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_var.h>
67 #include <net/if_clone.h>
68 #include <net/if_dl.h>
69 #include <net/if_media.h>
70 #include <net/if_types.h>
71 #include <net/route.h>
74 #include <netinet/in.h>
76 #include <net/if_tapvar.h>
77 #include <net/if_tap.h>
80 #define CDEV_NAME "tap"
81 #define TAPDEBUG if (tapdebug) printf
83 static const char tapname[] = "tap";
84 static const char vmnetname[] = "vmnet";
85 #define TAPMAXUNIT 0x7fff
86 #define VMNET_DEV_MASK CLONE_FLAG0
89 static int tapmodevent(module_t, int, void *);
92 static void tapclone(void *, struct ucred *, char *, int,
94 static void tapcreate(struct cdev *);
96 /* network interface */
97 static void tapifstart(struct ifnet *);
98 static int tapifioctl(struct ifnet *, u_long, caddr_t);
99 static void tapifinit(void *);
101 static int tap_clone_create(struct if_clone *, int, caddr_t);
102 static void tap_clone_destroy(struct ifnet *);
103 static struct if_clone *tap_cloner;
104 static int vmnet_clone_create(struct if_clone *, int, caddr_t);
105 static void vmnet_clone_destroy(struct ifnet *);
106 static struct if_clone *vmnet_cloner;
108 /* character device */
109 static d_open_t tapopen;
110 static d_close_t tapclose;
111 static d_read_t tapread;
112 static d_write_t tapwrite;
113 static d_ioctl_t tapioctl;
114 static d_poll_t tappoll;
115 static d_kqfilter_t tapkqfilter;
118 static int tapkqread(struct knote *, long);
119 static int tapkqwrite(struct knote *, long);
120 static void tapkqdetach(struct knote *);
122 static struct filterops tap_read_filterops = {
125 .f_detach = tapkqdetach,
126 .f_event = tapkqread,
129 static struct filterops tap_write_filterops = {
132 .f_detach = tapkqdetach,
133 .f_event = tapkqwrite,
136 static struct cdevsw tap_cdevsw = {
137 .d_version = D_VERSION,
138 .d_flags = D_NEEDMINOR,
146 .d_kqfilter = tapkqfilter,
150 * All global variables in if_tap.c are locked with tapmtx, with the
151 * exception of tapdebug, which is accessed unlocked; tapclones is
154 static struct mtx tapmtx;
155 static int tapdebug = 0; /* debug flag */
156 static int tapuopen = 0; /* allow user open() */
157 static int tapuponopen = 0; /* IFF_UP on open() */
158 static int tapdclone = 1; /* enable devfs cloning */
159 static int tapclosedeladdrs = 1; /* del addrs on close */
160 static SLIST_HEAD(, tap_softc) taphead; /* first device */
161 static struct clonedevs *tapclones;
163 MALLOC_DECLARE(M_TAP);
164 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
165 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
167 SYSCTL_DECL(_net_link);
168 static SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0,
169 "Ethernet tunnel software network interface");
170 SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0,
171 "Allow user to open /dev/tap (based on node permissions)");
172 SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0,
173 "Bring interface up when /dev/tap is opened");
174 SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RWTUN, &tapdclone, 0,
175 "Enably legacy devfs interface creation");
176 SYSCTL_INT(_net_link_tap, OID_AUTO, deladdrs_on_close, CTLFLAG_RW,
177 &tapclosedeladdrs, 0, "Delete addresses and routes when /dev/tap is "
179 SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, "");
181 DEV_MODULE(if_tap, tapmodevent, NULL);
184 tap_clone_create(struct if_clone *ifc, int unit, caddr_t params)
189 /* Find any existing device, or allocate new unit number. */
190 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, 0);
192 dev = make_dev(&tap_cdevsw, unit, UID_ROOT, GID_WHEEL, 0600,
193 "%s%d", tapname, unit);
200 /* vmnet devices are tap devices in disguise */
202 vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params)
207 /* Find any existing device, or allocate new unit number. */
208 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, VMNET_DEV_MASK);
210 dev = make_dev(&tap_cdevsw, unit | VMNET_DEV_MASK, UID_ROOT,
211 GID_WHEEL, 0600, "%s%d", vmnetname, unit);
219 tap_destroy(struct tap_softc *tp)
221 struct ifnet *ifp = tp->tap_ifp;
223 CURVNET_SET(ifp->if_vnet);
224 destroy_dev(tp->tap_dev);
225 seldrain(&tp->tap_rsel);
226 knlist_clear(&tp->tap_rsel.si_note, 0);
227 knlist_destroy(&tp->tap_rsel.si_note);
231 mtx_destroy(&tp->tap_mtx);
237 tap_clone_destroy(struct ifnet *ifp)
239 struct tap_softc *tp = ifp->if_softc;
242 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next);
247 /* vmnet devices are tap devices in disguise */
249 vmnet_clone_destroy(struct ifnet *ifp)
251 tap_clone_destroy(ifp);
257 * module event handler
260 tapmodevent(module_t mod, int type, void *data)
262 static eventhandler_tag eh_tag = NULL;
263 struct tap_softc *tp = NULL;
264 struct ifnet *ifp = NULL;
269 /* intitialize device */
271 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF);
272 SLIST_INIT(&taphead);
274 clone_setup(&tapclones);
275 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000);
276 if (eh_tag == NULL) {
277 clone_cleanup(&tapclones);
278 mtx_destroy(&tapmtx);
281 tap_cloner = if_clone_simple(tapname, tap_clone_create,
282 tap_clone_destroy, 0);
283 vmnet_cloner = if_clone_simple(vmnetname, vmnet_clone_create,
284 vmnet_clone_destroy, 0);
289 * The EBUSY algorithm here can't quite atomically
290 * guarantee that this is race-free since we have to
291 * release the tap mtx to deregister the clone handler.
294 SLIST_FOREACH(tp, &taphead, tap_next) {
295 mtx_lock(&tp->tap_mtx);
296 if (tp->tap_flags & TAP_OPEN) {
297 mtx_unlock(&tp->tap_mtx);
301 mtx_unlock(&tp->tap_mtx);
305 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag);
306 if_clone_detach(tap_cloner);
307 if_clone_detach(vmnet_cloner);
308 drain_dev_clone_events();
311 while ((tp = SLIST_FIRST(&taphead)) != NULL) {
312 SLIST_REMOVE_HEAD(&taphead, tap_next);
317 TAPDEBUG("detaching %s\n", ifp->if_xname);
323 clone_cleanup(&tapclones);
325 mtx_destroy(&tapmtx);
340 * We need to support two kind of devices - tap and vmnet
343 tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev)
345 char devname[SPECNAMELEN + 1];
346 int i, unit, append_unit;
353 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0))
360 /* We're interested in only tap/vmnet devices. */
361 if (strcmp(name, tapname) == 0) {
363 } else if (strcmp(name, vmnetname) == 0) {
365 extra = VMNET_DEV_MASK;
366 } else if (dev_stdclone(name, NULL, tapname, &unit) != 1) {
367 if (dev_stdclone(name, NULL, vmnetname, &unit) != 1) {
370 extra = VMNET_DEV_MASK;
377 CURVNET_SET(CRED_TO_VNET(cred));
378 /* find any existing device, or allocate new unit number */
379 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra);
383 * We were passed 'tun' or 'tap', with no unit specified
384 * so we'll need to append it now.
386 namelen = snprintf(devname, sizeof(devname), "%s%d", name,
391 *dev = make_dev_credf(MAKEDEV_REF, &tap_cdevsw, unit | extra,
392 cred, UID_ROOT, GID_WHEEL, 0600, "%s", name);
395 if_clone_create(name, namelen, NULL);
403 * to create interface
406 tapcreate(struct cdev *dev)
408 struct ifnet *ifp = NULL;
409 struct tap_softc *tp = NULL;
410 unsigned short macaddr_hi;
411 uint32_t macaddr_mid;
413 const char *name = NULL;
416 /* allocate driver storage and create device */
417 tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
418 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF);
420 SLIST_INSERT_HEAD(&taphead, tp, tap_next);
423 unit = dev2unit(dev);
425 /* select device: tap or vmnet */
426 if (unit & VMNET_DEV_MASK) {
428 tp->tap_flags |= TAP_VMNET;
434 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev));
436 /* generate fake MAC address: 00 bd xx xx xx unit_no */
437 macaddr_hi = htons(0x00bd);
438 macaddr_mid = (uint32_t) ticks;
439 bcopy(&macaddr_hi, eaddr, sizeof(short));
440 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t));
441 eaddr[5] = (u_char)unit;
443 /* fill the rest and attach interface */
444 ifp = tp->tap_ifp = if_alloc(IFT_ETHER);
446 panic("%s%d: can not if_alloc()", name, unit);
448 if_initname(ifp, name, unit);
449 ifp->if_init = tapifinit;
450 ifp->if_start = tapifstart;
451 ifp->if_ioctl = tapifioctl;
452 ifp->if_mtu = ETHERMTU;
453 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
454 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
455 ifp->if_capabilities |= IFCAP_LINKSTATE;
456 ifp->if_capenable |= IFCAP_LINKSTATE;
461 ether_ifattach(ifp, eaddr);
463 mtx_lock(&tp->tap_mtx);
464 tp->tap_flags |= TAP_INITED;
465 mtx_unlock(&tp->tap_mtx);
467 knlist_init_mtx(&tp->tap_rsel.si_note, &tp->tap_mtx);
469 TAPDEBUG("interface %s is created. minor = %#x\n",
470 ifp->if_xname, dev2unit(dev));
477 * to open tunnel. must be superuser
480 tapopen(struct cdev *dev, int flag, int mode, struct thread *td)
482 struct tap_softc *tp = NULL;
483 struct ifnet *ifp = NULL;
487 error = priv_check(td, PRIV_NET_TAP);
492 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT)
497 mtx_lock(&tp->tap_mtx);
498 if (tp->tap_flags & TAP_OPEN) {
499 mtx_unlock(&tp->tap_mtx);
503 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr));
504 tp->tap_pid = td->td_proc->p_pid;
505 tp->tap_flags |= TAP_OPEN;
508 ifp->if_drv_flags |= IFF_DRV_RUNNING;
509 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
511 ifp->if_flags |= IFF_UP;
512 if_link_state_change(ifp, LINK_STATE_UP);
513 mtx_unlock(&tp->tap_mtx);
515 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev));
524 * close the device - mark i/f down & delete routing info
527 tapclose(struct cdev *dev, int foo, int bar, struct thread *td)
530 struct tap_softc *tp = dev->si_drv1;
531 struct ifnet *ifp = tp->tap_ifp;
533 /* junk all pending output */
534 mtx_lock(&tp->tap_mtx);
535 CURVNET_SET(ifp->if_vnet);
536 IF_DRAIN(&ifp->if_snd);
539 * do not bring the interface down, and do not anything with
540 * interface, if we are in VMnet mode. just close the device.
543 if (tapclosedeladdrs == 1 && ((tp->tap_flags & TAP_VMNET) == 0) &&
544 (ifp->if_flags & IFF_UP)) {
545 mtx_unlock(&tp->tap_mtx);
547 mtx_lock(&tp->tap_mtx);
548 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
549 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
550 mtx_unlock(&tp->tap_mtx);
551 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
552 rtinit(ifa, (int)RTM_DELETE, 0);
555 mtx_lock(&tp->tap_mtx);
559 if_link_state_change(ifp, LINK_STATE_DOWN);
562 funsetown(&tp->tap_sigio);
563 selwakeuppri(&tp->tap_rsel, PZERO+1);
564 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
566 tp->tap_flags &= ~TAP_OPEN;
568 mtx_unlock(&tp->tap_mtx);
570 TAPDEBUG("%s is closed. minor = %#x\n",
571 ifp->if_xname, dev2unit(dev));
580 * network interface initialization function
585 struct tap_softc *tp = (struct tap_softc *)xtp;
586 struct ifnet *ifp = tp->tap_ifp;
588 TAPDEBUG("initializing %s\n", ifp->if_xname);
590 mtx_lock(&tp->tap_mtx);
591 ifp->if_drv_flags |= IFF_DRV_RUNNING;
592 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
593 mtx_unlock(&tp->tap_mtx);
595 /* attempt to start output */
603 * Process an ioctl request on network interface
606 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
608 struct tap_softc *tp = ifp->if_softc;
609 struct ifreq *ifr = (struct ifreq *)data;
610 struct ifstat *ifs = NULL;
611 struct ifmediareq *ifmr = NULL;
612 int dummy, error = 0;
615 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */
621 ifmr = (struct ifmediareq *)data;
622 dummy = ifmr->ifm_count;
624 ifmr->ifm_status = IFM_AVALID;
625 ifmr->ifm_active = IFM_ETHER;
626 if (tp->tap_flags & TAP_OPEN)
627 ifmr->ifm_status |= IFM_ACTIVE;
628 ifmr->ifm_current = ifmr->ifm_active;
630 int media = IFM_ETHER;
631 error = copyout(&media, ifmr->ifm_ulist,
637 ifp->if_mtu = ifr->ifr_mtu;
641 ifs = (struct ifstat *)data;
642 mtx_lock(&tp->tap_mtx);
643 if (tp->tap_pid != 0)
644 snprintf(ifs->ascii, sizeof(ifs->ascii),
645 "\tOpened by PID %d\n", tp->tap_pid);
647 ifs->ascii[0] = '\0';
648 mtx_unlock(&tp->tap_mtx);
652 error = ether_ioctl(ifp, cmd, data);
663 * queue packets from higher level ready to put out
666 tapifstart(struct ifnet *ifp)
668 struct tap_softc *tp = ifp->if_softc;
670 TAPDEBUG("%s starting\n", ifp->if_xname);
673 * do not junk pending output if we are in VMnet mode.
674 * XXX: can this do any harm because of queue overflow?
677 mtx_lock(&tp->tap_mtx);
678 if (((tp->tap_flags & TAP_VMNET) == 0) &&
679 ((tp->tap_flags & TAP_READY) != TAP_READY)) {
683 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname,
687 IF_DEQUEUE(&ifp->if_snd, m);
690 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
694 mtx_unlock(&tp->tap_mtx);
699 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
701 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
702 if (tp->tap_flags & TAP_RWAIT) {
703 tp->tap_flags &= ~TAP_RWAIT;
707 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
708 mtx_unlock(&tp->tap_mtx);
709 pgsigio(&tp->tap_sigio, SIGIO, 0);
710 mtx_lock(&tp->tap_mtx);
713 selwakeuppri(&tp->tap_rsel, PZERO+1);
714 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0);
715 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); /* obytes are counted in ether_output */
718 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
719 mtx_unlock(&tp->tap_mtx);
726 * the cdevsw interface is now pretty minimal
729 tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
731 struct tap_softc *tp = dev->si_drv1;
732 struct ifnet *ifp = tp->tap_ifp;
733 struct tapinfo *tapp = NULL;
735 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
736 defined(COMPAT_FREEBSD4)
742 tapp = (struct tapinfo *)data;
743 mtx_lock(&tp->tap_mtx);
744 ifp->if_mtu = tapp->mtu;
745 ifp->if_type = tapp->type;
746 ifp->if_baudrate = tapp->baudrate;
747 mtx_unlock(&tp->tap_mtx);
751 tapp = (struct tapinfo *)data;
752 mtx_lock(&tp->tap_mtx);
753 tapp->mtu = ifp->if_mtu;
754 tapp->type = ifp->if_type;
755 tapp->baudrate = ifp->if_baudrate;
756 mtx_unlock(&tp->tap_mtx);
760 tapdebug = *(int *)data;
764 *(int *)data = tapdebug;
768 struct ifreq *ifr = (struct ifreq *) data;
770 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ);
777 mtx_lock(&tp->tap_mtx);
779 tp->tap_flags |= TAP_ASYNC;
781 tp->tap_flags &= ~TAP_ASYNC;
782 mtx_unlock(&tp->tap_mtx);
786 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
789 IFQ_LOCK(&ifp->if_snd);
790 IFQ_POLL_NOLOCK(&ifp->if_snd, mb);
791 for (*(int *)data = 0; mb != NULL;
793 *(int *)data += mb->m_len;
794 IFQ_UNLOCK(&ifp->if_snd);
800 return (fsetown(*(int *)data, &tp->tap_sigio));
803 *(int *)data = fgetown(&tp->tap_sigio);
806 /* this is deprecated, FIOSETOWN should be used instead */
808 return (fsetown(-(*(int *)data), &tp->tap_sigio));
810 /* this is deprecated, FIOGETOWN should be used instead */
812 *(int *)data = -fgetown(&tp->tap_sigio);
815 /* VMware/VMnet port ioctl's */
817 #if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \
818 defined(COMPAT_FREEBSD4)
820 ival = IOCPARM_IVAL(data);
821 data = (caddr_t)&ival;
824 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
827 f &= ~IFF_CANTCHANGE;
830 mtx_lock(&tp->tap_mtx);
831 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
832 mtx_unlock(&tp->tap_mtx);
835 case SIOCGIFADDR: /* get MAC address of the remote side */
836 mtx_lock(&tp->tap_mtx);
837 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
838 mtx_unlock(&tp->tap_mtx);
841 case SIOCSIFADDR: /* set MAC address of the remote side */
842 mtx_lock(&tp->tap_mtx);
843 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
844 mtx_unlock(&tp->tap_mtx);
857 * the cdevsw read interface - reads a packet at a time, or at
858 * least as much of a packet as can be read
861 tapread(struct cdev *dev, struct uio *uio, int flag)
863 struct tap_softc *tp = dev->si_drv1;
864 struct ifnet *ifp = tp->tap_ifp;
865 struct mbuf *m = NULL;
868 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev));
870 mtx_lock(&tp->tap_mtx);
871 if ((tp->tap_flags & TAP_READY) != TAP_READY) {
872 mtx_unlock(&tp->tap_mtx);
875 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n",
876 ifp->if_xname, dev2unit(dev), tp->tap_flags);
881 tp->tap_flags &= ~TAP_RWAIT;
883 /* sleep until we get a packet */
885 IF_DEQUEUE(&ifp->if_snd, m);
888 if (flag & O_NONBLOCK) {
889 mtx_unlock(&tp->tap_mtx);
890 return (EWOULDBLOCK);
893 tp->tap_flags |= TAP_RWAIT;
894 error = mtx_sleep(tp, &tp->tap_mtx, PCATCH | (PZERO + 1),
897 mtx_unlock(&tp->tap_mtx);
902 mtx_unlock(&tp->tap_mtx);
904 /* feed packet to bpf */
907 /* xfer packet to user space */
908 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) {
909 len = min(uio->uio_resid, m->m_len);
913 error = uiomove(mtod(m, void *), len, uio);
918 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname,
930 * the cdevsw write interface - an atomic write is a packet - or else!
933 tapwrite(struct cdev *dev, struct uio *uio, int flag)
935 struct ether_header *eh;
936 struct tap_softc *tp = dev->si_drv1;
937 struct ifnet *ifp = tp->tap_ifp;
940 TAPDEBUG("%s writing, minor = %#x\n",
941 ifp->if_xname, dev2unit(dev));
943 if (uio->uio_resid == 0)
946 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
947 TAPDEBUG("%s invalid packet len = %zd, minor = %#x\n",
948 ifp->if_xname, uio->uio_resid, dev2unit(dev));
953 if ((m = m_uiotombuf(uio, M_NOWAIT, 0, ETHER_ALIGN,
954 M_PKTHDR)) == NULL) {
955 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
959 m->m_pkthdr.rcvif = ifp;
962 * Only pass a unicast frame to ether_input(), if it would actually
963 * have been received by non-virtual hardware.
965 if (m->m_len < sizeof(struct ether_header)) {
969 eh = mtod(m, struct ether_header *);
971 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 &&
972 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
973 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) {
978 /* Pass packet up to parent. */
979 CURVNET_SET(ifp->if_vnet);
980 (*ifp->if_input)(ifp, m);
982 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); /* ibytes are counted in parent */
991 * the poll interface, this is only useful on reads
992 * really. the write detect always returns true, write never blocks
993 * anyway, it either accepts the packet or drops it
996 tappoll(struct cdev *dev, int events, struct thread *td)
998 struct tap_softc *tp = dev->si_drv1;
999 struct ifnet *ifp = tp->tap_ifp;
1002 TAPDEBUG("%s polling, minor = %#x\n",
1003 ifp->if_xname, dev2unit(dev));
1005 if (events & (POLLIN | POLLRDNORM)) {
1006 IFQ_LOCK(&ifp->if_snd);
1007 if (!IFQ_IS_EMPTY(&ifp->if_snd)) {
1008 TAPDEBUG("%s have data in queue. len = %d, " \
1009 "minor = %#x\n", ifp->if_xname,
1010 ifp->if_snd.ifq_len, dev2unit(dev));
1012 revents |= (events & (POLLIN | POLLRDNORM));
1014 TAPDEBUG("%s waiting for data, minor = %#x\n",
1015 ifp->if_xname, dev2unit(dev));
1017 selrecord(td, &tp->tap_rsel);
1019 IFQ_UNLOCK(&ifp->if_snd);
1022 if (events & (POLLOUT | POLLWRNORM))
1023 revents |= (events & (POLLOUT | POLLWRNORM));
1032 * support for kevent() system call
1035 tapkqfilter(struct cdev *dev, struct knote *kn)
1037 struct tap_softc *tp = dev->si_drv1;
1038 struct ifnet *ifp = tp->tap_ifp;
1040 switch (kn->kn_filter) {
1042 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n",
1043 ifp->if_xname, dev2unit(dev));
1044 kn->kn_fop = &tap_read_filterops;
1048 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n",
1049 ifp->if_xname, dev2unit(dev));
1050 kn->kn_fop = &tap_write_filterops;
1054 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n",
1055 ifp->if_xname, dev2unit(dev));
1061 knlist_add(&tp->tap_rsel.si_note, kn, 0);
1070 * Return true if there is data in the interface queue
1073 tapkqread(struct knote *kn, long hint)
1076 struct tap_softc *tp = kn->kn_hook;
1077 struct cdev *dev = tp->tap_dev;
1078 struct ifnet *ifp = tp->tap_ifp;
1080 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) {
1081 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n",
1082 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev));
1085 TAPDEBUG("%s waiting for data, minor = %#x\n",
1086 ifp->if_xname, dev2unit(dev));
1097 * Always can write. Return the MTU in kn->data
1100 tapkqwrite(struct knote *kn, long hint)
1102 struct tap_softc *tp = kn->kn_hook;
1103 struct ifnet *ifp = tp->tap_ifp;
1105 kn->kn_data = ifp->if_mtu;
1112 tapkqdetach(struct knote *kn)
1114 struct tap_softc *tp = kn->kn_hook;
1116 knlist_remove(&tp->tap_rsel.si_note, kn, 0);