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", tapname, 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_destroy(&tp->tap_rsel.si_note);
227 mtx_destroy(&tp->tap_mtx);
233 tap_clone_destroy(struct ifnet *ifp)
235 struct tap_softc *tp = ifp->if_softc;
238 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
243 /* vmnet devices are tap devices in disguise */
245 vmnet_clone_destroy(struct ifnet *ifp)
247 tap_clone_destroy(ifp);
253 * module event handler
256 tapmodevent(module_t mod, int type, void *data)
258 static eventhandler_tag eh_tag = NULL;
259 struct tap_softc *tp = NULL;
260 struct ifnet *ifp = NULL;
265 /* intitialize device */
267 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
268 SLIST_INIT(&taphead);
270 clone_setup(&tapclones);
271 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
272 if (eh_tag == NULL) {
273 clone_cleanup(&tapclones);
274 mtx_destroy(&tapmtx);
277 tap_cloner = if_clone_simple(tapname, tap_clone_create,
278 tap_clone_destroy, 0);
279 vmnet_cloner = if_clone_simple(vmnetname, vmnet_clone_create,
280 vmnet_clone_destroy, 0);
285 * The EBUSY algorithm here can't quite atomically
286 * guarantee that this is race-free since we have to
287 * release the tap mtx to deregister the clone handler.
290 SLIST_FOREACH(tp, &taphead, tap_next) {
291 mtx_lock(&tp->tap_mtx);
292 if (tp->tap_flags & TAP_OPEN) {
293 mtx_unlock(&tp->tap_mtx);
297 mtx_unlock(&tp->tap_mtx);
301 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
302 if_clone_detach(tap_cloner);
303 if_clone_detach(vmnet_cloner);
304 drain_dev_clone_events();
307 while ((tp = SLIST_FIRST(&taphead)) != NULL) {
308 SLIST_REMOVE_HEAD(&taphead, tap_next);
313 TAPDEBUG("detaching %s\n", ifp->if_xname);
319 clone_cleanup(&tapclones);
321 mtx_destroy(&tapmtx);
336 * We need to support two kind of devices - tap and vmnet
339 tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
341 char devname[SPECNAMELEN + 1];
342 int i, unit, append_unit;
349 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0))
356 /* We're interested in only tap/vmnet devices. */
357 if (strcmp(name, tapname) == 0) {
359 } else if (strcmp(name, vmnetname) == 0) {
361 extra = VMNET_DEV_MASK;
362 } else if (dev_stdclone(name, NULL, tapname, &unit) != 1) {
363 if (dev_stdclone(name, NULL, vmnetname, &unit) != 1) {
366 extra = VMNET_DEV_MASK;
373 CURVNET_SET(CRED_TO_VNET(cred));
374 /* find any existing device, or allocate new unit number */
375 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
379 * We were passed 'tun' or 'tap', with no unit specified
380 * so we'll need to append it now.
382 namelen = snprintf(devname, sizeof(devname), "%s%d", name,
387 *dev = make_dev_credf(MAKEDEV_REF, &tap_cdevsw, unit | extra,
388 cred, UID_ROOT, GID_WHEEL, 0600, "%s", name);
391 if_clone_create(name, namelen, NULL);
399 * to create interface
402 tapcreate(struct cdev *dev)
404 struct ifnet *ifp = NULL;
405 struct tap_softc *tp = NULL;
406 unsigned short macaddr_hi;
407 uint32_t macaddr_mid;
409 const char *name = NULL;
412 dev->si_flags &= ~SI_CHEAPCLONE;
414 /* allocate driver storage and create device */
415 tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
416 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
418 SLIST_INSERT_HEAD(&taphead, tp, tap_next);
421 unit = dev2unit(dev);
423 /* select device: tap or vmnet */
424 if (unit & VMNET_DEV_MASK) {
426 tp->tap_flags |= TAP_VMNET;
432 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev));
434 /* generate fake MAC address: 00 bd xx xx xx unit_no */
435 macaddr_hi = htons(0x00bd);
436 macaddr_mid = (uint32_t) ticks;
437 bcopy(&macaddr_hi, eaddr, sizeof(short));
438 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t));
439 eaddr[5] = (u_char)unit;
441 /* fill the rest and attach interface */
442 ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
444 panic("%s%d: can not if_alloc()", name, unit);
446 if_initname(ifp, name, unit);
447 ifp->if_init = tapifinit;
448 ifp->if_start = tapifstart;
449 ifp->if_ioctl = tapifioctl;
450 ifp->if_mtu = ETHERMTU;
451 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
452 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
453 ifp->if_capabilities |= IFCAP_LINKSTATE;
454 ifp->if_capenable |= IFCAP_LINKSTATE;
459 ether_ifattach(ifp, eaddr);
461 mtx_lock(&tp->tap_mtx);
462 tp->tap_flags |= TAP_INITED;
463 mtx_unlock(&tp->tap_mtx);
465 knlist_init_mtx(&tp->tap_rsel.si_note, &tp->tap_mtx);
467 TAPDEBUG("interface %s is created. minor = %#x\n",
468 ifp->if_xname, dev2unit(dev));
475 * to open tunnel. must be superuser
478 tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
480 struct tap_softc *tp = NULL;
481 struct ifnet *ifp = NULL;
485 error = priv_check(td, PRIV_NET_TAP);
490 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
495 mtx_lock(&tp->tap_mtx);
496 if (tp->tap_flags & TAP_OPEN) {
497 mtx_unlock(&tp->tap_mtx);
501 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
502 tp->tap_pid = td->td_proc->p_pid;
503 tp->tap_flags |= TAP_OPEN;
506 ifp->if_drv_flags |= IFF_DRV_RUNNING;
507 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
509 ifp->if_flags |= IFF_UP;
510 if_link_state_change(ifp, LINK_STATE_UP);
511 mtx_unlock(&tp->tap_mtx);
513 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev));
522 * close the device - mark i/f down & delete routing info
525 tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
528 struct tap_softc *tp = dev->si_drv1;
529 struct ifnet *ifp = tp->tap_ifp;
531 /* junk all pending output */
532 mtx_lock(&tp->tap_mtx);
533 CURVNET_SET(ifp->if_vnet);
534 IF_DRAIN(&ifp->if_snd);
537 * do not bring the interface down, and do not anything with
538 * interface, if we are in VMnet mode. just close the device.
541 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
542 mtx_unlock(&tp->tap_mtx);
544 mtx_lock(&tp->tap_mtx);
545 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
546 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
547 mtx_unlock(&tp->tap_mtx);
548 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
549 rtinit(ifa, (int)RTM_DELETE, 0);
552 mtx_lock(&tp->tap_mtx);
556 if_link_state_change(ifp, LINK_STATE_DOWN);
559 funsetown(&tp->tap_sigio);
560 selwakeuppri(&tp->tap_rsel, PZERO+1);
561 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
563 tp->tap_flags &= ~TAP_OPEN;
565 mtx_unlock(&tp->tap_mtx);
567 TAPDEBUG("%s is closed. minor = %#x\n",
568 ifp->if_xname, dev2unit(dev));
577 * network interface initialization function
582 struct tap_softc *tp = (struct tap_softc *)xtp;
583 struct ifnet *ifp = tp->tap_ifp;
585 TAPDEBUG("initializing %s\n", ifp->if_xname);
587 mtx_lock(&tp->tap_mtx);
588 ifp->if_drv_flags |= IFF_DRV_RUNNING;
589 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
590 mtx_unlock(&tp->tap_mtx);
592 /* attempt to start output */
600 * Process an ioctl request on network interface
603 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
605 struct tap_softc *tp = ifp->if_softc;
606 struct ifreq *ifr = (struct ifreq *)data;
607 struct ifstat *ifs = NULL;
608 struct ifmediareq *ifmr = NULL;
609 int dummy, error = 0;
612 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
618 ifmr = (struct ifmediareq *)data;
619 dummy = ifmr->ifm_count;
621 ifmr->ifm_status = IFM_AVALID;
622 ifmr->ifm_active = IFM_ETHER;
623 if (tp->tap_flags & TAP_OPEN)
624 ifmr->ifm_status |= IFM_ACTIVE;
625 ifmr->ifm_current = ifmr->ifm_active;
627 int media = IFM_ETHER;
628 error = copyout(&media, ifmr->ifm_ulist,
634 ifp->if_mtu = ifr->ifr_mtu;
638 ifs = (struct ifstat *)data;
639 dummy = strlen(ifs->ascii);
640 mtx_lock(&tp->tap_mtx);
641 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii))
642 snprintf(ifs->ascii + dummy,
643 sizeof(ifs->ascii) - dummy,
644 "\tOpened by PID %d\n", tp->tap_pid);
645 mtx_unlock(&tp->tap_mtx);
649 error = ether_ioctl(ifp, cmd, data);
660 * queue packets from higher level ready to put out
663 tapifstart(struct ifnet *ifp)
665 struct tap_softc *tp = ifp->if_softc;
667 TAPDEBUG("%s starting\n", ifp->if_xname);
670 * do not junk pending output if we are in VMnet mode.
671 * XXX: can this do any harm because of queue overflow?
674 mtx_lock(&tp->tap_mtx);
675 if (((tp->tap_flags & TAP_VMNET) == 0) &&
676 ((tp->tap_flags & TAP_READY) != TAP_READY)) {
680 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
684 IF_DEQUEUE(&ifp->if_snd, m);
691 mtx_unlock(&tp->tap_mtx);
696 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
698 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
699 if (tp->tap_flags & TAP_RWAIT) {
700 tp->tap_flags &= ~TAP_RWAIT;
704 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
705 mtx_unlock(&tp->tap_mtx);
706 pgsigio(&tp->tap_sigio, SIGIO, 0);
707 mtx_lock(&tp->tap_mtx);
710 selwakeuppri(&tp->tap_rsel, PZERO+1);
711 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
712 ifp->if_opackets ++; /* obytes are counted in ether_output */
715 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
716 mtx_unlock(&tp->tap_mtx);
723 * the cdevsw interface is now pretty minimal
726 tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
728 struct tap_softc *tp = dev->si_drv1;
729 struct ifnet *ifp = tp->tap_ifp;
730 struct tapinfo *tapp = NULL;
732 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
733 defined(COMPAT_FREEBSD4)
739 tapp = (struct tapinfo *)data;
740 mtx_lock(&tp->tap_mtx);
741 ifp->if_mtu = tapp->mtu;
742 ifp->if_type = tapp->type;
743 ifp->if_baudrate = tapp->baudrate;
744 mtx_unlock(&tp->tap_mtx);
748 tapp = (struct tapinfo *)data;
749 mtx_lock(&tp->tap_mtx);
750 tapp->mtu = ifp->if_mtu;
751 tapp->type = ifp->if_type;
752 tapp->baudrate = ifp->if_baudrate;
753 mtx_unlock(&tp->tap_mtx);
757 tapdebug = *(int *)data;
761 *(int *)data = tapdebug;
765 struct ifreq *ifr = (struct ifreq *) data;
767 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
774 mtx_lock(&tp->tap_mtx);
776 tp->tap_flags |= TAP_ASYNC;
778 tp->tap_flags &= ~TAP_ASYNC;
779 mtx_unlock(&tp->tap_mtx);
783 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
786 IFQ_LOCK(&ifp->if_snd);
787 IFQ_POLL_NOLOCK(&ifp->if_snd, mb);
788 for (*(int *)data = 0; mb != NULL;
790 *(int *)data += mb->m_len;
791 IFQ_UNLOCK(&ifp->if_snd);
797 return (fsetown(*(int *)data, &tp->tap_sigio));
800 *(int *)data = fgetown(&tp->tap_sigio);
803 /* this is deprecated, FIOSETOWN should be used instead */
805 return (fsetown(-(*(int *)data), &tp->tap_sigio));
807 /* this is deprecated, FIOGETOWN should be used instead */
809 *(int *)data = -fgetown(&tp->tap_sigio);
812 /* VMware/VMnet port ioctl's */
814 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
815 defined(COMPAT_FREEBSD4)
817 ival = IOCPARM_IVAL(data);
818 data = (caddr_t)&ival;
821 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
824 f &= ~IFF_CANTCHANGE;
827 mtx_lock(&tp->tap_mtx);
828 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
829 mtx_unlock(&tp->tap_mtx);
832 case OSIOCGIFADDR: /* get MAC address of the remote side */
834 mtx_lock(&tp->tap_mtx);
835 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
836 mtx_unlock(&tp->tap_mtx);
839 case SIOCSIFADDR: /* set MAC address of the remote side */
840 mtx_lock(&tp->tap_mtx);
841 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
842 mtx_unlock(&tp->tap_mtx);
855 * the cdevsw read interface - reads a packet at a time, or at
856 * least as much of a packet as can be read
859 tapread(struct cdev *dev, struct uio *uio, int flag)
861 struct tap_softc *tp = dev->si_drv1;
862 struct ifnet *ifp = tp->tap_ifp;
863 struct mbuf *m = NULL;
866 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev));
868 mtx_lock(&tp->tap_mtx);
869 if ((tp->tap_flags & TAP_READY) != TAP_READY) {
870 mtx_unlock(&tp->tap_mtx);
873 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
874 ifp->if_xname, dev2unit(dev), tp->tap_flags);
879 tp->tap_flags &= ~TAP_RWAIT;
881 /* sleep until we get a packet */
883 IF_DEQUEUE(&ifp->if_snd, m);
886 if (flag & O_NONBLOCK) {
887 mtx_unlock(&tp->tap_mtx);
888 return (EWOULDBLOCK);
891 tp->tap_flags |= TAP_RWAIT;
892 error = mtx_sleep(tp, &tp->tap_mtx, PCATCH | (PZERO + 1),
895 mtx_unlock(&tp->tap_mtx);
900 mtx_unlock(&tp->tap_mtx);
902 /* feed packet to bpf */
905 /* xfer packet to user space */
906 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
907 len = min(uio->uio_resid, m->m_len);
911 error = uiomove(mtod(m, void *), len, uio);
916 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
928 * the cdevsw write interface - an atomic write is a packet - or else!
931 tapwrite(struct cdev *dev, struct uio *uio, int flag)
933 struct ether_header *eh;
934 struct tap_softc *tp = dev->si_drv1;
935 struct ifnet *ifp = tp->tap_ifp;
938 TAPDEBUG("%s writing, minor = %#x\n",
939 ifp->if_xname, dev2unit(dev));
941 if (uio->uio_resid == 0)
944 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
945 TAPDEBUG("%s invalid packet len = %zd, minor = %#x\n",
946 ifp->if_xname, uio->uio_resid, dev2unit(dev));
951 if ((m = m_uiotombuf(uio, M_NOWAIT, 0, ETHER_ALIGN,
952 M_PKTHDR)) == NULL) {
957 m->m_pkthdr.rcvif = ifp;
960 * Only pass a unicast frame to ether_input(), if it would actually
961 * have been received by non-virtual hardware.
963 if (m->m_len < sizeof(struct ether_header)) {
967 eh = mtod(m, struct ether_header *);
969 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
970 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
971 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
976 /* Pass packet up to parent. */
977 CURVNET_SET(ifp->if_vnet);
978 (*ifp->if_input)(ifp, m);
980 ifp->if_ipackets ++; /* ibytes are counted in parent */
989 * the poll interface, this is only useful on reads
990 * really. the write detect always returns true, write never blocks
991 * anyway, it either accepts the packet or drops it
994 tappoll(struct cdev *dev, int events, struct thread *td)
996 struct tap_softc *tp = dev->si_drv1;
997 struct ifnet *ifp = tp->tap_ifp;
1000 TAPDEBUG("%s polling, minor = %#x\n",
1001 ifp->if_xname, dev2unit(dev));
1003 if (events & (POLLIN | POLLRDNORM)) {
1004 IFQ_LOCK(&ifp->if_snd);
1005 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
1006 TAPDEBUG("%s have data in queue. len = %d, " \
1007 "minor = %#x\n", ifp->if_xname,
1008 ifp->if_snd.ifq_len, dev2unit(dev));
1010 revents |= (events & (POLLIN | POLLRDNORM));
1012 TAPDEBUG("%s waiting for data, minor = %#x\n",
1013 ifp->if_xname, dev2unit(dev));
1015 selrecord(td, &tp->tap_rsel);
1017 IFQ_UNLOCK(&ifp->if_snd);
1020 if (events & (POLLOUT | POLLWRNORM))
1021 revents |= (events & (POLLOUT | POLLWRNORM));
1030 * support for kevent() system call
1033 tapkqfilter(struct cdev *dev, struct knote *kn)
1035 struct tap_softc *tp = dev->si_drv1;
1036 struct ifnet *ifp = tp->tap_ifp;
1038 switch (kn->kn_filter) {
1040 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1041 ifp->if_xname, dev2unit(dev));
1042 kn->kn_fop = &tap_read_filterops;
1046 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1047 ifp->if_xname, dev2unit(dev));
1048 kn->kn_fop = &tap_write_filterops;
1052 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1053 ifp->if_xname, dev2unit(dev));
1059 knlist_add(&tp->tap_rsel.si_note, kn, 0);
1068 * Return true if there is data in the interface queue
1071 tapkqread(struct knote *kn, long hint)
1074 struct tap_softc *tp = kn->kn_hook;
1075 struct cdev *dev = tp->tap_dev;
1076 struct ifnet *ifp = tp->tap_ifp;
1078 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1079 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1080 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev));
1083 TAPDEBUG("%s waiting for data, minor = %#x\n",
1084 ifp->if_xname, dev2unit(dev));
1095 * Always can write. Return the MTU in kn->data
1098 tapkqwrite(struct knote *kn, long hint)
1100 struct tap_softc *tp = kn->kn_hook;
1101 struct ifnet *ifp = tp->tap_ifp;
1103 kn->kn_data = ifp->if_mtu;
1110 tapkqdetach(struct knote *kn)
1112 struct tap_softc *tp = kn->kn_hook;
1114 knlist_remove(&tp->tap_rsel.si_note, kn, 0);