struct socket *so, __unused int flags, struct mbuf *m, struct sockaddr *nam,
struct mbuf *control, __unused struct proc *p)
{
if (m->m_len < sizeof(struct ip) &&
(m = m_pullup(m, sizeof(struct ip))) == 0) {
OSAddAtomic(1, &ipstat.ips_toosmall);
m_freem(m);
return EINVAL;
}
return div_output(so, m, SIN(nam), control);
}
#if 0
static int
div_pcblist SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
int error, i, n;
struct inpcb *inp, **inp_list;
inp_gen_t gencnt;
struct xinpgen xig;
lck_rw_lock_exclusive(divcbinfo.ipi_lock);
if (req->oldptr == USER_ADDR_NULL) {
n = divcbinfo.ipi_count;
req->oldidx = 2 * (sizeof xig)
+ (n + n / 8) * sizeof(struct xinpcb);
lck_rw_done(divcbinfo.ipi_lock);
return 0;
}
if (req->newptr != USER_ADDR_NULL) {
lck_rw_done(divcbinfo.ipi_lock);
return EPERM;
}
gencnt = divcbinfo.ipi_gencnt;
n = divcbinfo.ipi_count;
bzero(&xig, sizeof(xig));
xig.xig_len = sizeof xig;
xig.xig_count = n;
xig.xig_gen = gencnt;
xig.xig_sogen = so_gencnt;
error = SYSCTL_OUT(req, &xig, sizeof xig);
if (error) {
lck_rw_done(divcbinfo.ipi_lock);
return error;
}
inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
if (inp_list == 0) {
lck_rw_done(divcbinfo.ipi_lock);
return ENOMEM;
}
for (inp = LIST_FIRST(divcbinfo.ipi_listhead), i = 0; inp && i < n;
inp = LIST_NEXT(inp, inp_list)) {
#ifdef __APPLE__
if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
#else
if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp))
#endif
{ inp_list[i++] = inp;}
}
n = i;
error = 0;
for (i = 0; i < n; i++) {
inp = inp_list[i];
if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
struct xinpcb xi;
bzero(&xi, sizeof(xi));
xi.xi_len = sizeof xi;
inpcb_to_compat(inp, &xi.xi_inp);
if (inp->inp_socket) {
sotoxsocket(inp->inp_socket, &xi.xi_socket);
}
error = SYSCTL_OUT(req, &xi, sizeof xi);
}
}
if (!error) {
bzero(&xig, sizeof(xig));
xig.xig_len = sizeof xig;
xig.xig_gen = divcbinfo.ipi_gencnt;
xig.xig_sogen = so_gencnt;
xig.xig_count = divcbinfo.ipi_count;
error = SYSCTL_OUT(req, &xig, sizeof xig);
}
FREE(inp_list, M_TEMP);
lck_rw_done(divcbinfo.ipi_lock);
return error;
}
#endif
__private_extern__ int
div_lock(struct socket *so, int refcount, void *lr)
{
void *lr_saved;
if (lr == NULL) {
lr_saved = __builtin_return_address(0);
} else {
lr_saved = lr;
}
#ifdef MORE_DICVLOCK_DEBUG
printf("div_lock: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x "
"lr=0x%llx\n", (uint64_t)VM_KERNEL_ADDRPERM(so),
(uint64_t)VM_KERNEL_ADDRPERM(so->so_pcb), so->so_pcb ?
(uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)) : NULL,
so->so_usecount, (uint64_t)VM_KERNEL_ADDRPERM(lr_saved));
#endif
if (so->so_pcb) {
lck_mtx_lock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
} else {
panic("div_lock: so=%p NO PCB! lr=%p lrh= lrh= %s\n",
so, lr_saved, solockhistory_nr(so));
}
if (so->so_usecount < 0) {
panic("div_lock: so=%p so_pcb=%p lr=%p ref=%x lrh= %s\n",
so, so->so_pcb, lr_saved, so->so_usecount,
solockhistory_nr(so));
}
if (refcount) {
so->so_usecount++;
}
so->lock_lr[so->next_lock_lr] = lr_saved;
so->next_lock_lr = (so->next_lock_lr + 1) % SO_LCKDBG_MAX;
return 0;
}
__private_extern__ int
div_unlock(struct socket *so, int refcount, void *lr)
{
void *lr_saved;
lck_mtx_t * mutex_held;
struct inpcb *inp = sotoinpcb(so);
if (lr == NULL) {
lr_saved = __builtin_return_address(0);
} else {
lr_saved = lr;
}
#ifdef MORE_DICVLOCK_DEBUG
printf("div_unlock: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x "
"lr=0x%llx\n", (uint64_t)VM_KERNEL_ADDRPERM(so),
(uint64_t)VM_KERNEL_ADDRPERM(so->so_pcb), so->so_pcb ?
(uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)) : NULL,
so->so_usecount, lr_saved);
#endif
if (refcount) {
so->so_usecount--;
}
if (so->so_usecount < 0) {
panic("div_unlock: so=%p usecount=%x lrh= %s\n",
so, so->so_usecount, solockhistory_nr(so));
}
if (so->so_pcb == NULL) {
panic("div_unlock: so=%p NO PCB usecount=%x lr=%p lrh= %s\n",
so, so->so_usecount, lr_saved, solockhistory_nr(so));
}
mutex_held = &((struct inpcb *)so->so_pcb)->inpcb_mtx;
if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
lck_rw_lock_exclusive(divcbinfo.ipi_lock);
if (inp->inp_state != INPCB_STATE_DEAD) {
in_pcbdetach(inp);
}
in_pcbdispose(inp);
lck_rw_done(divcbinfo.ipi_lock);
return 0;
}
LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
so->unlock_lr[so->next_unlock_lr] = lr_saved;
so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
lck_mtx_unlock(mutex_held);
return 0;
}
__private_extern__ lck_mtx_t *
div_getlock(struct socket *so, __unused int flags)
{
struct inpcb *inpcb = (struct inpcb *)so->so_pcb;
if (so->so_pcb) {
if (so->so_usecount < 0) {
panic("div_getlock: so=%p usecount=%x lrh= %s\n",
so, so->so_usecount, solockhistory_nr(so));
}
return &inpcb->inpcb_mtx;
} else {
panic("div_getlock: so=%p NULL NO PCB lrh= %s\n",
so, solockhistory_nr(so));
return so->so_proto->pr_domain->dom_mtx;
}
}
struct pr_usrreqs div_usrreqs = {
.pru_abort = div_abort,
.pru_attach = div_attach,
.pru_bind = div_bind,
.pru_control = in_control,
.pru_detach = div_detach,
.pru_disconnect = div_disconnect,
.pru_peeraddr = in_getpeeraddr,
.pru_send = div_send,
.pru_shutdown = div_shutdown,
.pru_sockaddr = in_getsockaddr,
.pru_sosend = sosend,
.pru_soreceive = soreceive,
};