2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (C) 1999-2000 by Maksim Yevmenkin <m_evmenkin@yahoo.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * -------------------------------------------------------------------------
31 * Copyright (c) 1988, Julian Onions <jpo@cs.nott.ac.uk>
32 * Nottingham University 1987.
37 * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $
42 #include <sys/param.h>
45 #include <sys/fcntl.h>
46 #include <sys/filio.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
51 #include <sys/module.h>
55 #include <sys/selinfo.h>
56 #include <sys/signalvar.h>
57 #include <sys/socket.h>
58 #include <sys/sockio.h>
60 #include <sys/sysctl.h>
61 #include <sys/systm.h>
62 #include <sys/ttycom.h>
64 #include <sys/queue.h>
67 #include <net/ethernet.h>
69 #include <net/if_var.h>
70 #include <net/if_clone.h>
71 #include <net/if_dl.h>
72 #include <net/if_media.h>
73 #include <net/if_types.h>
74 #include <net/route.h>
77 #include <netinet/in.h>
79 #include <net/if_tapvar.h>
80 #include <net/if_tap.h>
82 #define CDEV_NAME "tap"
83 #define TAPDEBUG if (tapdebug) printf
85 static const char tapname[] = "tap";
86 static const char vmnetname[] = "vmnet";
87 #define TAPMAXUNIT 0x7fff
88 #define VMNET_DEV_MASK CLONE_FLAG0
91 static int tapmodevent(module_t, int, void *);
94 static void tapclone(void *, struct ucred *, char *, int,
96 static void tapcreate(struct cdev *);
98 /* network interface */
99 static void tapifstart(struct ifnet *);
100 static int tapifioctl(struct ifnet *, u_long, caddr_t);
101 static void tapifinit(void *);
103 static int tap_clone_create(struct if_clone *, int, caddr_t);
104 static void tap_clone_destroy(struct ifnet *);
105 static struct if_clone *tap_cloner;
106 static int vmnet_clone_create(struct if_clone *, int, caddr_t);
107 static void vmnet_clone_destroy(struct ifnet *);
108 static struct if_clone *vmnet_cloner;
110 /* character device */
111 static d_open_t tapopen;
112 static d_close_t tapclose;
113 static d_read_t tapread;
114 static d_write_t tapwrite;
115 static d_ioctl_t tapioctl;
116 static d_poll_t tappoll;
117 static d_kqfilter_t tapkqfilter;
120 static int tapkqread(struct knote *, long);
121 static int tapkqwrite(struct knote *, long);
122 static void tapkqdetach(struct knote *);
124 static struct filterops tap_read_filterops = {
127 .f_detach = tapkqdetach,
128 .f_event = tapkqread,
131 static struct filterops tap_write_filterops = {
134 .f_detach = tapkqdetach,
135 .f_event = tapkqwrite,
138 static struct cdevsw tap_cdevsw = {
139 .d_version = D_VERSION,
140 .d_flags = D_NEEDMINOR,
148 .d_kqfilter = tapkqfilter,
152 * All global variables in if_tap.c are locked with tapmtx, with the
153 * exception of tapdebug, which is accessed unlocked; tapclones is
156 static struct mtx tapmtx;
157 static int tapdebug = 0; /* debug flag */
158 static int tapuopen = 0; /* allow user open() */
159 static int tapuponopen = 0; /* IFF_UP on open() */
160 static int tapdclone = 1; /* enable devfs cloning */
161 static SLIST_HEAD(, tap_softc) taphead; /* first device */
162 static struct clonedevs *tapclones;
164 MALLOC_DECLARE(M_TAP);
165 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
166 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
168 static struct sx tap_ioctl_sx;
169 SX_SYSINIT(tap_ioctl_sx, &tap_ioctl_sx, "tap_ioctl");
171 SYSCTL_DECL(_net_link);
172 static SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
173 "Ethernet tunnel software network interface");
174 SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
175 "Allow user to open /dev/tap (based on node permissions)");
176 SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0,
177 "Bring interface up when /dev/tap is opened");
178 SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RWTUN, &tapdclone, 0,
179 "Enable legacy devfs interface creation");
180 SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
182 DEV_MODULE(if_tap, tapmodevent, NULL);
183 MODULE_VERSION(if_tun, 1);
186 tap_clone_create(struct if_clone *ifc, int unit, caddr_t params)
191 /* Find any existing device, or allocate new unit number. */
192 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, 0);
194 dev = make_dev(&tap_cdevsw, unit, UID_ROOT, GID_WHEEL, 0600,
195 "%s%d", tapname, unit);
202 /* vmnet devices are tap devices in disguise */
204 vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params)
209 /* Find any existing device, or allocate new unit number. */
210 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, VMNET_DEV_MASK);
212 dev = make_dev(&tap_cdevsw, unit | VMNET_DEV_MASK, UID_ROOT,
213 GID_WHEEL, 0600, "%s%d", vmnetname, unit);
221 tap_destroy(struct tap_softc *tp)
223 struct ifnet *ifp = tp->tap_ifp;
225 CURVNET_SET(ifp->if_vnet);
226 sx_xlock(&tap_ioctl_sx);
227 ifp->if_softc = NULL;
228 sx_xunlock(&tap_ioctl_sx);
230 destroy_dev(tp->tap_dev);
231 seldrain(&tp->tap_rsel);
232 knlist_clear(&tp->tap_rsel.si_note, 0);
233 knlist_destroy(&tp->tap_rsel.si_note);
237 mtx_destroy(&tp->tap_mtx);
243 tap_clone_destroy(struct ifnet *ifp)
245 struct tap_softc *tp = ifp->if_softc;
248 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
253 /* vmnet devices are tap devices in disguise */
255 vmnet_clone_destroy(struct ifnet *ifp)
257 tap_clone_destroy(ifp);
263 * module event handler
266 tapmodevent(module_t mod, int type, void *data)
268 static eventhandler_tag eh_tag = NULL;
269 struct tap_softc *tp = NULL;
270 struct ifnet *ifp = NULL;
275 /* intitialize device */
277 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
278 SLIST_INIT(&taphead);
280 clone_setup(&tapclones);
281 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
282 if (eh_tag == NULL) {
283 clone_cleanup(&tapclones);
284 mtx_destroy(&tapmtx);
287 tap_cloner = if_clone_simple(tapname, tap_clone_create,
288 tap_clone_destroy, 0);
289 vmnet_cloner = if_clone_simple(vmnetname, vmnet_clone_create,
290 vmnet_clone_destroy, 0);
295 * The EBUSY algorithm here can't quite atomically
296 * guarantee that this is race-free since we have to
297 * release the tap mtx to deregister the clone handler.
300 SLIST_FOREACH(tp, &taphead, tap_next) {
301 mtx_lock(&tp->tap_mtx);
302 if (tp->tap_flags & TAP_OPEN) {
303 mtx_unlock(&tp->tap_mtx);
307 mtx_unlock(&tp->tap_mtx);
311 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
312 if_clone_detach(tap_cloner);
313 if_clone_detach(vmnet_cloner);
314 drain_dev_clone_events();
317 while ((tp = SLIST_FIRST(&taphead)) != NULL) {
318 SLIST_REMOVE_HEAD(&taphead, tap_next);
323 TAPDEBUG("detaching %s\n", ifp->if_xname);
329 clone_cleanup(&tapclones);
331 mtx_destroy(&tapmtx);
346 * We need to support two kind of devices - tap and vmnet
349 tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
351 char devname[SPECNAMELEN + 1];
352 int i, unit, append_unit;
359 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE) != 0))
366 /* We're interested in only tap/vmnet devices. */
367 if (strcmp(name, tapname) == 0) {
369 } else if (strcmp(name, vmnetname) == 0) {
371 extra = VMNET_DEV_MASK;
372 } else if (dev_stdclone(name, NULL, tapname, &unit) != 1) {
373 if (dev_stdclone(name, NULL, vmnetname, &unit) != 1) {
376 extra = VMNET_DEV_MASK;
383 CURVNET_SET(CRED_TO_VNET(cred));
384 /* find any existing device, or allocate new unit number */
385 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
389 * We were passed 'tun' or 'tap', with no unit specified
390 * so we'll need to append it now.
392 namelen = snprintf(devname, sizeof(devname), "%s%d", name,
397 *dev = make_dev_credf(MAKEDEV_REF, &tap_cdevsw, unit | extra,
398 cred, UID_ROOT, GID_WHEEL, 0600, "%s", name);
401 if_clone_create(name, namelen, NULL);
409 * to create interface
412 tapcreate(struct cdev *dev)
414 struct ifnet *ifp = NULL;
415 struct tap_softc *tp = NULL;
416 unsigned short macaddr_hi;
417 uint32_t macaddr_mid;
419 const char *name = NULL;
422 /* allocate driver storage and create device */
423 tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
424 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
426 SLIST_INSERT_HEAD(&taphead, tp, tap_next);
429 unit = dev2unit(dev);
431 /* select device: tap or vmnet */
432 if (unit & VMNET_DEV_MASK) {
434 tp->tap_flags |= TAP_VMNET;
440 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev));
442 /* generate fake MAC address: 00 bd xx xx xx unit_no */
443 macaddr_hi = htons(0x00bd);
444 macaddr_mid = (uint32_t) ticks;
445 bcopy(&macaddr_hi, eaddr, sizeof(short));
446 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t));
447 eaddr[5] = (u_char)unit;
449 /* fill the rest and attach interface */
450 ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
452 panic("%s%d: can not if_alloc()", name, unit);
454 if_initname(ifp, name, unit);
455 ifp->if_init = tapifinit;
456 ifp->if_start = tapifstart;
457 ifp->if_ioctl = tapifioctl;
458 ifp->if_mtu = ETHERMTU;
459 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
460 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
461 ifp->if_capabilities |= IFCAP_LINKSTATE;
462 ifp->if_capenable |= IFCAP_LINKSTATE;
467 ether_ifattach(ifp, eaddr);
469 mtx_lock(&tp->tap_mtx);
470 tp->tap_flags |= TAP_INITED;
471 mtx_unlock(&tp->tap_mtx);
473 knlist_init_mtx(&tp->tap_rsel.si_note, &tp->tap_mtx);
475 TAPDEBUG("interface %s is created. minor = %#x\n",
476 ifp->if_xname, dev2unit(dev));
483 * to open tunnel. must be superuser
486 tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
488 struct tap_softc *tp = NULL;
489 struct ifnet *ifp = NULL;
493 error = priv_check(td, PRIV_NET_TAP);
498 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
503 mtx_lock(&tp->tap_mtx);
504 if (tp->tap_flags & TAP_OPEN) {
505 mtx_unlock(&tp->tap_mtx);
509 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
510 tp->tap_pid = td->td_proc->p_pid;
511 tp->tap_flags |= TAP_OPEN;
514 ifp->if_drv_flags |= IFF_DRV_RUNNING;
515 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
517 ifp->if_flags |= IFF_UP;
518 if_link_state_change(ifp, LINK_STATE_UP);
519 mtx_unlock(&tp->tap_mtx);
521 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev));
530 * close the device - mark i/f down & delete routing info
533 tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
536 struct tap_softc *tp = dev->si_drv1;
537 struct ifnet *ifp = tp->tap_ifp;
539 /* junk all pending output */
540 mtx_lock(&tp->tap_mtx);
541 CURVNET_SET(ifp->if_vnet);
542 IF_DRAIN(&ifp->if_snd);
545 * Do not bring the interface down, and do not anything with
546 * interface, if we are in VMnet mode. Just close the device.
548 if (((tp->tap_flags & TAP_VMNET) == 0) &&
549 (ifp->if_flags & (IFF_UP | IFF_LINK0)) == IFF_UP) {
550 mtx_unlock(&tp->tap_mtx);
552 mtx_lock(&tp->tap_mtx);
553 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
554 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
555 mtx_unlock(&tp->tap_mtx);
556 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
557 rtinit(ifa, (int)RTM_DELETE, 0);
560 mtx_lock(&tp->tap_mtx);
564 if_link_state_change(ifp, LINK_STATE_DOWN);
567 funsetown(&tp->tap_sigio);
568 selwakeuppri(&tp->tap_rsel, PZERO+1);
569 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
571 tp->tap_flags &= ~TAP_OPEN;
573 mtx_unlock(&tp->tap_mtx);
575 TAPDEBUG("%s is closed. minor = %#x\n",
576 ifp->if_xname, dev2unit(dev));
585 * network interface initialization function
590 struct tap_softc *tp = (struct tap_softc *)xtp;
591 struct ifnet *ifp = tp->tap_ifp;
593 TAPDEBUG("initializing %s\n", ifp->if_xname);
595 mtx_lock(&tp->tap_mtx);
596 ifp->if_drv_flags |= IFF_DRV_RUNNING;
597 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
598 mtx_unlock(&tp->tap_mtx);
600 /* attempt to start output */
608 * Process an ioctl request on network interface
611 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
613 struct tap_softc *tp;
614 struct ifreq *ifr = (struct ifreq *)data;
615 struct ifstat *ifs = NULL;
616 struct ifmediareq *ifmr = NULL;
617 int dummy, error = 0;
619 sx_xlock(&tap_ioctl_sx);
626 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
632 ifmr = (struct ifmediareq *)data;
633 dummy = ifmr->ifm_count;
635 ifmr->ifm_status = IFM_AVALID;
636 ifmr->ifm_active = IFM_ETHER;
637 if (tp->tap_flags & TAP_OPEN)
638 ifmr->ifm_status |= IFM_ACTIVE;
639 ifmr->ifm_current = ifmr->ifm_active;
641 int media = IFM_ETHER;
642 error = copyout(&media, ifmr->ifm_ulist,
648 ifp->if_mtu = ifr->ifr_mtu;
652 ifs = (struct ifstat *)data;
653 mtx_lock(&tp->tap_mtx);
654 if (tp->tap_pid != 0)
655 snprintf(ifs->ascii, sizeof(ifs->ascii),
656 "\tOpened by PID %d\n", tp->tap_pid);
658 ifs->ascii[0] = '\0';
659 mtx_unlock(&tp->tap_mtx);
663 error = ether_ioctl(ifp, cmd, data);
668 sx_xunlock(&tap_ioctl_sx);
676 * queue packets from higher level ready to put out
679 tapifstart(struct ifnet *ifp)
681 struct tap_softc *tp = ifp->if_softc;
683 TAPDEBUG("%s starting\n", ifp->if_xname);
686 * do not junk pending output if we are in VMnet mode.
687 * XXX: can this do any harm because of queue overflow?
690 mtx_lock(&tp->tap_mtx);
691 if (((tp->tap_flags & TAP_VMNET) == 0) &&
692 ((tp->tap_flags & TAP_READY) != TAP_READY)) {
696 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
700 IF_DEQUEUE(&ifp->if_snd, m);
703 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
707 mtx_unlock(&tp->tap_mtx);
712 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
714 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
715 if (tp->tap_flags & TAP_RWAIT) {
716 tp->tap_flags &= ~TAP_RWAIT;
720 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
721 mtx_unlock(&tp->tap_mtx);
722 pgsigio(&tp->tap_sigio, SIGIO, 0);
723 mtx_lock(&tp->tap_mtx);
726 selwakeuppri(&tp->tap_rsel, PZERO+1);
727 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
728 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); /* obytes are counted in ether_output */
731 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
732 mtx_unlock(&tp->tap_mtx);
739 * the cdevsw interface is now pretty minimal
742 tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
745 struct tap_softc *tp = dev->si_drv1;
746 struct ifnet *ifp = tp->tap_ifp;
747 struct tapinfo *tapp = NULL;
750 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
751 defined(COMPAT_FREEBSD4)
757 tapp = (struct tapinfo *)data;
758 if (ifp->if_type != tapp->type)
760 mtx_lock(&tp->tap_mtx);
761 if (ifp->if_mtu != tapp->mtu) {
762 strlcpy(ifr.ifr_name, if_name(ifp), IFNAMSIZ);
763 ifr.ifr_mtu = tapp->mtu;
764 CURVNET_SET(ifp->if_vnet);
765 error = ifhwioctl(SIOCSIFMTU, ifp,
769 mtx_unlock(&tp->tap_mtx);
773 ifp->if_baudrate = tapp->baudrate;
774 mtx_unlock(&tp->tap_mtx);
778 tapp = (struct tapinfo *)data;
779 mtx_lock(&tp->tap_mtx);
780 tapp->mtu = ifp->if_mtu;
781 tapp->type = ifp->if_type;
782 tapp->baudrate = ifp->if_baudrate;
783 mtx_unlock(&tp->tap_mtx);
787 tapdebug = *(int *)data;
791 *(int *)data = tapdebug;
795 struct ifreq *ifr = (struct ifreq *) data;
797 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
804 mtx_lock(&tp->tap_mtx);
806 tp->tap_flags |= TAP_ASYNC;
808 tp->tap_flags &= ~TAP_ASYNC;
809 mtx_unlock(&tp->tap_mtx);
813 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
816 IFQ_LOCK(&ifp->if_snd);
817 IFQ_POLL_NOLOCK(&ifp->if_snd, mb);
818 for (*(int *)data = 0; mb != NULL;
820 *(int *)data += mb->m_len;
821 IFQ_UNLOCK(&ifp->if_snd);
827 return (fsetown(*(int *)data, &tp->tap_sigio));
830 *(int *)data = fgetown(&tp->tap_sigio);
833 /* this is deprecated, FIOSETOWN should be used instead */
835 return (fsetown(-(*(int *)data), &tp->tap_sigio));
837 /* this is deprecated, FIOGETOWN should be used instead */
839 *(int *)data = -fgetown(&tp->tap_sigio);
842 /* VMware/VMnet port ioctl's */
844 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
845 defined(COMPAT_FREEBSD4)
847 ival = IOCPARM_IVAL(data);
848 data = (caddr_t)&ival;
851 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
854 f &= ~IFF_CANTCHANGE;
857 mtx_lock(&tp->tap_mtx);
858 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
859 mtx_unlock(&tp->tap_mtx);
862 case SIOCGIFADDR: /* get MAC address of the remote side */
863 mtx_lock(&tp->tap_mtx);
864 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
865 mtx_unlock(&tp->tap_mtx);
868 case SIOCSIFADDR: /* set MAC address of the remote side */
869 mtx_lock(&tp->tap_mtx);
870 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
871 mtx_unlock(&tp->tap_mtx);
884 * the cdevsw read interface - reads a packet at a time, or at
885 * least as much of a packet as can be read
888 tapread(struct cdev *dev, struct uio *uio, int flag)
890 struct tap_softc *tp = dev->si_drv1;
891 struct ifnet *ifp = tp->tap_ifp;
892 struct mbuf *m = NULL;
895 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev));
897 mtx_lock(&tp->tap_mtx);
898 if ((tp->tap_flags & TAP_READY) != TAP_READY) {
899 mtx_unlock(&tp->tap_mtx);
902 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
903 ifp->if_xname, dev2unit(dev), tp->tap_flags);
908 tp->tap_flags &= ~TAP_RWAIT;
910 /* sleep until we get a packet */
912 IF_DEQUEUE(&ifp->if_snd, m);
915 if (flag & O_NONBLOCK) {
916 mtx_unlock(&tp->tap_mtx);
917 return (EWOULDBLOCK);
920 tp->tap_flags |= TAP_RWAIT;
921 error = mtx_sleep(tp, &tp->tap_mtx, PCATCH | (PZERO + 1),
924 mtx_unlock(&tp->tap_mtx);
929 mtx_unlock(&tp->tap_mtx);
931 /* feed packet to bpf */
934 /* xfer packet to user space */
935 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
936 len = min(uio->uio_resid, m->m_len);
940 error = uiomove(mtod(m, void *), len, uio);
945 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
957 * the cdevsw write interface - an atomic write is a packet - or else!
960 tapwrite(struct cdev *dev, struct uio *uio, int flag)
962 struct ether_header *eh;
963 struct tap_softc *tp = dev->si_drv1;
964 struct ifnet *ifp = tp->tap_ifp;
967 TAPDEBUG("%s writing, minor = %#x\n",
968 ifp->if_xname, dev2unit(dev));
970 if (uio->uio_resid == 0)
973 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
974 TAPDEBUG("%s invalid packet len = %zd, minor = %#x\n",
975 ifp->if_xname, uio->uio_resid, dev2unit(dev));
980 if ((m = m_uiotombuf(uio, M_NOWAIT, 0, ETHER_ALIGN,
981 M_PKTHDR)) == NULL) {
982 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
986 m->m_pkthdr.rcvif = ifp;
989 * Only pass a unicast frame to ether_input(), if it would actually
990 * have been received by non-virtual hardware.
992 if (m->m_len < sizeof(struct ether_header)) {
996 eh = mtod(m, struct ether_header *);
998 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
999 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
1000 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
1005 /* Pass packet up to parent. */
1006 CURVNET_SET(ifp->if_vnet);
1007 (*ifp->if_input)(ifp, m);
1009 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); /* ibytes are counted in parent */
1018 * the poll interface, this is only useful on reads
1019 * really. the write detect always returns true, write never blocks
1020 * anyway, it either accepts the packet or drops it
1023 tappoll(struct cdev *dev, int events, struct thread *td)
1025 struct tap_softc *tp = dev->si_drv1;
1026 struct ifnet *ifp = tp->tap_ifp;
1029 TAPDEBUG("%s polling, minor = %#x\n",
1030 ifp->if_xname, dev2unit(dev));
1032 if (events & (POLLIN | POLLRDNORM)) {
1033 IFQ_LOCK(&ifp->if_snd);
1034 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
1035 TAPDEBUG("%s have data in queue. len = %d, " \
1036 "minor = %#x\n", ifp->if_xname,
1037 ifp->if_snd.ifq_len, dev2unit(dev));
1039 revents |= (events & (POLLIN | POLLRDNORM));
1041 TAPDEBUG("%s waiting for data, minor = %#x\n",
1042 ifp->if_xname, dev2unit(dev));
1044 selrecord(td, &tp->tap_rsel);
1046 IFQ_UNLOCK(&ifp->if_snd);
1049 if (events & (POLLOUT | POLLWRNORM))
1050 revents |= (events & (POLLOUT | POLLWRNORM));
1059 * support for kevent() system call
1062 tapkqfilter(struct cdev *dev, struct knote *kn)
1064 struct tap_softc *tp = dev->si_drv1;
1065 struct ifnet *ifp = tp->tap_ifp;
1067 switch (kn->kn_filter) {
1069 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1070 ifp->if_xname, dev2unit(dev));
1071 kn->kn_fop = &tap_read_filterops;
1075 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1076 ifp->if_xname, dev2unit(dev));
1077 kn->kn_fop = &tap_write_filterops;
1081 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1082 ifp->if_xname, dev2unit(dev));
1088 knlist_add(&tp->tap_rsel.si_note, kn, 0);
1097 * Return true if there is data in the interface queue
1100 tapkqread(struct knote *kn, long hint)
1103 struct tap_softc *tp = kn->kn_hook;
1104 struct cdev *dev = tp->tap_dev;
1105 struct ifnet *ifp = tp->tap_ifp;
1107 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1108 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1109 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev));
1112 TAPDEBUG("%s waiting for data, minor = %#x\n",
1113 ifp->if_xname, dev2unit(dev));
1124 * Always can write. Return the MTU in kn->data
1127 tapkqwrite(struct knote *kn, long hint)
1129 struct tap_softc *tp = kn->kn_hook;
1130 struct ifnet *ifp = tp->tap_ifp;
1132 kn->kn_data = ifp->if_mtu;
1139 tapkqdetach(struct knote *kn)
1141 struct tap_softc *tp = kn->kn_hook;
1143 knlist_remove(&tp->tap_rsel.si_note, kn, 0);