ppc/pnv: create the ICP object under PnvCore
[qemu/ar7.git] / slirp / slirp.c
blob9a50918346b1e4e4e0e49df808f25c9b9cc2e908
1 /*
2 * libslirp glue
4 * Copyright (c) 2004-2008 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
24 #include "qemu/osdep.h"
25 #include "qemu-common.h"
26 #include "qemu/timer.h"
27 #include "qemu/error-report.h"
28 #include "sysemu/char.h"
29 #include "slirp.h"
30 #include "hw/hw.h"
31 #include "qemu/cutils.h"
33 #ifndef _WIN32
34 #include <net/if.h>
35 #endif
37 /* host loopback address */
38 struct in_addr loopback_addr;
39 /* host loopback network mask */
40 unsigned long loopback_mask;
42 /* emulated hosts use the MAC addr 52:55:IP:IP:IP:IP */
43 static const uint8_t special_ethaddr[ETH_ALEN] = {
44 0x52, 0x55, 0x00, 0x00, 0x00, 0x00
47 u_int curtime;
49 static QTAILQ_HEAD(slirp_instances, Slirp) slirp_instances =
50 QTAILQ_HEAD_INITIALIZER(slirp_instances);
52 static struct in_addr dns_addr;
53 #ifndef _WIN32
54 static struct in6_addr dns6_addr;
55 #endif
56 static u_int dns_addr_time;
57 #ifndef _WIN32
58 static u_int dns6_addr_time;
59 #endif
61 #define TIMEOUT_FAST 2 /* milliseconds */
62 #define TIMEOUT_SLOW 499 /* milliseconds */
63 /* for the aging of certain requests like DNS */
64 #define TIMEOUT_DEFAULT 1000 /* milliseconds */
66 #ifdef _WIN32
68 int get_dns_addr(struct in_addr *pdns_addr)
70 FIXED_INFO *FixedInfo=NULL;
71 ULONG BufLen;
72 DWORD ret;
73 IP_ADDR_STRING *pIPAddr;
74 struct in_addr tmp_addr;
76 if (dns_addr.s_addr != 0 && (curtime - dns_addr_time) < TIMEOUT_DEFAULT) {
77 *pdns_addr = dns_addr;
78 return 0;
81 FixedInfo = (FIXED_INFO *)GlobalAlloc(GPTR, sizeof(FIXED_INFO));
82 BufLen = sizeof(FIXED_INFO);
84 if (ERROR_BUFFER_OVERFLOW == GetNetworkParams(FixedInfo, &BufLen)) {
85 if (FixedInfo) {
86 GlobalFree(FixedInfo);
87 FixedInfo = NULL;
89 FixedInfo = GlobalAlloc(GPTR, BufLen);
92 if ((ret = GetNetworkParams(FixedInfo, &BufLen)) != ERROR_SUCCESS) {
93 printf("GetNetworkParams failed. ret = %08x\n", (u_int)ret );
94 if (FixedInfo) {
95 GlobalFree(FixedInfo);
96 FixedInfo = NULL;
98 return -1;
101 pIPAddr = &(FixedInfo->DnsServerList);
102 inet_aton(pIPAddr->IpAddress.String, &tmp_addr);
103 *pdns_addr = tmp_addr;
104 dns_addr = tmp_addr;
105 dns_addr_time = curtime;
106 if (FixedInfo) {
107 GlobalFree(FixedInfo);
108 FixedInfo = NULL;
110 return 0;
113 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
115 return -1;
118 static void winsock_cleanup(void)
120 WSACleanup();
123 #else
125 static int get_dns_addr_cached(void *pdns_addr, void *cached_addr,
126 socklen_t addrlen,
127 struct stat *cached_stat, u_int *cached_time)
129 struct stat old_stat;
130 if (curtime - *cached_time < TIMEOUT_DEFAULT) {
131 memcpy(pdns_addr, cached_addr, addrlen);
132 return 0;
134 old_stat = *cached_stat;
135 if (stat("/etc/resolv.conf", cached_stat) != 0) {
136 return -1;
138 if (cached_stat->st_dev == old_stat.st_dev
139 && cached_stat->st_ino == old_stat.st_ino
140 && cached_stat->st_size == old_stat.st_size
141 && cached_stat->st_mtime == old_stat.st_mtime) {
142 memcpy(pdns_addr, cached_addr, addrlen);
143 return 0;
145 return 1;
148 static int get_dns_addr_resolv_conf(int af, void *pdns_addr, void *cached_addr,
149 socklen_t addrlen, uint32_t *scope_id,
150 u_int *cached_time)
152 char buff[512];
153 char buff2[257];
154 FILE *f;
155 int found = 0;
156 void *tmp_addr = alloca(addrlen);
157 unsigned if_index;
159 f = fopen("/etc/resolv.conf", "r");
160 if (!f)
161 return -1;
163 #ifdef DEBUG
164 fprintf(stderr, "IP address of your DNS(s): ");
165 #endif
166 while (fgets(buff, 512, f) != NULL) {
167 if (sscanf(buff, "nameserver%*[ \t]%256s", buff2) == 1) {
168 char *c = strchr(buff2, '%');
169 if (c) {
170 if_index = if_nametoindex(c + 1);
171 *c = '\0';
172 } else {
173 if_index = 0;
176 if (!inet_pton(af, buff2, tmp_addr)) {
177 continue;
179 /* If it's the first one, set it to dns_addr */
180 if (!found) {
181 memcpy(pdns_addr, tmp_addr, addrlen);
182 memcpy(cached_addr, tmp_addr, addrlen);
183 if (scope_id) {
184 *scope_id = if_index;
186 *cached_time = curtime;
188 #ifdef DEBUG
189 else
190 fprintf(stderr, ", ");
191 #endif
192 if (++found > 3) {
193 #ifdef DEBUG
194 fprintf(stderr, "(more)");
195 #endif
196 break;
198 #ifdef DEBUG
199 else {
200 char s[INET6_ADDRSTRLEN];
201 const char *res = inet_ntop(af, tmp_addr, s, sizeof(s));
202 if (!res) {
203 res = "(string conversion error)";
205 fprintf(stderr, "%s", res);
207 #endif
210 fclose(f);
211 if (!found)
212 return -1;
213 return 0;
216 int get_dns_addr(struct in_addr *pdns_addr)
218 static struct stat dns_addr_stat;
220 if (dns_addr.s_addr != 0) {
221 int ret;
222 ret = get_dns_addr_cached(pdns_addr, &dns_addr, sizeof(dns_addr),
223 &dns_addr_stat, &dns_addr_time);
224 if (ret <= 0) {
225 return ret;
228 return get_dns_addr_resolv_conf(AF_INET, pdns_addr, &dns_addr,
229 sizeof(dns_addr), NULL, &dns_addr_time);
232 int get_dns6_addr(struct in6_addr *pdns6_addr, uint32_t *scope_id)
234 static struct stat dns6_addr_stat;
236 if (!in6_zero(&dns6_addr)) {
237 int ret;
238 ret = get_dns_addr_cached(pdns6_addr, &dns6_addr, sizeof(dns6_addr),
239 &dns6_addr_stat, &dns6_addr_time);
240 if (ret <= 0) {
241 return ret;
244 return get_dns_addr_resolv_conf(AF_INET6, pdns6_addr, &dns6_addr,
245 sizeof(dns6_addr),
246 scope_id, &dns6_addr_time);
249 #endif
251 static void slirp_init_once(void)
253 static int initialized;
254 #ifdef _WIN32
255 WSADATA Data;
256 #endif
258 if (initialized) {
259 return;
261 initialized = 1;
263 #ifdef _WIN32
264 WSAStartup(MAKEWORD(2,0), &Data);
265 atexit(winsock_cleanup);
266 #endif
268 loopback_addr.s_addr = htonl(INADDR_LOOPBACK);
269 loopback_mask = htonl(IN_CLASSA_NET);
272 static void slirp_state_save(QEMUFile *f, void *opaque);
273 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id);
275 Slirp *slirp_init(int restricted, bool in_enabled, struct in_addr vnetwork,
276 struct in_addr vnetmask, struct in_addr vhost,
277 bool in6_enabled,
278 struct in6_addr vprefix_addr6, uint8_t vprefix_len,
279 struct in6_addr vhost6, const char *vhostname,
280 const char *tftp_path, const char *bootfile,
281 struct in_addr vdhcp_start, struct in_addr vnameserver,
282 struct in6_addr vnameserver6, const char **vdnssearch,
283 void *opaque)
285 Slirp *slirp = g_malloc0(sizeof(Slirp));
287 slirp_init_once();
289 slirp->grand = g_rand_new();
290 slirp->restricted = restricted;
292 slirp->in_enabled = in_enabled;
293 slirp->in6_enabled = in6_enabled;
295 if_init(slirp);
296 ip_init(slirp);
297 ip6_init(slirp);
299 /* Initialise mbufs *after* setting the MTU */
300 m_init(slirp);
302 slirp->vnetwork_addr = vnetwork;
303 slirp->vnetwork_mask = vnetmask;
304 slirp->vhost_addr = vhost;
305 slirp->vprefix_addr6 = vprefix_addr6;
306 slirp->vprefix_len = vprefix_len;
307 slirp->vhost_addr6 = vhost6;
308 if (vhostname) {
309 pstrcpy(slirp->client_hostname, sizeof(slirp->client_hostname),
310 vhostname);
312 slirp->tftp_prefix = g_strdup(tftp_path);
313 slirp->bootp_filename = g_strdup(bootfile);
314 slirp->vdhcp_startaddr = vdhcp_start;
315 slirp->vnameserver_addr = vnameserver;
316 slirp->vnameserver_addr6 = vnameserver6;
318 if (vdnssearch) {
319 translate_dnssearch(slirp, vdnssearch);
322 slirp->opaque = opaque;
324 register_savevm(NULL, "slirp", 0, 4,
325 slirp_state_save, slirp_state_load, slirp);
327 QTAILQ_INSERT_TAIL(&slirp_instances, slirp, entry);
329 return slirp;
332 void slirp_cleanup(Slirp *slirp)
334 QTAILQ_REMOVE(&slirp_instances, slirp, entry);
336 unregister_savevm(NULL, "slirp", slirp);
338 ip_cleanup(slirp);
339 ip6_cleanup(slirp);
340 m_cleanup(slirp);
342 g_rand_free(slirp->grand);
344 g_free(slirp->vdnssearch);
345 g_free(slirp->tftp_prefix);
346 g_free(slirp->bootp_filename);
347 g_free(slirp);
350 #define CONN_CANFSEND(so) (((so)->so_state & (SS_FCANTSENDMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
351 #define CONN_CANFRCV(so) (((so)->so_state & (SS_FCANTRCVMORE|SS_ISFCONNECTED)) == SS_ISFCONNECTED)
353 static void slirp_update_timeout(uint32_t *timeout)
355 Slirp *slirp;
356 uint32_t t;
358 if (*timeout <= TIMEOUT_FAST) {
359 return;
362 t = MIN(1000, *timeout);
364 /* If we have tcp timeout with slirp, then we will fill @timeout with
365 * more precise value.
367 QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
368 if (slirp->time_fasttimo) {
369 *timeout = TIMEOUT_FAST;
370 return;
372 if (slirp->do_slowtimo) {
373 t = MIN(TIMEOUT_SLOW, t);
376 *timeout = t;
379 void slirp_pollfds_fill(GArray *pollfds, uint32_t *timeout)
381 Slirp *slirp;
382 struct socket *so, *so_next;
384 if (QTAILQ_EMPTY(&slirp_instances)) {
385 return;
389 * First, TCP sockets
392 QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
394 * *_slowtimo needs calling if there are IP fragments
395 * in the fragment queue, or there are TCP connections active
397 slirp->do_slowtimo = ((slirp->tcb.so_next != &slirp->tcb) ||
398 (&slirp->ipq.ip_link != slirp->ipq.ip_link.next));
400 for (so = slirp->tcb.so_next; so != &slirp->tcb;
401 so = so_next) {
402 int events = 0;
404 so_next = so->so_next;
406 so->pollfds_idx = -1;
409 * See if we need a tcp_fasttimo
411 if (slirp->time_fasttimo == 0 &&
412 so->so_tcpcb->t_flags & TF_DELACK) {
413 slirp->time_fasttimo = curtime; /* Flag when want a fasttimo */
417 * NOFDREF can include still connecting to local-host,
418 * newly socreated() sockets etc. Don't want to select these.
420 if (so->so_state & SS_NOFDREF || so->s == -1) {
421 continue;
425 * Set for reading sockets which are accepting
427 if (so->so_state & SS_FACCEPTCONN) {
428 GPollFD pfd = {
429 .fd = so->s,
430 .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
432 so->pollfds_idx = pollfds->len;
433 g_array_append_val(pollfds, pfd);
434 continue;
438 * Set for writing sockets which are connecting
440 if (so->so_state & SS_ISFCONNECTING) {
441 GPollFD pfd = {
442 .fd = so->s,
443 .events = G_IO_OUT | G_IO_ERR,
445 so->pollfds_idx = pollfds->len;
446 g_array_append_val(pollfds, pfd);
447 continue;
451 * Set for writing if we are connected, can send more, and
452 * we have something to send
454 if (CONN_CANFSEND(so) && so->so_rcv.sb_cc) {
455 events |= G_IO_OUT | G_IO_ERR;
459 * Set for reading (and urgent data) if we are connected, can
460 * receive more, and we have room for it XXX /2 ?
462 if (CONN_CANFRCV(so) &&
463 (so->so_snd.sb_cc < (so->so_snd.sb_datalen/2))) {
464 events |= G_IO_IN | G_IO_HUP | G_IO_ERR | G_IO_PRI;
467 if (events) {
468 GPollFD pfd = {
469 .fd = so->s,
470 .events = events,
472 so->pollfds_idx = pollfds->len;
473 g_array_append_val(pollfds, pfd);
478 * UDP sockets
480 for (so = slirp->udb.so_next; so != &slirp->udb;
481 so = so_next) {
482 so_next = so->so_next;
484 so->pollfds_idx = -1;
487 * See if it's timed out
489 if (so->so_expire) {
490 if (so->so_expire <= curtime) {
491 udp_detach(so);
492 continue;
493 } else {
494 slirp->do_slowtimo = true; /* Let socket expire */
499 * When UDP packets are received from over the
500 * link, they're sendto()'d straight away, so
501 * no need for setting for writing
502 * Limit the number of packets queued by this session
503 * to 4. Note that even though we try and limit this
504 * to 4 packets, the session could have more queued
505 * if the packets needed to be fragmented
506 * (XXX <= 4 ?)
508 if ((so->so_state & SS_ISFCONNECTED) && so->so_queued <= 4) {
509 GPollFD pfd = {
510 .fd = so->s,
511 .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
513 so->pollfds_idx = pollfds->len;
514 g_array_append_val(pollfds, pfd);
519 * ICMP sockets
521 for (so = slirp->icmp.so_next; so != &slirp->icmp;
522 so = so_next) {
523 so_next = so->so_next;
525 so->pollfds_idx = -1;
528 * See if it's timed out
530 if (so->so_expire) {
531 if (so->so_expire <= curtime) {
532 icmp_detach(so);
533 continue;
534 } else {
535 slirp->do_slowtimo = true; /* Let socket expire */
539 if (so->so_state & SS_ISFCONNECTED) {
540 GPollFD pfd = {
541 .fd = so->s,
542 .events = G_IO_IN | G_IO_HUP | G_IO_ERR,
544 so->pollfds_idx = pollfds->len;
545 g_array_append_val(pollfds, pfd);
549 slirp_update_timeout(timeout);
552 void slirp_pollfds_poll(GArray *pollfds, int select_error)
554 Slirp *slirp;
555 struct socket *so, *so_next;
556 int ret;
558 if (QTAILQ_EMPTY(&slirp_instances)) {
559 return;
562 curtime = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
564 QTAILQ_FOREACH(slirp, &slirp_instances, entry) {
566 * See if anything has timed out
568 if (slirp->time_fasttimo &&
569 ((curtime - slirp->time_fasttimo) >= TIMEOUT_FAST)) {
570 tcp_fasttimo(slirp);
571 slirp->time_fasttimo = 0;
573 if (slirp->do_slowtimo &&
574 ((curtime - slirp->last_slowtimo) >= TIMEOUT_SLOW)) {
575 ip_slowtimo(slirp);
576 tcp_slowtimo(slirp);
577 slirp->last_slowtimo = curtime;
581 * Check sockets
583 if (!select_error) {
585 * Check TCP sockets
587 for (so = slirp->tcb.so_next; so != &slirp->tcb;
588 so = so_next) {
589 int revents;
591 so_next = so->so_next;
593 revents = 0;
594 if (so->pollfds_idx != -1) {
595 revents = g_array_index(pollfds, GPollFD,
596 so->pollfds_idx).revents;
599 if (so->so_state & SS_NOFDREF || so->s == -1) {
600 continue;
604 * Check for URG data
605 * This will soread as well, so no need to
606 * test for G_IO_IN below if this succeeds
608 if (revents & G_IO_PRI) {
609 ret = sorecvoob(so);
610 if (ret < 0) {
611 /* Socket error might have resulted in the socket being
612 * removed, do not try to do anything more with it. */
613 continue;
617 * Check sockets for reading
619 else if (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR)) {
621 * Check for incoming connections
623 if (so->so_state & SS_FACCEPTCONN) {
624 tcp_connect(so);
625 continue;
626 } /* else */
627 ret = soread(so);
629 /* Output it if we read something */
630 if (ret > 0) {
631 tcp_output(sototcpcb(so));
633 if (ret < 0) {
634 /* Socket error might have resulted in the socket being
635 * removed, do not try to do anything more with it. */
636 continue;
641 * Check sockets for writing
643 if (!(so->so_state & SS_NOFDREF) &&
644 (revents & (G_IO_OUT | G_IO_ERR))) {
646 * Check for non-blocking, still-connecting sockets
648 if (so->so_state & SS_ISFCONNECTING) {
649 /* Connected */
650 so->so_state &= ~SS_ISFCONNECTING;
652 ret = send(so->s, (const void *) &ret, 0, 0);
653 if (ret < 0) {
654 /* XXXXX Must fix, zero bytes is a NOP */
655 if (errno == EAGAIN || errno == EWOULDBLOCK ||
656 errno == EINPROGRESS || errno == ENOTCONN) {
657 continue;
660 /* else failed */
661 so->so_state &= SS_PERSISTENT_MASK;
662 so->so_state |= SS_NOFDREF;
664 /* else so->so_state &= ~SS_ISFCONNECTING; */
667 * Continue tcp_input
669 tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
670 so->so_ffamily);
671 /* continue; */
672 } else {
673 ret = sowrite(so);
676 * XXXXX If we wrote something (a lot), there
677 * could be a need for a window update.
678 * In the worst case, the remote will send
679 * a window probe to get things going again
684 * Probe a still-connecting, non-blocking socket
685 * to check if it's still alive
687 #ifdef PROBE_CONN
688 if (so->so_state & SS_ISFCONNECTING) {
689 ret = qemu_recv(so->s, &ret, 0, 0);
691 if (ret < 0) {
692 /* XXX */
693 if (errno == EAGAIN || errno == EWOULDBLOCK ||
694 errno == EINPROGRESS || errno == ENOTCONN) {
695 continue; /* Still connecting, continue */
698 /* else failed */
699 so->so_state &= SS_PERSISTENT_MASK;
700 so->so_state |= SS_NOFDREF;
702 /* tcp_input will take care of it */
703 } else {
704 ret = send(so->s, &ret, 0, 0);
705 if (ret < 0) {
706 /* XXX */
707 if (errno == EAGAIN || errno == EWOULDBLOCK ||
708 errno == EINPROGRESS || errno == ENOTCONN) {
709 continue;
711 /* else failed */
712 so->so_state &= SS_PERSISTENT_MASK;
713 so->so_state |= SS_NOFDREF;
714 } else {
715 so->so_state &= ~SS_ISFCONNECTING;
719 tcp_input((struct mbuf *)NULL, sizeof(struct ip), so,
720 so->so_ffamily);
721 } /* SS_ISFCONNECTING */
722 #endif
726 * Now UDP sockets.
727 * Incoming packets are sent straight away, they're not buffered.
728 * Incoming UDP data isn't buffered either.
730 for (so = slirp->udb.so_next; so != &slirp->udb;
731 so = so_next) {
732 int revents;
734 so_next = so->so_next;
736 revents = 0;
737 if (so->pollfds_idx != -1) {
738 revents = g_array_index(pollfds, GPollFD,
739 so->pollfds_idx).revents;
742 if (so->s != -1 &&
743 (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
744 sorecvfrom(so);
749 * Check incoming ICMP relies.
751 for (so = slirp->icmp.so_next; so != &slirp->icmp;
752 so = so_next) {
753 int revents;
755 so_next = so->so_next;
757 revents = 0;
758 if (so->pollfds_idx != -1) {
759 revents = g_array_index(pollfds, GPollFD,
760 so->pollfds_idx).revents;
763 if (so->s != -1 &&
764 (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR))) {
765 icmp_receive(so);
770 if_start(slirp);
774 static void arp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
776 struct slirp_arphdr *ah = (struct slirp_arphdr *)(pkt + ETH_HLEN);
777 uint8_t arp_reply[MAX(ETH_HLEN + sizeof(struct slirp_arphdr), 64)];
778 struct ethhdr *reh = (struct ethhdr *)arp_reply;
779 struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_reply + ETH_HLEN);
780 int ar_op;
781 struct ex_list *ex_ptr;
783 if (!slirp->in_enabled) {
784 return;
787 ar_op = ntohs(ah->ar_op);
788 switch(ar_op) {
789 case ARPOP_REQUEST:
790 if (ah->ar_tip == ah->ar_sip) {
791 /* Gratuitous ARP */
792 arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
793 return;
796 if ((ah->ar_tip & slirp->vnetwork_mask.s_addr) ==
797 slirp->vnetwork_addr.s_addr) {
798 if (ah->ar_tip == slirp->vnameserver_addr.s_addr ||
799 ah->ar_tip == slirp->vhost_addr.s_addr)
800 goto arp_ok;
801 for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
802 if (ex_ptr->ex_addr.s_addr == ah->ar_tip)
803 goto arp_ok;
805 return;
806 arp_ok:
807 memset(arp_reply, 0, sizeof(arp_reply));
809 arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
811 /* ARP request for alias/dns mac address */
812 memcpy(reh->h_dest, pkt + ETH_ALEN, ETH_ALEN);
813 memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
814 memcpy(&reh->h_source[2], &ah->ar_tip, 4);
815 reh->h_proto = htons(ETH_P_ARP);
817 rah->ar_hrd = htons(1);
818 rah->ar_pro = htons(ETH_P_IP);
819 rah->ar_hln = ETH_ALEN;
820 rah->ar_pln = 4;
821 rah->ar_op = htons(ARPOP_REPLY);
822 memcpy(rah->ar_sha, reh->h_source, ETH_ALEN);
823 rah->ar_sip = ah->ar_tip;
824 memcpy(rah->ar_tha, ah->ar_sha, ETH_ALEN);
825 rah->ar_tip = ah->ar_sip;
826 slirp_output(slirp->opaque, arp_reply, sizeof(arp_reply));
828 break;
829 case ARPOP_REPLY:
830 arp_table_add(slirp, ah->ar_sip, ah->ar_sha);
831 break;
832 default:
833 break;
837 void slirp_input(Slirp *slirp, const uint8_t *pkt, int pkt_len)
839 struct mbuf *m;
840 int proto;
842 if (pkt_len < ETH_HLEN)
843 return;
845 proto = ntohs(*(uint16_t *)(pkt + 12));
846 switch(proto) {
847 case ETH_P_ARP:
848 arp_input(slirp, pkt, pkt_len);
849 break;
850 case ETH_P_IP:
851 case ETH_P_IPV6:
852 m = m_get(slirp);
853 if (!m)
854 return;
855 /* Note: we add 2 to align the IP header on 4 bytes,
856 * and add the margin for the tcpiphdr overhead */
857 if (M_FREEROOM(m) < pkt_len + TCPIPHDR_DELTA + 2) {
858 m_inc(m, pkt_len + TCPIPHDR_DELTA + 2);
860 m->m_len = pkt_len + TCPIPHDR_DELTA + 2;
861 memcpy(m->m_data + TCPIPHDR_DELTA + 2, pkt, pkt_len);
863 m->m_data += TCPIPHDR_DELTA + 2 + ETH_HLEN;
864 m->m_len -= TCPIPHDR_DELTA + 2 + ETH_HLEN;
866 if (proto == ETH_P_IP) {
867 ip_input(m);
868 } else if (proto == ETH_P_IPV6) {
869 ip6_input(m);
871 break;
873 case ETH_P_NCSI:
874 ncsi_input(slirp, pkt, pkt_len);
875 break;
877 default:
878 break;
882 /* Prepare the IPv4 packet to be sent to the ethernet device. Returns 1 if no
883 * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
884 * is ready to go.
886 static int if_encap4(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
887 uint8_t ethaddr[ETH_ALEN])
889 const struct ip *iph = (const struct ip *)ifm->m_data;
891 if (iph->ip_dst.s_addr == 0) {
892 /* 0.0.0.0 can not be a destination address, something went wrong,
893 * avoid making it worse */
894 return 1;
896 if (!arp_table_search(slirp, iph->ip_dst.s_addr, ethaddr)) {
897 uint8_t arp_req[ETH_HLEN + sizeof(struct slirp_arphdr)];
898 struct ethhdr *reh = (struct ethhdr *)arp_req;
899 struct slirp_arphdr *rah = (struct slirp_arphdr *)(arp_req + ETH_HLEN);
901 if (!ifm->resolution_requested) {
902 /* If the client addr is not known, send an ARP request */
903 memset(reh->h_dest, 0xff, ETH_ALEN);
904 memcpy(reh->h_source, special_ethaddr, ETH_ALEN - 4);
905 memcpy(&reh->h_source[2], &slirp->vhost_addr, 4);
906 reh->h_proto = htons(ETH_P_ARP);
907 rah->ar_hrd = htons(1);
908 rah->ar_pro = htons(ETH_P_IP);
909 rah->ar_hln = ETH_ALEN;
910 rah->ar_pln = 4;
911 rah->ar_op = htons(ARPOP_REQUEST);
913 /* source hw addr */
914 memcpy(rah->ar_sha, special_ethaddr, ETH_ALEN - 4);
915 memcpy(&rah->ar_sha[2], &slirp->vhost_addr, 4);
917 /* source IP */
918 rah->ar_sip = slirp->vhost_addr.s_addr;
920 /* target hw addr (none) */
921 memset(rah->ar_tha, 0, ETH_ALEN);
923 /* target IP */
924 rah->ar_tip = iph->ip_dst.s_addr;
925 slirp->client_ipaddr = iph->ip_dst;
926 slirp_output(slirp->opaque, arp_req, sizeof(arp_req));
927 ifm->resolution_requested = true;
929 /* Expire request and drop outgoing packet after 1 second */
930 ifm->expiration_date = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + 1000000000ULL;
932 return 0;
933 } else {
934 memcpy(eh->h_source, special_ethaddr, ETH_ALEN - 4);
935 /* XXX: not correct */
936 memcpy(&eh->h_source[2], &slirp->vhost_addr, 4);
937 eh->h_proto = htons(ETH_P_IP);
939 /* Send this */
940 return 2;
944 /* Prepare the IPv6 packet to be sent to the ethernet device. Returns 1 if no
945 * packet should be sent, 0 if the packet must be re-queued, 2 if the packet
946 * is ready to go.
948 static int if_encap6(Slirp *slirp, struct mbuf *ifm, struct ethhdr *eh,
949 uint8_t ethaddr[ETH_ALEN])
951 const struct ip6 *ip6h = mtod(ifm, const struct ip6 *);
952 if (!ndp_table_search(slirp, ip6h->ip_dst, ethaddr)) {
953 if (!ifm->resolution_requested) {
954 ndp_send_ns(slirp, ip6h->ip_dst);
955 ifm->resolution_requested = true;
956 ifm->expiration_date =
957 qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + 1000000000ULL;
959 return 0;
960 } else {
961 eh->h_proto = htons(ETH_P_IPV6);
962 in6_compute_ethaddr(ip6h->ip_src, eh->h_source);
964 /* Send this */
965 return 2;
969 /* Output the IP packet to the ethernet device. Returns 0 if the packet must be
970 * re-queued.
972 int if_encap(Slirp *slirp, struct mbuf *ifm)
974 uint8_t buf[1600];
975 struct ethhdr *eh = (struct ethhdr *)buf;
976 uint8_t ethaddr[ETH_ALEN];
977 const struct ip *iph = (const struct ip *)ifm->m_data;
978 int ret;
980 if (ifm->m_len + ETH_HLEN > sizeof(buf)) {
981 return 1;
984 switch (iph->ip_v) {
985 case IPVERSION:
986 ret = if_encap4(slirp, ifm, eh, ethaddr);
987 if (ret < 2) {
988 return ret;
990 break;
992 case IP6VERSION:
993 ret = if_encap6(slirp, ifm, eh, ethaddr);
994 if (ret < 2) {
995 return ret;
997 break;
999 default:
1000 g_assert_not_reached();
1001 break;
1004 memcpy(eh->h_dest, ethaddr, ETH_ALEN);
1005 DEBUG_ARGS((dfd, " src = %02x:%02x:%02x:%02x:%02x:%02x\n",
1006 eh->h_source[0], eh->h_source[1], eh->h_source[2],
1007 eh->h_source[3], eh->h_source[4], eh->h_source[5]));
1008 DEBUG_ARGS((dfd, " dst = %02x:%02x:%02x:%02x:%02x:%02x\n",
1009 eh->h_dest[0], eh->h_dest[1], eh->h_dest[2],
1010 eh->h_dest[3], eh->h_dest[4], eh->h_dest[5]));
1011 memcpy(buf + sizeof(struct ethhdr), ifm->m_data, ifm->m_len);
1012 slirp_output(slirp->opaque, buf, ifm->m_len + ETH_HLEN);
1013 return 1;
1016 /* Drop host forwarding rule, return 0 if found. */
1017 int slirp_remove_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
1018 int host_port)
1020 struct socket *so;
1021 struct socket *head = (is_udp ? &slirp->udb : &slirp->tcb);
1022 struct sockaddr_in addr;
1023 int port = htons(host_port);
1024 socklen_t addr_len;
1026 for (so = head->so_next; so != head; so = so->so_next) {
1027 addr_len = sizeof(addr);
1028 if ((so->so_state & SS_HOSTFWD) &&
1029 getsockname(so->s, (struct sockaddr *)&addr, &addr_len) == 0 &&
1030 addr.sin_addr.s_addr == host_addr.s_addr &&
1031 addr.sin_port == port) {
1032 close(so->s);
1033 sofree(so);
1034 return 0;
1038 return -1;
1041 int slirp_add_hostfwd(Slirp *slirp, int is_udp, struct in_addr host_addr,
1042 int host_port, struct in_addr guest_addr, int guest_port)
1044 if (!guest_addr.s_addr) {
1045 guest_addr = slirp->vdhcp_startaddr;
1047 if (is_udp) {
1048 if (!udp_listen(slirp, host_addr.s_addr, htons(host_port),
1049 guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
1050 return -1;
1051 } else {
1052 if (!tcp_listen(slirp, host_addr.s_addr, htons(host_port),
1053 guest_addr.s_addr, htons(guest_port), SS_HOSTFWD))
1054 return -1;
1056 return 0;
1059 int slirp_add_exec(Slirp *slirp, int do_pty, const void *args,
1060 struct in_addr *guest_addr, int guest_port)
1062 if (!guest_addr->s_addr) {
1063 guest_addr->s_addr = slirp->vnetwork_addr.s_addr |
1064 (htonl(0x0204) & ~slirp->vnetwork_mask.s_addr);
1066 if ((guest_addr->s_addr & slirp->vnetwork_mask.s_addr) !=
1067 slirp->vnetwork_addr.s_addr ||
1068 guest_addr->s_addr == slirp->vhost_addr.s_addr ||
1069 guest_addr->s_addr == slirp->vnameserver_addr.s_addr) {
1070 return -1;
1072 return add_exec(&slirp->exec_list, do_pty, (char *)args, *guest_addr,
1073 htons(guest_port));
1076 ssize_t slirp_send(struct socket *so, const void *buf, size_t len, int flags)
1078 if (so->s == -1 && so->extra) {
1079 /* XXX this blocks entire thread. Rewrite to use
1080 * qemu_chr_fe_write and background I/O callbacks */
1081 qemu_chr_fe_write_all(so->extra, buf, len);
1082 return len;
1085 return send(so->s, buf, len, flags);
1088 static struct socket *
1089 slirp_find_ctl_socket(Slirp *slirp, struct in_addr guest_addr, int guest_port)
1091 struct socket *so;
1093 for (so = slirp->tcb.so_next; so != &slirp->tcb; so = so->so_next) {
1094 if (so->so_faddr.s_addr == guest_addr.s_addr &&
1095 htons(so->so_fport) == guest_port) {
1096 return so;
1099 return NULL;
1102 size_t slirp_socket_can_recv(Slirp *slirp, struct in_addr guest_addr,
1103 int guest_port)
1105 struct iovec iov[2];
1106 struct socket *so;
1108 so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1110 if (!so || so->so_state & SS_NOFDREF) {
1111 return 0;
1114 if (!CONN_CANFRCV(so) || so->so_snd.sb_cc >= (so->so_snd.sb_datalen/2)) {
1115 return 0;
1118 return sopreprbuf(so, iov, NULL);
1121 void slirp_socket_recv(Slirp *slirp, struct in_addr guest_addr, int guest_port,
1122 const uint8_t *buf, int size)
1124 int ret;
1125 struct socket *so = slirp_find_ctl_socket(slirp, guest_addr, guest_port);
1127 if (!so)
1128 return;
1130 ret = soreadbuf(so, (const char *)buf, size);
1132 if (ret > 0)
1133 tcp_output(sototcpcb(so));
1136 static void slirp_tcp_save(QEMUFile *f, struct tcpcb *tp)
1138 int i;
1140 qemu_put_sbe16(f, tp->t_state);
1141 for (i = 0; i < TCPT_NTIMERS; i++)
1142 qemu_put_sbe16(f, tp->t_timer[i]);
1143 qemu_put_sbe16(f, tp->t_rxtshift);
1144 qemu_put_sbe16(f, tp->t_rxtcur);
1145 qemu_put_sbe16(f, tp->t_dupacks);
1146 qemu_put_be16(f, tp->t_maxseg);
1147 qemu_put_sbyte(f, tp->t_force);
1148 qemu_put_be16(f, tp->t_flags);
1149 qemu_put_be32(f, tp->snd_una);
1150 qemu_put_be32(f, tp->snd_nxt);
1151 qemu_put_be32(f, tp->snd_up);
1152 qemu_put_be32(f, tp->snd_wl1);
1153 qemu_put_be32(f, tp->snd_wl2);
1154 qemu_put_be32(f, tp->iss);
1155 qemu_put_be32(f, tp->snd_wnd);
1156 qemu_put_be32(f, tp->rcv_wnd);
1157 qemu_put_be32(f, tp->rcv_nxt);
1158 qemu_put_be32(f, tp->rcv_up);
1159 qemu_put_be32(f, tp->irs);
1160 qemu_put_be32(f, tp->rcv_adv);
1161 qemu_put_be32(f, tp->snd_max);
1162 qemu_put_be32(f, tp->snd_cwnd);
1163 qemu_put_be32(f, tp->snd_ssthresh);
1164 qemu_put_sbe16(f, tp->t_idle);
1165 qemu_put_sbe16(f, tp->t_rtt);
1166 qemu_put_be32(f, tp->t_rtseq);
1167 qemu_put_sbe16(f, tp->t_srtt);
1168 qemu_put_sbe16(f, tp->t_rttvar);
1169 qemu_put_be16(f, tp->t_rttmin);
1170 qemu_put_be32(f, tp->max_sndwnd);
1171 qemu_put_byte(f, tp->t_oobflags);
1172 qemu_put_byte(f, tp->t_iobc);
1173 qemu_put_sbe16(f, tp->t_softerror);
1174 qemu_put_byte(f, tp->snd_scale);
1175 qemu_put_byte(f, tp->rcv_scale);
1176 qemu_put_byte(f, tp->request_r_scale);
1177 qemu_put_byte(f, tp->requested_s_scale);
1178 qemu_put_be32(f, tp->ts_recent);
1179 qemu_put_be32(f, tp->ts_recent_age);
1180 qemu_put_be32(f, tp->last_ack_sent);
1183 static void slirp_sbuf_save(QEMUFile *f, struct sbuf *sbuf)
1185 uint32_t off;
1187 qemu_put_be32(f, sbuf->sb_cc);
1188 qemu_put_be32(f, sbuf->sb_datalen);
1189 off = (uint32_t)(sbuf->sb_wptr - sbuf->sb_data);
1190 qemu_put_sbe32(f, off);
1191 off = (uint32_t)(sbuf->sb_rptr - sbuf->sb_data);
1192 qemu_put_sbe32(f, off);
1193 qemu_put_buffer(f, (unsigned char*)sbuf->sb_data, sbuf->sb_datalen);
1196 static void slirp_socket_save(QEMUFile *f, struct socket *so)
1198 qemu_put_be32(f, so->so_urgc);
1199 qemu_put_be16(f, so->so_ffamily);
1200 switch (so->so_ffamily) {
1201 case AF_INET:
1202 qemu_put_be32(f, so->so_faddr.s_addr);
1203 qemu_put_be16(f, so->so_fport);
1204 break;
1205 default:
1206 error_report("so_ffamily unknown, unable to save so_faddr and"
1207 " so_fport");
1209 qemu_put_be16(f, so->so_lfamily);
1210 switch (so->so_lfamily) {
1211 case AF_INET:
1212 qemu_put_be32(f, so->so_laddr.s_addr);
1213 qemu_put_be16(f, so->so_lport);
1214 break;
1215 default:
1216 error_report("so_ffamily unknown, unable to save so_laddr and"
1217 " so_lport");
1219 qemu_put_byte(f, so->so_iptos);
1220 qemu_put_byte(f, so->so_emu);
1221 qemu_put_byte(f, so->so_type);
1222 qemu_put_be32(f, so->so_state);
1223 slirp_sbuf_save(f, &so->so_rcv);
1224 slirp_sbuf_save(f, &so->so_snd);
1225 slirp_tcp_save(f, so->so_tcpcb);
1228 static void slirp_bootp_save(QEMUFile *f, Slirp *slirp)
1230 int i;
1232 for (i = 0; i < NB_BOOTP_CLIENTS; i++) {
1233 qemu_put_be16(f, slirp->bootp_clients[i].allocated);
1234 qemu_put_buffer(f, slirp->bootp_clients[i].macaddr, 6);
1238 static void slirp_state_save(QEMUFile *f, void *opaque)
1240 Slirp *slirp = opaque;
1241 struct ex_list *ex_ptr;
1243 for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next)
1244 if (ex_ptr->ex_pty == 3) {
1245 struct socket *so;
1246 so = slirp_find_ctl_socket(slirp, ex_ptr->ex_addr,
1247 ntohs(ex_ptr->ex_fport));
1248 if (!so)
1249 continue;
1251 qemu_put_byte(f, 42);
1252 slirp_socket_save(f, so);
1254 qemu_put_byte(f, 0);
1256 qemu_put_be16(f, slirp->ip_id);
1258 slirp_bootp_save(f, slirp);
1261 static void slirp_tcp_load(QEMUFile *f, struct tcpcb *tp)
1263 int i;
1265 tp->t_state = qemu_get_sbe16(f);
1266 for (i = 0; i < TCPT_NTIMERS; i++)
1267 tp->t_timer[i] = qemu_get_sbe16(f);
1268 tp->t_rxtshift = qemu_get_sbe16(f);
1269 tp->t_rxtcur = qemu_get_sbe16(f);
1270 tp->t_dupacks = qemu_get_sbe16(f);
1271 tp->t_maxseg = qemu_get_be16(f);
1272 tp->t_force = qemu_get_sbyte(f);
1273 tp->t_flags = qemu_get_be16(f);
1274 tp->snd_una = qemu_get_be32(f);
1275 tp->snd_nxt = qemu_get_be32(f);
1276 tp->snd_up = qemu_get_be32(f);
1277 tp->snd_wl1 = qemu_get_be32(f);
1278 tp->snd_wl2 = qemu_get_be32(f);
1279 tp->iss = qemu_get_be32(f);
1280 tp->snd_wnd = qemu_get_be32(f);
1281 tp->rcv_wnd = qemu_get_be32(f);
1282 tp->rcv_nxt = qemu_get_be32(f);
1283 tp->rcv_up = qemu_get_be32(f);
1284 tp->irs = qemu_get_be32(f);
1285 tp->rcv_adv = qemu_get_be32(f);
1286 tp->snd_max = qemu_get_be32(f);
1287 tp->snd_cwnd = qemu_get_be32(f);
1288 tp->snd_ssthresh = qemu_get_be32(f);
1289 tp->t_idle = qemu_get_sbe16(f);
1290 tp->t_rtt = qemu_get_sbe16(f);
1291 tp->t_rtseq = qemu_get_be32(f);
1292 tp->t_srtt = qemu_get_sbe16(f);
1293 tp->t_rttvar = qemu_get_sbe16(f);
1294 tp->t_rttmin = qemu_get_be16(f);
1295 tp->max_sndwnd = qemu_get_be32(f);
1296 tp->t_oobflags = qemu_get_byte(f);
1297 tp->t_iobc = qemu_get_byte(f);
1298 tp->t_softerror = qemu_get_sbe16(f);
1299 tp->snd_scale = qemu_get_byte(f);
1300 tp->rcv_scale = qemu_get_byte(f);
1301 tp->request_r_scale = qemu_get_byte(f);
1302 tp->requested_s_scale = qemu_get_byte(f);
1303 tp->ts_recent = qemu_get_be32(f);
1304 tp->ts_recent_age = qemu_get_be32(f);
1305 tp->last_ack_sent = qemu_get_be32(f);
1306 tcp_template(tp);
1309 static int slirp_sbuf_load(QEMUFile *f, struct sbuf *sbuf)
1311 uint32_t off, sb_cc, sb_datalen;
1313 sb_cc = qemu_get_be32(f);
1314 sb_datalen = qemu_get_be32(f);
1316 sbreserve(sbuf, sb_datalen);
1318 if (sbuf->sb_datalen != sb_datalen)
1319 return -ENOMEM;
1321 sbuf->sb_cc = sb_cc;
1323 off = qemu_get_sbe32(f);
1324 sbuf->sb_wptr = sbuf->sb_data + off;
1325 off = qemu_get_sbe32(f);
1326 sbuf->sb_rptr = sbuf->sb_data + off;
1327 qemu_get_buffer(f, (unsigned char*)sbuf->sb_data, sbuf->sb_datalen);
1329 return 0;
1332 static int slirp_socket_load(QEMUFile *f, struct socket *so, int version_id)
1334 if (tcp_attach(so) < 0)
1335 return -ENOMEM;
1337 so->so_urgc = qemu_get_be32(f);
1338 if (version_id <= 3) {
1339 so->so_ffamily = AF_INET;
1340 so->so_faddr.s_addr = qemu_get_be32(f);
1341 so->so_laddr.s_addr = qemu_get_be32(f);
1342 so->so_fport = qemu_get_be16(f);
1343 so->so_lport = qemu_get_be16(f);
1344 } else {
1345 so->so_ffamily = qemu_get_be16(f);
1346 switch (so->so_ffamily) {
1347 case AF_INET:
1348 so->so_faddr.s_addr = qemu_get_be32(f);
1349 so->so_fport = qemu_get_be16(f);
1350 break;
1351 default:
1352 error_report(
1353 "so_ffamily unknown, unable to restore so_faddr and so_lport");
1355 so->so_lfamily = qemu_get_be16(f);
1356 switch (so->so_lfamily) {
1357 case AF_INET:
1358 so->so_laddr.s_addr = qemu_get_be32(f);
1359 so->so_lport = qemu_get_be16(f);
1360 break;
1361 default:
1362 error_report(
1363 "so_ffamily unknown, unable to restore so_laddr and so_lport");
1366 so->so_iptos = qemu_get_byte(f);
1367 so->so_emu = qemu_get_byte(f);
1368 so->so_type = qemu_get_byte(f);
1369 so->so_state = qemu_get_be32(f);
1370 if (slirp_sbuf_load(f, &so->so_rcv) < 0)
1371 return -ENOMEM;
1372 if (slirp_sbuf_load(f, &so->so_snd) < 0)
1373 return -ENOMEM;
1374 slirp_tcp_load(f, so->so_tcpcb);
1376 return 0;
1379 static void slirp_bootp_load(QEMUFile *f, Slirp *slirp)
1381 int i;
1383 for (i = 0; i < NB_BOOTP_CLIENTS; i++) {
1384 slirp->bootp_clients[i].allocated = qemu_get_be16(f);
1385 qemu_get_buffer(f, slirp->bootp_clients[i].macaddr, 6);
1389 static int slirp_state_load(QEMUFile *f, void *opaque, int version_id)
1391 Slirp *slirp = opaque;
1392 struct ex_list *ex_ptr;
1394 while (qemu_get_byte(f)) {
1395 int ret;
1396 struct socket *so = socreate(slirp);
1398 if (!so)
1399 return -ENOMEM;
1401 ret = slirp_socket_load(f, so, version_id);
1403 if (ret < 0)
1404 return ret;
1406 if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) !=
1407 slirp->vnetwork_addr.s_addr) {
1408 return -EINVAL;
1410 for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
1411 if (ex_ptr->ex_pty == 3 &&
1412 so->so_faddr.s_addr == ex_ptr->ex_addr.s_addr &&
1413 so->so_fport == ex_ptr->ex_fport) {
1414 break;
1417 if (!ex_ptr)
1418 return -EINVAL;
1420 so->extra = (void *)ex_ptr->ex_exec;
1423 if (version_id >= 2) {
1424 slirp->ip_id = qemu_get_be16(f);
1427 if (version_id >= 3) {
1428 slirp_bootp_load(f, slirp);
1431 return 0;