2 * Copyright (C) 1999-2000 by Maksim Yevmenkin <m_evmenkin@yahoo.com>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * -------------------------------------------------------------------------
29 * Copyright (c) 1988, Julian Onions <jpo@cs.nott.ac.uk>
30 * Nottingham University 1987.
35 * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $
38 #include "opt_compat.h"
41 #include <sys/param.h>
43 #include <sys/fcntl.h>
44 #include <sys/filio.h>
46 #include <sys/kernel.h>
47 #include <sys/malloc.h>
49 #include <sys/module.h>
53 #include <sys/selinfo.h>
54 #include <sys/signalvar.h>
55 #include <sys/socket.h>
56 #include <sys/sockio.h>
57 #include <sys/sysctl.h>
58 #include <sys/systm.h>
59 #include <sys/ttycom.h>
61 #include <sys/queue.h>
64 #include <net/ethernet.h>
66 #include <net/if_clone.h>
67 #include <net/if_dl.h>
68 #include <net/if_media.h>
69 #include <net/if_types.h>
70 #include <net/route.h>
73 #include <netinet/in.h>
75 #include <net/if_tapvar.h>
76 #include <net/if_tap.h>
79 #define CDEV_NAME "tap"
80 #define TAPDEBUG if (tapdebug) printf
82 static const char tapname[] = "tap";
83 static const char vmnetname[] = "vmnet";
84 #define TAPMAXUNIT 0x7fff
85 #define VMNET_DEV_MASK CLONE_FLAG0
88 static int tapmodevent(module_t, int, void *);
91 static void tapclone(void *, struct ucred *, char *, int,
93 static void tapcreate(struct cdev *);
95 /* network interface */
96 static void tapifstart(struct ifnet *);
97 static int tapifioctl(struct ifnet *, u_long, caddr_t);
98 static void tapifinit(void *);
100 static int tap_clone_create(struct if_clone *, int, caddr_t);
101 static void tap_clone_destroy(struct ifnet *);
102 static struct if_clone *tap_cloner;
103 static int vmnet_clone_create(struct if_clone *, int, caddr_t);
104 static void vmnet_clone_destroy(struct ifnet *);
105 static struct if_clone *vmnet_cloner;
107 /* character device */
108 static d_open_t tapopen;
109 static d_close_t tapclose;
110 static d_read_t tapread;
111 static d_write_t tapwrite;
112 static d_ioctl_t tapioctl;
113 static d_poll_t tappoll;
114 static d_kqfilter_t tapkqfilter;
117 static int tapkqread(struct knote *, long);
118 static int tapkqwrite(struct knote *, long);
119 static void tapkqdetach(struct knote *);
121 static struct filterops tap_read_filterops = {
124 .f_detach = tapkqdetach,
125 .f_event = tapkqread,
128 static struct filterops tap_write_filterops = {
131 .f_detach = tapkqdetach,
132 .f_event = tapkqwrite,
135 static struct cdevsw tap_cdevsw = {
136 .d_version = D_VERSION,
137 .d_flags = D_NEEDMINOR,
145 .d_kqfilter = tapkqfilter,
149 * All global variables in if_tap.c are locked with tapmtx, with the
150 * exception of tapdebug, which is accessed unlocked; tapclones is
153 static struct mtx tapmtx;
154 static int tapdebug = 0; /* debug flag */
155 static int tapuopen = 0; /* allow user open() */
156 static int tapuponopen = 0; /* IFF_UP on open() */
157 static int tapdclone = 1; /* enable devfs cloning */
158 static SLIST_HEAD(, tap_softc) taphead; /* first device */
159 static struct clonedevs *tapclones;
161 MALLOC_DECLARE(M_TAP);
162 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
163 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
165 SYSCTL_DECL(_net_link);
166 static SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
167 "Ethernet tunnel software network interface");
168 SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
169 "Allow user to open /dev/tap (based on node permissions)");
170 SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0,
171 "Bring interface up when /dev/tap is opened");
172 SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tapdclone, 0,
173 "Enably legacy devfs interface creation");
174 SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
176 TUNABLE_INT("net.link.tap.devfs_cloning", &tapdclone);
178 DEV_MODULE(if_tap, tapmodevent, NULL);
181 tap_clone_create(struct if_clone *ifc, int unit, caddr_t params)
186 /* Find any existing device, or allocate new unit number. */
187 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, 0);
189 dev = make_dev(&tap_cdevsw, unit, UID_ROOT, GID_WHEEL, 0600,
190 "%s%d", tapname, unit);
197 /* vmnet devices are tap devices in disguise */
199 vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params)
204 /* Find any existing device, or allocate new unit number. */
205 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, VMNET_DEV_MASK);
207 dev = make_dev(&tap_cdevsw, unit | VMNET_DEV_MASK, UID_ROOT,
208 GID_WHEEL, 0600, "%s%d", vmnetname, unit);
216 tap_destroy(struct tap_softc *tp)
218 struct ifnet *ifp = tp->tap_ifp;
220 CURVNET_SET(ifp->if_vnet);
221 destroy_dev(tp->tap_dev);
222 seldrain(&tp->tap_rsel);
223 knlist_clear(&tp->tap_rsel.si_note, 0);
224 knlist_destroy(&tp->tap_rsel.si_note);
228 mtx_destroy(&tp->tap_mtx);
234 tap_clone_destroy(struct ifnet *ifp)
236 struct tap_softc *tp = ifp->if_softc;
239 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
244 /* vmnet devices are tap devices in disguise */
246 vmnet_clone_destroy(struct ifnet *ifp)
248 tap_clone_destroy(ifp);
254 * module event handler
257 tapmodevent(module_t mod, int type, void *data)
259 static eventhandler_tag eh_tag = NULL;
260 struct tap_softc *tp = NULL;
261 struct ifnet *ifp = NULL;
266 /* intitialize device */
268 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
269 SLIST_INIT(&taphead);
271 clone_setup(&tapclones);
272 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
273 if (eh_tag == NULL) {
274 clone_cleanup(&tapclones);
275 mtx_destroy(&tapmtx);
278 tap_cloner = if_clone_simple(tapname, tap_clone_create,
279 tap_clone_destroy, 0);
280 vmnet_cloner = if_clone_simple(vmnetname, vmnet_clone_create,
281 vmnet_clone_destroy, 0);
286 * The EBUSY algorithm here can't quite atomically
287 * guarantee that this is race-free since we have to
288 * release the tap mtx to deregister the clone handler.
291 SLIST_FOREACH(tp, &taphead, tap_next) {
292 mtx_lock(&tp->tap_mtx);
293 if (tp->tap_flags & TAP_OPEN) {
294 mtx_unlock(&tp->tap_mtx);
298 mtx_unlock(&tp->tap_mtx);
302 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
303 if_clone_detach(tap_cloner);
304 if_clone_detach(vmnet_cloner);
305 drain_dev_clone_events();
308 while ((tp = SLIST_FIRST(&taphead)) != NULL) {
309 SLIST_REMOVE_HEAD(&taphead, tap_next);
314 TAPDEBUG("detaching %s\n", ifp->if_xname);
320 clone_cleanup(&tapclones);
322 mtx_destroy(&tapmtx);
337 * We need to support two kind of devices - tap and vmnet
340 tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
342 char devname[SPECNAMELEN + 1];
343 int i, unit, append_unit;
350 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0))
357 /* We're interested in only tap/vmnet devices. */
358 if (strcmp(name, tapname) == 0) {
360 } else if (strcmp(name, vmnetname) == 0) {
362 extra = VMNET_DEV_MASK;
363 } else if (dev_stdclone(name, NULL, tapname, &unit) != 1) {
364 if (dev_stdclone(name, NULL, vmnetname, &unit) != 1) {
367 extra = VMNET_DEV_MASK;
374 CURVNET_SET(CRED_TO_VNET(cred));
375 /* find any existing device, or allocate new unit number */
376 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
380 * We were passed 'tun' or 'tap', with no unit specified
381 * so we'll need to append it now.
383 namelen = snprintf(devname, sizeof(devname), "%s%d", name,
388 *dev = make_dev_credf(MAKEDEV_REF, &tap_cdevsw, unit | extra,
389 cred, UID_ROOT, GID_WHEEL, 0600, "%s", name);
392 if_clone_create(name, namelen, NULL);
400 * to create interface
403 tapcreate(struct cdev *dev)
405 struct ifnet *ifp = NULL;
406 struct tap_softc *tp = NULL;
407 unsigned short macaddr_hi;
408 uint32_t macaddr_mid;
410 const char *name = NULL;
413 /* allocate driver storage and create device */
414 tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
415 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
417 SLIST_INSERT_HEAD(&taphead, tp, tap_next);
420 unit = dev2unit(dev);
422 /* select device: tap or vmnet */
423 if (unit & VMNET_DEV_MASK) {
425 tp->tap_flags |= TAP_VMNET;
431 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev));
433 /* generate fake MAC address: 00 bd xx xx xx unit_no */
434 macaddr_hi = htons(0x00bd);
435 macaddr_mid = (uint32_t) ticks;
436 bcopy(&macaddr_hi, eaddr, sizeof(short));
437 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t));
438 eaddr[5] = (u_char)unit;
440 /* fill the rest and attach interface */
441 ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
443 panic("%s%d: can not if_alloc()", name, unit);
445 if_initname(ifp, name, unit);
446 ifp->if_init = tapifinit;
447 ifp->if_start = tapifstart;
448 ifp->if_ioctl = tapifioctl;
449 ifp->if_mtu = ETHERMTU;
450 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
451 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
452 ifp->if_capabilities |= IFCAP_LINKSTATE;
453 ifp->if_capenable |= IFCAP_LINKSTATE;
458 ether_ifattach(ifp, eaddr);
460 mtx_lock(&tp->tap_mtx);
461 tp->tap_flags |= TAP_INITED;
462 mtx_unlock(&tp->tap_mtx);
464 knlist_init_mtx(&tp->tap_rsel.si_note, &tp->tap_mtx);
466 TAPDEBUG("interface %s is created. minor = %#x\n",
467 ifp->if_xname, dev2unit(dev));
474 * to open tunnel. must be superuser
477 tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
479 struct tap_softc *tp = NULL;
480 struct ifnet *ifp = NULL;
484 error = priv_check(td, PRIV_NET_TAP);
489 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
494 mtx_lock(&tp->tap_mtx);
495 if (tp->tap_flags & TAP_OPEN) {
496 mtx_unlock(&tp->tap_mtx);
500 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
501 tp->tap_pid = td->td_proc->p_pid;
502 tp->tap_flags |= TAP_OPEN;
505 ifp->if_drv_flags |= IFF_DRV_RUNNING;
506 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
508 ifp->if_flags |= IFF_UP;
509 if_link_state_change(ifp, LINK_STATE_UP);
510 mtx_unlock(&tp->tap_mtx);
512 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev));
521 * close the device - mark i/f down & delete routing info
524 tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
527 struct tap_softc *tp = dev->si_drv1;
528 struct ifnet *ifp = tp->tap_ifp;
530 /* junk all pending output */
531 mtx_lock(&tp->tap_mtx);
532 CURVNET_SET(ifp->if_vnet);
533 IF_DRAIN(&ifp->if_snd);
536 * do not bring the interface down, and do not anything with
537 * interface, if we are in VMnet mode. just close the device.
540 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
541 mtx_unlock(&tp->tap_mtx);
543 mtx_lock(&tp->tap_mtx);
544 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
545 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
546 mtx_unlock(&tp->tap_mtx);
547 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
548 rtinit(ifa, (int)RTM_DELETE, 0);
551 mtx_lock(&tp->tap_mtx);
555 if_link_state_change(ifp, LINK_STATE_DOWN);
558 funsetown(&tp->tap_sigio);
559 selwakeuppri(&tp->tap_rsel, PZERO+1);
560 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
562 tp->tap_flags &= ~TAP_OPEN;
564 mtx_unlock(&tp->tap_mtx);
566 TAPDEBUG("%s is closed. minor = %#x\n",
567 ifp->if_xname, dev2unit(dev));
576 * network interface initialization function
581 struct tap_softc *tp = (struct tap_softc *)xtp;
582 struct ifnet *ifp = tp->tap_ifp;
584 TAPDEBUG("initializing %s\n", ifp->if_xname);
586 mtx_lock(&tp->tap_mtx);
587 ifp->if_drv_flags |= IFF_DRV_RUNNING;
588 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
589 mtx_unlock(&tp->tap_mtx);
591 /* attempt to start output */
599 * Process an ioctl request on network interface
602 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
604 struct tap_softc *tp = ifp->if_softc;
605 struct ifreq *ifr = (struct ifreq *)data;
606 struct ifstat *ifs = NULL;
607 struct ifmediareq *ifmr = NULL;
608 int dummy, error = 0;
611 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
617 ifmr = (struct ifmediareq *)data;
618 dummy = ifmr->ifm_count;
620 ifmr->ifm_status = IFM_AVALID;
621 ifmr->ifm_active = IFM_ETHER;
622 if (tp->tap_flags & TAP_OPEN)
623 ifmr->ifm_status |= IFM_ACTIVE;
624 ifmr->ifm_current = ifmr->ifm_active;
626 int media = IFM_ETHER;
627 error = copyout(&media, ifmr->ifm_ulist,
633 ifp->if_mtu = ifr->ifr_mtu;
637 ifs = (struct ifstat *)data;
638 dummy = strlen(ifs->ascii);
639 mtx_lock(&tp->tap_mtx);
640 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii))
641 snprintf(ifs->ascii + dummy,
642 sizeof(ifs->ascii) - dummy,
643 "\tOpened by PID %d\n", tp->tap_pid);
644 mtx_unlock(&tp->tap_mtx);
648 error = ether_ioctl(ifp, cmd, data);
659 * queue packets from higher level ready to put out
662 tapifstart(struct ifnet *ifp)
664 struct tap_softc *tp = ifp->if_softc;
666 TAPDEBUG("%s starting\n", ifp->if_xname);
669 * do not junk pending output if we are in VMnet mode.
670 * XXX: can this do any harm because of queue overflow?
673 mtx_lock(&tp->tap_mtx);
674 if (((tp->tap_flags & TAP_VMNET) == 0) &&
675 ((tp->tap_flags & TAP_READY) != TAP_READY)) {
679 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
683 IF_DEQUEUE(&ifp->if_snd, m);
690 mtx_unlock(&tp->tap_mtx);
695 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
697 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
698 if (tp->tap_flags & TAP_RWAIT) {
699 tp->tap_flags &= ~TAP_RWAIT;
703 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
704 mtx_unlock(&tp->tap_mtx);
705 pgsigio(&tp->tap_sigio, SIGIO, 0);
706 mtx_lock(&tp->tap_mtx);
709 selwakeuppri(&tp->tap_rsel, PZERO+1);
710 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
711 ifp->if_opackets ++; /* obytes are counted in ether_output */
714 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
715 mtx_unlock(&tp->tap_mtx);
722 * the cdevsw interface is now pretty minimal
725 tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
727 struct tap_softc *tp = dev->si_drv1;
728 struct ifnet *ifp = tp->tap_ifp;
729 struct tapinfo *tapp = NULL;
731 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
732 defined(COMPAT_FREEBSD4)
738 tapp = (struct tapinfo *)data;
739 mtx_lock(&tp->tap_mtx);
740 ifp->if_mtu = tapp->mtu;
741 ifp->if_type = tapp->type;
742 ifp->if_baudrate = tapp->baudrate;
743 mtx_unlock(&tp->tap_mtx);
747 tapp = (struct tapinfo *)data;
748 mtx_lock(&tp->tap_mtx);
749 tapp->mtu = ifp->if_mtu;
750 tapp->type = ifp->if_type;
751 tapp->baudrate = ifp->if_baudrate;
752 mtx_unlock(&tp->tap_mtx);
756 tapdebug = *(int *)data;
760 *(int *)data = tapdebug;
764 struct ifreq *ifr = (struct ifreq *) data;
766 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
773 mtx_lock(&tp->tap_mtx);
775 tp->tap_flags |= TAP_ASYNC;
777 tp->tap_flags &= ~TAP_ASYNC;
778 mtx_unlock(&tp->tap_mtx);
782 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
785 IFQ_LOCK(&ifp->if_snd);
786 IFQ_POLL_NOLOCK(&ifp->if_snd, mb);
787 for (*(int *)data = 0; mb != NULL;
789 *(int *)data += mb->m_len;
790 IFQ_UNLOCK(&ifp->if_snd);
796 return (fsetown(*(int *)data, &tp->tap_sigio));
799 *(int *)data = fgetown(&tp->tap_sigio);
802 /* this is deprecated, FIOSETOWN should be used instead */
804 return (fsetown(-(*(int *)data), &tp->tap_sigio));
806 /* this is deprecated, FIOGETOWN should be used instead */
808 *(int *)data = -fgetown(&tp->tap_sigio);
811 /* VMware/VMnet port ioctl's */
813 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
814 defined(COMPAT_FREEBSD4)
816 ival = IOCPARM_IVAL(data);
817 data = (caddr_t)&ival;
820 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
823 f &= ~IFF_CANTCHANGE;
826 mtx_lock(&tp->tap_mtx);
827 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
828 mtx_unlock(&tp->tap_mtx);
831 case OSIOCGIFADDR: /* get MAC address of the remote side */
833 mtx_lock(&tp->tap_mtx);
834 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
835 mtx_unlock(&tp->tap_mtx);
838 case SIOCSIFADDR: /* set MAC address of the remote side */
839 mtx_lock(&tp->tap_mtx);
840 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
841 mtx_unlock(&tp->tap_mtx);
854 * the cdevsw read interface - reads a packet at a time, or at
855 * least as much of a packet as can be read
858 tapread(struct cdev *dev, struct uio *uio, int flag)
860 struct tap_softc *tp = dev->si_drv1;
861 struct ifnet *ifp = tp->tap_ifp;
862 struct mbuf *m = NULL;
865 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev));
867 mtx_lock(&tp->tap_mtx);
868 if ((tp->tap_flags & TAP_READY) != TAP_READY) {
869 mtx_unlock(&tp->tap_mtx);
872 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
873 ifp->if_xname, dev2unit(dev), tp->tap_flags);
878 tp->tap_flags &= ~TAP_RWAIT;
880 /* sleep until we get a packet */
882 IF_DEQUEUE(&ifp->if_snd, m);
885 if (flag & O_NONBLOCK) {
886 mtx_unlock(&tp->tap_mtx);
887 return (EWOULDBLOCK);
890 tp->tap_flags |= TAP_RWAIT;
891 error = mtx_sleep(tp, &tp->tap_mtx, PCATCH | (PZERO + 1),
894 mtx_unlock(&tp->tap_mtx);
899 mtx_unlock(&tp->tap_mtx);
901 /* feed packet to bpf */
904 /* xfer packet to user space */
905 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
906 len = min(uio->uio_resid, m->m_len);
910 error = uiomove(mtod(m, void *), len, uio);
915 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
927 * the cdevsw write interface - an atomic write is a packet - or else!
930 tapwrite(struct cdev *dev, struct uio *uio, int flag)
932 struct ether_header *eh;
933 struct tap_softc *tp = dev->si_drv1;
934 struct ifnet *ifp = tp->tap_ifp;
937 TAPDEBUG("%s writing, minor = %#x\n",
938 ifp->if_xname, dev2unit(dev));
940 if (uio->uio_resid == 0)
943 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
944 TAPDEBUG("%s invalid packet len = %zd, minor = %#x\n",
945 ifp->if_xname, uio->uio_resid, dev2unit(dev));
950 if ((m = m_uiotombuf(uio, M_NOWAIT, 0, ETHER_ALIGN,
951 M_PKTHDR)) == NULL) {
956 m->m_pkthdr.rcvif = ifp;
959 * Only pass a unicast frame to ether_input(), if it would actually
960 * have been received by non-virtual hardware.
962 if (m->m_len < sizeof(struct ether_header)) {
966 eh = mtod(m, struct ether_header *);
968 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
969 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
970 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
975 /* Pass packet up to parent. */
976 CURVNET_SET(ifp->if_vnet);
977 (*ifp->if_input)(ifp, m);
979 ifp->if_ipackets ++; /* ibytes are counted in parent */
988 * the poll interface, this is only useful on reads
989 * really. the write detect always returns true, write never blocks
990 * anyway, it either accepts the packet or drops it
993 tappoll(struct cdev *dev, int events, struct thread *td)
995 struct tap_softc *tp = dev->si_drv1;
996 struct ifnet *ifp = tp->tap_ifp;
999 TAPDEBUG("%s polling, minor = %#x\n",
1000 ifp->if_xname, dev2unit(dev));
1002 if (events & (POLLIN | POLLRDNORM)) {
1003 IFQ_LOCK(&ifp->if_snd);
1004 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
1005 TAPDEBUG("%s have data in queue. len = %d, " \
1006 "minor = %#x\n", ifp->if_xname,
1007 ifp->if_snd.ifq_len, dev2unit(dev));
1009 revents |= (events & (POLLIN | POLLRDNORM));
1011 TAPDEBUG("%s waiting for data, minor = %#x\n",
1012 ifp->if_xname, dev2unit(dev));
1014 selrecord(td, &tp->tap_rsel);
1016 IFQ_UNLOCK(&ifp->if_snd);
1019 if (events & (POLLOUT | POLLWRNORM))
1020 revents |= (events & (POLLOUT | POLLWRNORM));
1029 * support for kevent() system call
1032 tapkqfilter(struct cdev *dev, struct knote *kn)
1034 struct tap_softc *tp = dev->si_drv1;
1035 struct ifnet *ifp = tp->tap_ifp;
1037 switch (kn->kn_filter) {
1039 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1040 ifp->if_xname, dev2unit(dev));
1041 kn->kn_fop = &tap_read_filterops;
1045 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1046 ifp->if_xname, dev2unit(dev));
1047 kn->kn_fop = &tap_write_filterops;
1051 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1052 ifp->if_xname, dev2unit(dev));
1058 knlist_add(&tp->tap_rsel.si_note, kn, 0);
1067 * Return true if there is data in the interface queue
1070 tapkqread(struct knote *kn, long hint)
1073 struct tap_softc *tp = kn->kn_hook;
1074 struct cdev *dev = tp->tap_dev;
1075 struct ifnet *ifp = tp->tap_ifp;
1077 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1078 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1079 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev));
1082 TAPDEBUG("%s waiting for data, minor = %#x\n",
1083 ifp->if_xname, dev2unit(dev));
1094 * Always can write. Return the MTU in kn->data
1097 tapkqwrite(struct knote *kn, long hint)
1099 struct tap_softc *tp = kn->kn_hook;
1100 struct ifnet *ifp = tp->tap_ifp;
1102 kn->kn_data = ifp->if_mtu;
1109 tapkqdetach(struct knote *kn)
1111 struct tap_softc *tp = kn->kn_hook;
1113 knlist_remove(&tp->tap_rsel.si_note, kn, 0);