/*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
#include <sys/sysctl.h>
#include <sys/syslog.h>
+#include <kern/zalloc.h>
+
#include <net/if.h>
#include <net/if_types.h>
#include <net/route.h>
#include <netinet/in_var.h>
#include <netinet/ip_var.h>
#if INET6
+#include <netinet6/in6_pcb.h>
#include <netinet6/ip6_var.h>
#endif
#include <netinet/ip_icmp.h>
#if IPSEC
#include <netinet6/ipsec.h>
+#include <netinet6/esp.h>
extern int ipsec_bypass;
-extern lck_mtx_t *sadb_mutex;
#endif /*IPSEC*/
SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
&udpcksum, 0, "");
+static u_int32_t udps_in_sw_cksum;
+SYSCTL_UINT(_net_inet_udp, OID_AUTO, in_sw_cksum, CTLFLAG_RD,
+ &udps_in_sw_cksum, 0,
+ "Number of received packets checksummed in software");
+
+static u_int64_t udps_in_sw_cksum_bytes;
+SYSCTL_QUAD(_net_inet_udp, OID_AUTO, in_sw_cksum_bytes, CTLFLAG_RD,
+ &udps_in_sw_cksum_bytes,
+ "Amount of received data checksummed in software");
+
+static u_int32_t udps_out_sw_cksum;
+SYSCTL_UINT(_net_inet_udp, OID_AUTO, out_sw_cksum, CTLFLAG_RD,
+ &udps_out_sw_cksum, 0,
+ "Number of transmitted packets checksummed in software");
+
+static u_int64_t udps_out_sw_cksum_bytes;
+SYSCTL_QUAD(_net_inet_udp, OID_AUTO, out_sw_cksum_bytes, CTLFLAG_RD,
+ &udps_out_sw_cksum_bytes,
+ "Amount of transmitted data checksummed in software");
+
int log_in_vain = 0;
SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
&log_in_vain, 0, "Log all incoming UDP packets");
#define UDBHASHSIZE 16
#endif
-extern int apple_hwcksum_rx;
extern int esp_udp_encap_port;
-extern u_long route_generation;
-extern void ipfwsyslog( int level, char *format,...);
+extern void ipfwsyslog( int level, const char *format,...);
extern int fw_verbose;
+static int udp_gc_done = FALSE; /* Garbage collection performed last slowtimo */
+#if IPFIREWALL
#define log_in_vain_log( a ) { \
if ( (log_in_vain == 3 ) && (fw_verbose == 2)) { /* Apple logging, log to ipfw.log */ \
ipfwsyslog a ; \
} \
else log a ; \
}
+#else
+#define log_in_vain_log( a ) { log a; }
+#endif
struct udpstat udpstat; /* from udp_var.h */
SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RD,
SYSCTL_INT(_net_inet_udp, OID_AUTO, pcbcount, CTLFLAG_RD,
&udbinfo.ipi_count, 0, "Number of active PCBs");
-static struct sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
+__private_extern__ int udp_use_randomport = 1;
+SYSCTL_INT(_net_inet_udp, OID_AUTO, randomize_ports, CTLFLAG_RW,
+ &udp_use_randomport, 0, "Randomize UDP port numbers");
+
#if INET6
struct udp_in6 {
struct sockaddr_in6 uin6_sin;
u_char uin6_init_done : 1;
-} udp_in6 = {
- { sizeof(udp_in6.uin6_sin), AF_INET6 },
- 0
};
struct udp_ip6 {
struct ip6_hdr uip6_ip6;
u_char uip6_init_done : 1;
-} udp_ip6;
-#endif /* INET6 */
-
-static void udp_append(struct inpcb *last, struct ip *ip,
- struct mbuf *n, int off);
-#if INET6
+};
static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip);
+static void udp_append(struct inpcb *last, struct ip *ip,
+ struct mbuf *n, int off, struct sockaddr_in *pudp_in,
+ struct udp_in6 *pudp_in6, struct udp_ip6 *pudp_ip6);
+#else
+static void udp_append(struct inpcb *last, struct ip *ip,
+ struct mbuf *n, int off, struct sockaddr_in *pudp_in);
#endif
static int udp_detach(struct socket *so);
* allocate lock group attribute and group for udp pcb mutexes
*/
pcbinfo->mtx_grp_attr = lck_grp_attr_alloc_init();
- lck_grp_attr_setdefault(pcbinfo->mtx_grp_attr);
pcbinfo->mtx_grp = lck_grp_alloc_init("udppcb", pcbinfo->mtx_grp_attr);
pcbinfo->mtx_attr = lck_attr_alloc_init();
- lck_attr_setdefault(pcbinfo->mtx_attr);
if ((pcbinfo->mtx = lck_rw_alloc_init(pcbinfo->mtx_grp, pcbinfo->mtx_attr)) == NULL)
return; /* pretty much dead if this fails... */
-
- in_pcb_nat_init(&udbinfo, AF_INET, IPPROTO_UDP, SOCK_DGRAM);
#else
udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets,
ZONE_INTERRUPT, 0);
#endif
-
-#if 0
- /* for pcb sharing testing only */
- stat = in_pcb_new_share_client(&udbinfo, &fake_owner);
- kprintf("udp_init in_pcb_new_share_client - stat = %d\n", stat);
-
- laddr.s_addr = 0x11646464;
- faddr.s_addr = 0x11646465;
-
- lport = 1500;
- in_pcb_grab_port(&udbinfo, 0, laddr, &lport, faddr, 1600, 0, fake_owner);
- kprintf("udp_init in_pcb_grab_port - stat = %d\n", stat);
-
- stat = in_pcb_rem_share_client(&udbinfo, fake_owner);
- kprintf("udp_init in_pcb_rem_share_client - stat = %d\n", stat);
-
- stat = in_pcb_new_share_client(&udbinfo, &fake_owner);
- kprintf("udp_init in_pcb_new_share_client(2) - stat = %d\n", stat);
-
- laddr.s_addr = 0x11646464;
- faddr.s_addr = 0x11646465;
-
- lport = 1500;
- stat = in_pcb_grab_port(&udbinfo, 0, laddr, &lport, faddr, 1600, 0, fake_owner);
- kprintf("udp_init in_pcb_grab_port(2) - stat = %d\n", stat);
-#endif
}
void
struct ip save_ip;
struct sockaddr *append_sa;
struct inpcbinfo *pcbinfo = &udbinfo;
+ struct sockaddr_in udp_in = {
+ sizeof (udp_in), AF_INET, 0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
+ };
+#if INET6
+ struct udp_in6 udp_in6 = {
+ { sizeof (udp_in6.uin6_sin), AF_INET6, 0, 0,
+ IN6ADDR_ANY_INIT, 0 },
+ 0
+ };
+ struct udp_ip6 udp_ip6;
+#endif /* INET6 */
udpstat.udps_ipackets++;
*(uint32_t*)&((struct ipovly *)ip)->ih_x1[0] = *(uint32_t*)&b[0];
*(uint32_t*)&((struct ipovly *)ip)->ih_x1[4] = *(uint32_t*)&b[4];
*(uint8_t*)&((struct ipovly *)ip)->ih_x1[8] = *(uint8_t*)&b[8];
+ udp_in_cksum_stats(len);
}
if (uh->uh_sum) {
udpstat.udps_badsum++;
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
- struct inpcb *last;
+
+ int reuse_sock = 0, mcast_delivered = 0;
+
lck_rw_lock_shared(pcbinfo->mtx);
/*
* Deliver a multicast or broadcast datagram to *all* sockets
* Locate pcb(s) for datagram.
* (Algorithm copied from raw_intr().)
*/
- last = NULL;
#if INET6
udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
#endif
LIST_FOREACH(inp, &udb, inp_list) {
-#ifdef __APPLE__
- /* Ignore nat/SharedIP dummy pcbs */
- if (inp->inp_socket == &udbinfo.nat_dummy_socket)
- continue;
-#endif
if (inp->inp_socket == NULL)
continue;
if (inp != sotoinpcb(inp->inp_socket))
- panic("udp_input: bad so back ptr inp=%x\n", inp);
+ panic("udp_input: bad so back ptr inp=%p\n", inp);
#if INET6
if ((inp->inp_vflag & INP_IPV4) == 0)
continue;
#endif
+
if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) {
continue;
}
}
}
- if (last != NULL) {
- struct mbuf *n;
+ reuse_sock = inp->inp_socket->so_options& (SO_REUSEPORT|SO_REUSEADDR);
+ {
#if IPSEC
int skipit = 0;
/* check AH/ESP integrity. */
if (ipsec_bypass == 0) {
- lck_mtx_lock(sadb_mutex);
- if (ipsec4_in_reject_so(m, last->inp_socket)) {
- ipsecstat.in_polvio++;
+ if (ipsec4_in_reject_so(m, inp->inp_socket)) {
+ IPSEC_STAT_INCREMENT(ipsecstat.in_polvio);
/* do not inject data to pcb */
skipit = 1;
}
- lck_mtx_unlock(sadb_mutex);
}
- if (skipit == 0)
+ if (skipit == 0)
#endif /*IPSEC*/
- if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
- udp_append(last, ip, n,
- iphlen +
- sizeof(struct udphdr));
+ {
+ struct mbuf *n = NULL;
+
+ if (reuse_sock)
+ n = m_copy(m, 0, M_COPYALL);
+#if INET6
+ udp_append(inp, ip, m,
+ iphlen + sizeof(struct udphdr),
+ &udp_in, &udp_in6, &udp_ip6);
+#else
+ udp_append(inp, ip, m,
+ iphlen + sizeof(struct udphdr),
+ &udp_in);
+#endif /* INET6 */
+ mcast_delivered++;
+
+ m = n;
}
- udp_unlock(last->inp_socket, 1, 0);
+ udp_unlock(inp->inp_socket, 1, 0);
}
- last = inp;
/*
* Don't look for additional matches if this one does
* not have either the SO_REUSEPORT or SO_REUSEADDR
* socket options set. This heuristic avoids searching
* through all pcbs in the common case of a non-shared
- * port. It * assumes that an application will never
+ * port. It assumes that an application will never
* clear these options after setting them.
*/
- if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
+ if (reuse_sock == 0 || m == NULL)
break;
+ /*
+ * Recompute IP and UDP header pointers for new mbuf
+ */
+ ip = mtod(m, struct ip *);
+ uh = (struct udphdr *)((caddr_t)ip + iphlen);
}
lck_rw_done(pcbinfo->mtx);
- if (last == NULL) {
+ if (mcast_delivered == 0) {
/*
* No matching pcb found; discard datagram.
* (No need to send an ICMP Port Unreachable
udpstat.udps_noportbcast++;
goto bad;
}
-#if IPSEC
- /* check AH/ESP integrity. */
- if (ipsec_bypass == 0 && m) {
- lck_mtx_lock(sadb_mutex);
- if (ipsec4_in_reject_so(m, last->inp_socket)) {
- ipsecstat.in_polvio++;
- lck_mtx_unlock(sadb_mutex);
- udp_unlock(last->inp_socket, 1, 0);
- goto bad;
- }
- lck_mtx_unlock(sadb_mutex);
- }
-#endif /*IPSEC*/
- udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
- udp_unlock(last->inp_socket, 1, 0);
+
+ if (m != NULL) /* free the extra copy of mbuf or skipped by IPSec */
+ m_freem(m);
+ KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
return;
}
}
#if IPSEC
if (ipsec_bypass == 0 && inp != NULL) {
- lck_mtx_lock(sadb_mutex);
if (ipsec4_in_reject_so(m, inp->inp_socket)) {
- ipsecstat.in_polvio++;
- lck_mtx_unlock(sadb_mutex);
+ IPSEC_STAT_INCREMENT(ipsecstat.in_polvio);
udp_unlock(inp->inp_socket, 1, 0);
goto bad;
}
- lck_mtx_unlock(sadb_mutex);
}
#endif /*IPSEC*/
#if INET6
if (inp->inp_vflag & INP_IPV6) {
in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
- append_sa = (struct sockaddr *)&udp_in6;
+ append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
} else
#endif
append_sa = (struct sockaddr *)&udp_in;
/*
* subroutine of udp_input(), mainly for source code readability.
- * caller must properly init udp_ip6 and udp_in6 beforehand.
*/
static void
-udp_append(last, ip, n, off)
- struct inpcb *last;
- struct ip *ip;
- struct mbuf *n;
- int off;
+#if INET6
+udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off,
+ struct sockaddr_in *pudp_in, struct udp_in6 *pudp_in6,
+ struct udp_ip6 *pudp_ip6)
+#else
+udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off,
+ struct sockaddr_in *pudp_in)
+#endif
{
struct sockaddr *append_sa;
struct mbuf *opts = 0;
+#if CONFIG_MACF_NET
+ if (mac_inpcb_check_deliver(last, n, AF_INET, SOCK_DGRAM) != 0) {
+ m_freem(n);
+ return;
+ }
+#endif
if (last->inp_flags & INP_CONTROLOPTS ||
last->inp_socket->so_options & SO_TIMESTAMP) {
#if INET6
if (last->inp_vflag & INP_IPV6) {
int savedflags;
- if (udp_ip6.uip6_init_done == 0) {
- ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
- udp_ip6.uip6_init_done = 1;
+ if (pudp_ip6->uip6_init_done == 0) {
+ ip_2_ip6_hdr(&pudp_ip6->uip6_ip6, ip);
+ pudp_ip6->uip6_init_done = 1;
}
savedflags = last->inp_flags;
last->inp_flags &= ~INP_UNMAPPABLEOPTS;
- ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
+ ip6_savecontrol(last, &opts, &pudp_ip6->uip6_ip6, n);
last->inp_flags = savedflags;
} else
#endif
}
#if INET6
if (last->inp_vflag & INP_IPV6) {
- if (udp_in6.uin6_init_done == 0) {
- in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
- udp_in6.uin6_init_done = 1;
+ if (pudp_in6->uin6_init_done == 0) {
+ in6_sin_2_v4mapsin6(pudp_in, &pudp_in6->uin6_sin);
+ pudp_in6->uin6_init_done = 1;
}
- append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
+ append_sa = (struct sockaddr *)&pudp_in6->uin6_sin;
} else
#endif
- append_sa = (struct sockaddr *)&udp_in;
+ append_sa = (struct sockaddr *)pudp_in;
m_adj(n, off);
if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts, NULL) == 0) {
udpstat.udps_fullsock++;
in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify);
}
+int
+udp_ctloutput(struct socket *so, struct sockopt *sopt)
+{
+ int error, optval;
+ struct inpcb *inp;
+
+ if (sopt->sopt_level != IPPROTO_UDP)
+ return (ip_ctloutput(so, sopt));
+
+ error = 0;
+ inp = sotoinpcb(so);
+
+ switch (sopt->sopt_dir) {
+ case SOPT_SET:
+ switch (sopt->sopt_name) {
+ case UDP_NOCKSUM:
+ /* This option is settable only for UDP over IPv4 */
+ if (!(inp->inp_vflag & INP_IPV4)) {
+ error = EINVAL;
+ break;
+ }
+
+ if ((error = sooptcopyin(sopt, &optval, sizeof (optval),
+ sizeof (optval))) != 0)
+ break;
+
+ if (optval != 0)
+ inp->inp_flags |= INP_UDP_NOCKSUM;
+ else
+ inp->inp_flags &= ~INP_UDP_NOCKSUM;
+ break;
+
+ default:
+ error = ENOPROTOOPT;
+ break;
+ }
+ break;
+
+ case SOPT_GET:
+ switch (sopt->sopt_name) {
+ case UDP_NOCKSUM:
+ optval = inp->inp_flags & INP_UDP_NOCKSUM;
+ break;
+
+ default:
+ error = ENOPROTOOPT;
+ break;
+ }
+ if (error == 0)
+ error = sooptcopyout(sopt, &optval, sizeof (optval));
+ break;
+ }
+ return (error);
+}
+
static int
udp_pcblist SYSCTL_HANDLER_ARGS
{
+#pragma unused(oidp, arg1, arg2)
int error, i, n;
struct inpcb *inp, **inp_list;
inp_gen_t gencnt;
SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
udp_pcblist, "S,xinpcb", "List of active UDP sockets");
+#if !CONFIG_EMBEDDED
+
+static int
+udp_pcblist64 SYSCTL_HANDLER_ARGS
+{
+#pragma unused(oidp, arg1, arg2)
+ int error, i, n;
+ struct inpcb *inp, **inp_list;
+ inp_gen_t gencnt;
+ struct xinpgen xig;
+
+ /*
+ * The process of preparing the TCB list is too time-consuming and
+ * resource-intensive to repeat twice on every request.
+ */
+ lck_rw_lock_shared(udbinfo.mtx);
+ if (req->oldptr == USER_ADDR_NULL) {
+ n = udbinfo.ipi_count;
+ req->oldidx = 2 * (sizeof xig)
+ + (n + n/8) * sizeof(struct xinpcb64);
+ lck_rw_done(udbinfo.mtx);
+ return 0;
+ }
+
+ if (req->newptr != USER_ADDR_NULL) {
+ lck_rw_done(udbinfo.mtx);
+ return EPERM;
+ }
+
+ /*
+ * OK, now we're committed to doing something.
+ */
+ gencnt = udbinfo.ipi_gencnt;
+ n = udbinfo.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(udbinfo.mtx);
+ return error;
+ }
+ /*
+ * We are done if there is no pcb
+ */
+ if (n == 0) {
+ lck_rw_done(udbinfo.mtx);
+ return 0;
+ }
+ inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
+ if (inp_list == 0) {
+ lck_rw_done(udbinfo.mtx);
+ return ENOMEM;
+ }
+
+ for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
+ inp = LIST_NEXT(inp, inp_list)) {
+ if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD)
+ 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 xinpcb64 xi;
+
+ bzero(&xi, sizeof(xi));
+ xi.xi_len = sizeof xi;
+ inpcb_to_xinpcb64(inp, &xi);
+ if (inp->inp_socket)
+ sotoxsocket64(inp->inp_socket, &xi.xi_socket);
+ error = SYSCTL_OUT(req, &xi, sizeof xi);
+ }
+ }
+ if (!error) {
+ /*
+ * Give the user an updated idea of our state.
+ * If the generation differs from what we told
+ * her before, she knows that something happened
+ * while we were processing this request, and it
+ * might be necessary to retry.
+ */
+ bzero(&xig, sizeof(xig));
+ xig.xig_len = sizeof xig;
+ xig.xig_gen = udbinfo.ipi_gencnt;
+ xig.xig_sogen = so_gencnt;
+ xig.xig_count = udbinfo.ipi_count;
+ error = SYSCTL_OUT(req, &xig, sizeof xig);
+ }
+ FREE(inp_list, M_TEMP);
+ lck_rw_done(udbinfo.mtx);
+ return error;
+}
+
+SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist64, CTLFLAG_RD, 0, 0,
+ udp_pcblist64, "S,xinpcb64", "List of active UDP sockets");
+
+#endif /* !CONFIG_EMBEDDED */
static __inline__ u_int16_t
get_socket_id(struct socket * s)
if (s == NULL) {
return (0);
}
- val = (u_int16_t)(((u_int32_t)s) / sizeof(struct socket));
+ val = (u_int16_t)(((uintptr_t)s) / sizeof(struct socket));
if (val == 0) {
val = 0xffff;
}
{
register struct udpiphdr *ui;
register int len = m->m_pkthdr.len;
- struct sockaddr_in *sin, src;
+ struct sockaddr_in *sin;
struct in_addr origladdr, laddr, faddr;
u_short lport, fport;
struct sockaddr_in *ifaddr;
int error = 0, udp_dodisconnect = 0;
-
+ struct socket *so = inp->inp_socket;
+ int soopts = 0;
+ struct mbuf *inpopts;
+ struct ip_moptions *mopts;
+ struct route ro;
+ struct ip_out_args ipoa;
+#if PKT_PRIORITY
+ mbuf_traffic_class_t mtc = MBUF_TC_NONE;
+#endif /* PKT_PRIORITY */
KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
- if (control)
- m_freem(control); /* XXX */
-
+ if (control != NULL) {
+#if PKT_PRIORITY
+ mtc = mbuf_traffic_class_from_control(control);
+#endif /* PKT_PRIORITY */
+ m_freem(control);
+ }
KERNEL_DEBUG(DBG_LAYER_OUT_BEG, inp->inp_fport, inp->inp_lport,
inp->inp_laddr.s_addr, inp->inp_faddr.s_addr,
(htons((u_short)len + sizeof (struct udphdr))));
goto release;
}
+ lck_mtx_assert(inp->inpcb_mtx, LCK_MTX_ASSERT_OWNED);
+
+ /* If socket was bound to an ifindex, tell ip_output about it */
+ ipoa.ipoa_ifscope = (inp->inp_flags & INP_BOUND_IF) ?
+ inp->inp_boundif : IFSCOPE_NONE;
+ soopts |= IP_OUTARGS;
+
/* If there was a routing change, discard cached route and check
* that we have a valid source address.
* Reacquire a new source address if INADDR_ANY was specified
*/
-
-#if 1
- lck_mtx_assert(inp->inpcb_mtx, LCK_MTX_ASSERT_OWNED);
-#endif
-
- if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->generation_id != route_generation) {
- if (ifa_foraddr(inp->inp_laddr.s_addr) == 0) { /* src address is gone */
- if (inp->inp_flags & INP_INADDR_ANY)
- inp->inp_faddr.s_addr = INADDR_ANY; /* new src will be set later */
- else {
+ if (inp->inp_route.ro_rt != NULL &&
+ inp->inp_route.ro_rt->generation_id != route_generation) {
+ struct in_ifaddr *ia;
+
+ /* src address is gone? */
+ if ((ia = ifa_foraddr(inp->inp_laddr.s_addr)) == NULL) {
+ if (inp->inp_flags & INP_INADDR_ANY) {
+ /* new src will be set later */
+ inp->inp_laddr.s_addr = INADDR_ANY;
+ } else {
error = EADDRNOTAVAIL;
goto release;
}
}
- rtfree(inp->inp_route.ro_rt);
- inp->inp_route.ro_rt = (struct rtentry *)0;
+ if (ia != NULL)
+ ifafree(&ia->ia_ifa);
+ if (inp->inp_route.ro_rt != NULL)
+ rtfree(inp->inp_route.ro_rt);
+ inp->inp_route.ro_rt = NULL;
}
origladdr= laddr = inp->inp_laddr;
}
}
+#if CONFIG_MACF_NET
+ mac_mbuf_label_associate_inpcb(inp, m);
+#endif
/*
* Calculate data length and get a mbuf
/*
* Set up checksum and output datagram.
*/
- if (udpcksum) {
+ if (udpcksum && !(inp->inp_flags & INP_UDP_NOCKSUM)) {
ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
m->m_pkthdr.csum_flags = CSUM_UDP;
}
#endif /*IPSEC*/
m->m_pkthdr.socket_id = get_socket_id(inp->inp_socket);
- error = ip_output_list(m, 0, inp->inp_options, &inp->inp_route,
- (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
- inp->inp_moptions);
+
+ inpopts = inp->inp_options;
+ soopts |= (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST));
+ mopts = inp->inp_moptions;
+
+ /* Copy the cached route and take an extra reference */
+ inp_route_copyout(inp, &ro);
+
+#if PKT_PRIORITY
+ set_traffic_class(m, so, mtc);
+#endif /* PKT_PRIORITY */
+
+ socket_unlock(so, 0);
+ /* XXX jgraessley please look at XXX */
+ error = ip_output_list(m, 0, inpopts, &ro, soopts, mopts, &ipoa);
+ socket_lock(so, 0);
+
+ /* Synchronize PCB cached route */
+ inp_route_copyin(inp, &ro);
if (udp_dodisconnect) {
+#if IFNET_ROUTE_REFCNT
+ /* Always discard the cached route for unconnected socket */
+ if (inp->inp_route.ro_rt != NULL) {
+ rtfree(inp->inp_route.ro_rt);
+ inp->inp_route.ro_rt = NULL;
+ }
+#endif /* IFNET_ROUTE_REFCNT */
in_pcbdisconnect(inp);
inp->inp_laddr = origladdr; /* XXX rehash? */
}
+#if IFNET_ROUTE_REFCNT
+ else if (inp->inp_route.ro_rt != NULL &&
+ (inp->inp_route.ro_rt->rt_flags & (RTF_MULTICAST|RTF_BROADCAST))) {
+ /* Always discard non-unicast cached route */
+ rtfree(inp->inp_route.ro_rt);
+ inp->inp_route.ro_rt = NULL;
+ }
+#endif /* IFNET_ROUTE_REFCNT */
+
KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0,0,0,0);
return (error);
abort:
if (udp_dodisconnect) {
- in_pcbdisconnect(inp);
- inp->inp_laddr = origladdr; /* XXX rehash? */
+#if IFNET_ROUTE_REFCNT
+ /* Always discard the cached route for unconnected socket */
+ if (inp->inp_route.ro_rt != NULL) {
+ rtfree(inp->inp_route.ro_rt);
+ inp->inp_route.ro_rt = NULL;
+ }
+#endif /* IFNET_ROUTE_REFCNT */
+ in_pcbdisconnect(inp);
+ inp->inp_laddr = origladdr; /* XXX rehash? */
}
+#if IFNET_ROUTE_REFCNT
+ else if (inp->inp_route.ro_rt != NULL &&
+ (inp->inp_route.ro_rt->rt_flags & (RTF_MULTICAST|RTF_BROADCAST))) {
+ /* Always discard non-unicast cached route */
+ rtfree(inp->inp_route.ro_rt);
+ inp->inp_route.ro_rt = NULL;
+ }
+#endif /* IFNET_ROUTE_REFCNT */
release:
m_freem(m);
return (error);
}
-u_long udp_sendspace = 9216; /* really max datagram size */
- /* 40 1K datagrams */
-SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
- &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
-
-u_long udp_recvspace = 40 * (1024 +
+u_int32_t udp_sendspace = 9216; /* really max datagram size */
+/* 40 1K datagrams */
+u_int32_t udp_recvspace = 40 * (1024 +
#if INET6
sizeof(struct sockaddr_in6)
#else
sizeof(struct sockaddr_in)
#endif
);
-SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
- &udp_recvspace, 0, "Maximum incoming UDP datagram size");
+
+/* Check that the values of udp send and recv space do not exceed sb_max */
+static int
+sysctl_udp_sospace(struct sysctl_oid *oidp, __unused void *arg1,
+ __unused int arg2, struct sysctl_req *req) {
+ u_int32_t new_value = 0, *space_p = NULL;
+ int changed = 0, error = 0;
+ u_quad_t sb_effective_max = (sb_max/ (MSIZE+MCLBYTES)) * MCLBYTES;
+
+ switch (oidp->oid_number) {
+ case UDPCTL_RECVSPACE:
+ space_p = &udp_recvspace;
+ break;
+ case UDPCTL_MAXDGRAM:
+ space_p = &udp_sendspace;
+ break;
+ default:
+ return EINVAL;
+ }
+ error = sysctl_io_number(req, *space_p, sizeof(u_int32_t),
+ &new_value, &changed);
+ if (changed) {
+ if (new_value > 0 && new_value <= sb_effective_max) {
+ *space_p = new_value;
+ } else {
+ error = ERANGE;
+ }
+ }
+ return error;
+}
+
+SYSCTL_PROC(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLTYPE_INT | CTLFLAG_RW,
+ &udp_recvspace, 0, &sysctl_udp_sospace, "IU", "Maximum incoming UDP datagram size");
+
+SYSCTL_PROC(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLTYPE_INT | CTLFLAG_RW,
+ &udp_sendspace, 0, &sysctl_udp_sospace, "IU", "Maximum outgoing UDP datagram size");
static int
udp_abort(struct socket *so)
inp = sotoinpcb(so);
if (inp == 0)
- panic("udp_abort: so=%x null inp\n", so); /* ??? possible? panic instead? */
+ panic("udp_abort: so=%p null inp\n", so); /* ??? possible? panic instead? */
soisdisconnected(so);
in_pcbdetach(inp);
return 0;
}
static int
-udp_attach(struct socket *so, int proto, struct proc *p)
+udp_attach(struct socket *so, __unused int proto, struct proc *p)
{
struct inpcb *inp;
int error;
inp = sotoinpcb(so);
if (inp != 0)
- panic ("udp_attach so=%x inp=%x\n", so, inp);
+ panic ("udp_attach so=%p inp=%p\n", so, inp);
error = in_pcballoc(so, &udbinfo, p);
if (error)
struct inpcb *inp;
int error;
+ if (nam->sa_family != 0 && nam->sa_family != AF_INET
+ && nam->sa_family != AF_INET6) {
+ return EAFNOSUPPORT;
+ }
inp = sotoinpcb(so);
if (inp == 0)
return EINVAL;
inp = sotoinpcb(so);
if (inp == 0)
- panic("udp_detach: so=%x null inp\n", so); /* ??? possible? panic instead? */
+ panic("udp_detach: so=%p null inp\n", so); /* ??? possible? panic instead? */
in_pcbdetach(inp);
inp->inp_state = INPCB_STATE_DEAD;
return 0;
}
static int
-udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
+udp_send(struct socket *so, __unused int flags, struct mbuf *m, struct sockaddr *addr,
struct mbuf *control, struct proc *p)
{
struct inpcb *inp;
m_freem(m);
return EINVAL;
}
+
return udp_output(inp, m, addr, control, p);
}
int
-udp_lock(so, refcount, debug)
- struct socket *so;
- int refcount, debug;
+udp_lock(struct socket *so, int refcount, void *debug)
{
- int lr_saved;
-#ifdef __ppc__
- if (debug == 0) {
- __asm__ volatile("mflr %0" : "=r" (lr_saved));
- }
- else lr_saved = debug;
-#endif
+ void *lr_saved;
+
+ if (debug == NULL)
+ lr_saved = __builtin_return_address(0);
+ else
+ lr_saved = debug;
if (so->so_pcb) {
- lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_NOTOWNED);
+ lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx,
+ LCK_MTX_ASSERT_NOTOWNED);
lck_mtx_lock(((struct inpcb *)so->so_pcb)->inpcb_mtx);
+ } else {
+ panic("udp_lock: so=%p NO PCB! lr=%p lrh= %s\n",
+ so, lr_saved, solockhistory_nr(so));
+ /* NOTREACHED */
}
- else {
- panic("udp_lock: so=%x NO PCB! lr=%x\n", so, lr_saved);
- lck_mtx_assert(so->so_proto->pr_domain->dom_mtx, LCK_MTX_ASSERT_NOTOWNED);
- lck_mtx_lock(so->so_proto->pr_domain->dom_mtx);
- }
-
- if (refcount)
+ if (refcount)
so->so_usecount++;
- so->reserved3= lr_saved;
+ so->lock_lr[so->next_lock_lr] = lr_saved;
+ so->next_lock_lr = (so->next_lock_lr+1) % SO_LCKDBG_MAX;
return (0);
}
int
-udp_unlock(so, refcount, debug)
- struct socket *so;
- int refcount;
- int debug;
+udp_unlock(struct socket *so, int refcount, void *debug)
{
- int lr_saved;
- struct inpcb *inp = sotoinpcb(so);
- struct inpcbinfo *pcbinfo = &udbinfo;
-#ifdef __ppc__
- if (debug == 0) {
- __asm__ volatile("mflr %0" : "=r" (lr_saved));
- }
- else lr_saved = debug;
-#endif
- if (refcount) {
+ void *lr_saved;
+
+ if (debug == NULL)
+ lr_saved = __builtin_return_address(0);
+ else
+ lr_saved = debug;
+
+ if (refcount)
so->so_usecount--;
-#if 0
- if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) {
- if (lck_rw_try_lock_exclusive(pcbinfo->mtx)) {
- in_pcbdispose(inp);
- lck_rw_done(pcbinfo->mtx);
- return(0);
- }
- }
-#endif
- }
+
if (so->so_pcb == NULL) {
- panic("udp_unlock: so=%x NO PCB! lr=%x\n", so, lr_saved);
- lck_mtx_assert(so->so_proto->pr_domain->dom_mtx, LCK_MTX_ASSERT_OWNED);
- lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx);
- }
- else {
- lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED);
+ panic("udp_unlock: so=%p NO PCB! lr=%p lrh= %s\n",
+ so, lr_saved, solockhistory_nr(so));
+ /* NOTREACHED */
+ } else {
+ lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx,
+ 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(((struct inpcb *)so->so_pcb)->inpcb_mtx);
}
- so->reserved4 = lr_saved;
return (0);
}
lck_mtx_t *
-udp_getlock(so, locktype)
- struct socket *so;
- int locktype;
+udp_getlock(struct socket *so, __unused int locktype)
{
struct inpcb *inp = sotoinpcb(so);
if (so->so_pcb)
return(inp->inpcb_mtx);
else {
- panic("udp_getlock: so=%x NULL so_pcb\n", so);
+ panic("udp_getlock: so=%p NULL so_pcb lrh= %s\n",
+ so, solockhistory_nr(so));
return (so->so_proto->pr_domain->dom_mtx);
}
}
struct socket *so;
struct inpcbinfo *pcbinfo = &udbinfo;
- lck_rw_lock_exclusive(pcbinfo->mtx);
+ if (lck_rw_try_lock_exclusive(pcbinfo->mtx) == FALSE) {
+ if (udp_gc_done == TRUE) {
+ udp_gc_done = FALSE;
+ return; /* couldn't get the lock, better lock next time */
+ }
+ lck_rw_lock_exclusive(pcbinfo->mtx);
+ }
+
+ udp_gc_done = TRUE;
for (inp = udb.lh_first; inp != NULL; inp = inpnxt) {
inpnxt = inp->inp_list.le_next;
-
- /* Ignore nat/SharedIP dummy pcbs */
- if (inp->inp_socket == &udbinfo.nat_dummy_socket)
- continue;
if (inp->inp_wantcnt != WNT_STOPUSING)
continue;
if (!lck_mtx_try_lock(inp->inpcb_mtx)) /* skip if busy, no hurry for cleanup... */
continue;
- if (so->so_usecount == 0)
+ if (so->so_usecount == 0) {
+ if (inp->inp_state != INPCB_STATE_DEAD) {
+#if INET6
+ if (INP_CHECK_SOCKAF(so, AF_INET6))
+ in6_pcbdetach(inp);
+ else
+#endif /* INET6 */
+ in_pcbdetach(inp);
+ }
in_pcbdispose(inp);
- else
+ } else {
lck_mtx_unlock(inp->inpcb_mtx);
+ }
}
lck_rw_done(pcbinfo->mtx);
}
return 1;
}
+void
+udp_in_cksum_stats(u_int32_t len)
+{
+ udps_in_sw_cksum++;
+ udps_in_sw_cksum_bytes += len;
+}
+
+void
+udp_out_cksum_stats(u_int32_t len)
+{
+ udps_out_sw_cksum++;
+ udps_out_sw_cksum_bytes += len;
+}
projects / apple / xnu.git / blobdiff