1/*
2 * Copyright (c) 2003-2005 The Regents of The University of Michigan
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are
7 * met: redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer;
9 * redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution;
12 * neither the name of the copyright holders nor the names of its
13 * contributors may be used to endorse or promote products derived from
14 * this software without specific prior written permission.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
17 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
18 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
19 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
20 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 *
28 * Authors: Nathan Binkert
29 */
30
31/* @file
32 * Interface to connect a simulated ethernet device to the real world
33 */
34
35#include "dev/net/ethertap.hh"
36
37#if defined(__OpenBSD__) || defined(__APPLE__)
38#include <sys/param.h>
39
40#endif
41
42#if USE_TUNTAP && defined(__linux__)
43#if 1 // Hide from the style checker since these have to be out of order.
44#include <sys/socket.h> // Has to be included before if.h for some reason.
45
46#endif
47
48#include <linux/if.h>
49#include <linux/if_tun.h>
50
51#endif
52
53#include <fcntl.h>
54#include <netinet/in.h>
55#include <sys/ioctl.h>
56#include <unistd.h>
57
58#include <cstring>
59#include <deque>
60#include <string>
61
62#include "base/logging.hh"
63#include "base/pollevent.hh"
64#include "base/socket.hh"
65#include "base/trace.hh"
66#include "debug/Ethernet.hh"
67#include "debug/EthernetData.hh"
68#include "dev/net/etherdump.hh"
69#include "dev/net/etherint.hh"
70#include "dev/net/etherpkt.hh"
71
72using namespace std;
73
74class TapEvent : public PollEvent
75{
76  protected:
77    EtherTapBase *tap;
78
79  public:
80    TapEvent(EtherTapBase *_tap, int fd, int e)
81        : PollEvent(fd, e), tap(_tap) {}
82
83    void
84    process(int revent) override
85    {
86        // Ensure that our event queue is active. It may not be since we get
87        // here from the PollQueue whenever a real packet happens to arrive.
88        EventQueue::ScopedMigration migrate(tap->eventQueue());
89
90        tap->recvReal(revent);
91    }
92};
93
94EtherTapBase::EtherTapBase(const Params *p)
95    : SimObject(p), buflen(p->bufsz), dump(p->dump), event(NULL),
96      interface(NULL),
97      txEvent([this]{ retransmit(); }, "EtherTapBase retransmit")
98{
99    buffer = new uint8_t[buflen];
100    interface = new EtherTapInt(name() + ".interface", this);
101}
102
103EtherTapBase::~EtherTapBase()
104{
105    delete buffer;
106    delete event;
107    delete interface;
108}
109
110void
111EtherTapBase::serialize(CheckpointOut &cp) const
112{
113    SERIALIZE_SCALAR(buflen);
114    uint8_t *buffer = (uint8_t *)this->buffer;
115    SERIALIZE_ARRAY(buffer, buflen);
116
117    bool tapevent_present = false;
118    if (event) {
119        tapevent_present = true;
120        SERIALIZE_SCALAR(tapevent_present);
121        event->serialize(cp);
122    } else {
123        SERIALIZE_SCALAR(tapevent_present);
124    }
125}
126
127void
128EtherTapBase::unserialize(CheckpointIn &cp)
129{
130    UNSERIALIZE_SCALAR(buflen);
131    uint8_t *buffer = (uint8_t *)this->buffer;
132    UNSERIALIZE_ARRAY(buffer, buflen);
133
134    bool tapevent_present;
135    UNSERIALIZE_SCALAR(tapevent_present);
136    if (tapevent_present) {
137        event = new TapEvent(this, 0, 0);
138        event->unserialize(cp);
139        if (event->queued())
140            pollQueue.schedule(event);
141    }
142}
143
144
145void
146EtherTapBase::pollFd(int fd)
147{
148    assert(!event);
149    event = new TapEvent(this, fd, POLLIN|POLLERR);
150    pollQueue.schedule(event);
151}
152
153void
154EtherTapBase::stopPolling()
155{
156    assert(event);
157    delete event;
158    event = NULL;
159}
160
161
162Port &
163EtherTapBase::getPort(const std::string &if_name, PortID idx)
164{
165    if (if_name == "tap")
166        return *interface;
167    return SimObject::getPort(if_name, idx);
168}
169
170bool
171EtherTapBase::recvSimulated(EthPacketPtr packet)
172{
173    if (dump)
174        dump->dump(packet);
175
176    DPRINTF(Ethernet, "EtherTap sim->real len=%d\n", packet->length);
177    DDUMP(EthernetData, packet->data, packet->length);
178
179    bool success = sendReal(packet->data, packet->length);
180
181    interface->recvDone();
182
183    return success;
184}
185
186void
187EtherTapBase::sendSimulated(void *data, size_t len)
188{
189    EthPacketPtr packet;
190    packet = make_shared<EthPacketData>(len);
191    packet->length = len;
192    packet->simLength = len;
193    memcpy(packet->data, data, len);
194
195    DPRINTF(Ethernet, "EtherTap real->sim len=%d\n", packet->length);
196    DDUMP(EthernetData, packet->data, packet->length);
197    if (!packetBuffer.empty() || !interface->sendPacket(packet)) {
198        DPRINTF(Ethernet, "bus busy...buffer for retransmission\n");
199        packetBuffer.push(packet);
200        if (!txEvent.scheduled())
201            schedule(txEvent, curTick() + retryTime);
202    } else if (dump) {
203        dump->dump(packet);
204    }
205}
206
207void
208EtherTapBase::retransmit()
209{
210    if (packetBuffer.empty())
211        return;
212
213    EthPacketPtr packet = packetBuffer.front();
214    if (interface->sendPacket(packet)) {
215        if (dump)
216            dump->dump(packet);
217        DPRINTF(Ethernet, "EtherTap retransmit\n");
218        packetBuffer.front() = NULL;
219        packetBuffer.pop();
220    }
221
222    if (!packetBuffer.empty() && !txEvent.scheduled())
223        schedule(txEvent, curTick() + retryTime);
224}
225
226
227class TapListener
228{
229  protected:
230    class Event : public PollEvent
231    {
232      protected:
233        TapListener *listener;
234
235      public:
236        Event(TapListener *l, int fd, int e) : PollEvent(fd, e), listener(l) {}
237
238        void process(int revent) override { listener->accept(); }
239    };
240
241    friend class Event;
242    Event *event;
243
244    void accept();
245
246  protected:
247    ListenSocket listener;
248    EtherTapStub *tap;
249    int port;
250
251  public:
252    TapListener(EtherTapStub *t, int p) : event(NULL), tap(t), port(p) {}
253    ~TapListener() { delete event; }
254
255    void listen();
256};
257
258void
259TapListener::listen()
260{
261    while (!listener.listen(port, true)) {
262        DPRINTF(Ethernet, "TapListener(listen): Can't bind port %d\n", port);
263        port++;
264    }
265
266    ccprintf(cerr, "Listening for tap connection on port %d\n", port);
267    event = new Event(this, listener.getfd(), POLLIN|POLLERR);
268    pollQueue.schedule(event);
269}
270
271void
272TapListener::accept()
273{
274    // As a consequence of being called from the PollQueue, we might
275    // have been called from a different thread. Migrate to "our"
276    // thread.
277    EventQueue::ScopedMigration migrate(tap->eventQueue());
278
279    if (!listener.islistening())
280        panic("TapListener(accept): cannot accept if we're not listening!");
281
282    int sfd = listener.accept(true);
283    if (sfd != -1)
284        tap->attach(sfd);
285}
286
287
288EtherTapStub::EtherTapStub(const Params *p) : EtherTapBase(p), socket(-1)
289{
290    if (ListenSocket::allDisabled())
291        fatal("All listeners are disabled! EtherTapStub can't work!");
292
293    listener = new TapListener(this, p->port);
294    listener->listen();
295}
296
297EtherTapStub::~EtherTapStub()
298{
299    delete listener;
300}
301
302void
303EtherTapStub::serialize(CheckpointOut &cp) const
304{
305    EtherTapBase::serialize(cp);
306
307    SERIALIZE_SCALAR(socket);
308    SERIALIZE_SCALAR(buffer_used);
309    SERIALIZE_SCALAR(frame_len);
310}
311
312void
313EtherTapStub::unserialize(CheckpointIn &cp)
314{
315    EtherTapBase::unserialize(cp);
316
317    UNSERIALIZE_SCALAR(socket);
318    UNSERIALIZE_SCALAR(buffer_used);
319    UNSERIALIZE_SCALAR(frame_len);
320}
321
322
323void
324EtherTapStub::attach(int fd)
325{
326    if (socket != -1)
327        close(fd);
328
329    buffer_used = 0;
330    frame_len = 0;
331    socket = fd;
332    DPRINTF(Ethernet, "EtherTapStub attached\n");
333    pollFd(socket);
334}
335
336void
337EtherTapStub::detach()
338{
339    DPRINTF(Ethernet, "EtherTapStub detached\n");
340    stopPolling();
341    close(socket);
342    socket = -1;
343}
344
345void
346EtherTapStub::recvReal(int revent)
347{
348    if (revent & POLLERR) {
349        detach();
350        return;
351    }
352
353    if (!(revent & POLLIN))
354        return;
355
356    // Read in as much of the new data as we can.
357    int len = read(socket, buffer + buffer_used, buflen - buffer_used);
358    if (len == 0) {
359        detach();
360        return;
361    }
362    buffer_used += len;
363
364    // If there's not enough data for the frame length, wait for more.
365    if (buffer_used < sizeof(uint32_t))
366        return;
367
368    if (frame_len == 0)
369        frame_len = ntohl(*(uint32_t *)buffer);
370
371    DPRINTF(Ethernet, "Received data from peer: len=%d buffer_used=%d "
372            "frame_len=%d\n", len, buffer_used, frame_len);
373
374    uint8_t *frame_start = &buffer[sizeof(uint32_t)];
375    while (frame_len != 0 && buffer_used >= frame_len + sizeof(uint32_t)) {
376        sendSimulated(frame_start, frame_len);
377
378        // Bookkeeping.
379        buffer_used -= frame_len + sizeof(uint32_t);
380        if (buffer_used > 0) {
381            // If there's still any data left, move it into position.
382            memmove(buffer, frame_start + frame_len, buffer_used);
383        }
384        frame_len = 0;
385
386        if (buffer_used >= sizeof(uint32_t))
387            frame_len = ntohl(*(uint32_t *)buffer);
388    }
389}
390
391bool
392EtherTapStub::sendReal(const void *data, size_t len)
393{
394    uint32_t frame_len = htonl(len);
395    ssize_t ret = write(socket, &frame_len, sizeof(frame_len));
396    if (ret != sizeof(frame_len))
397        return false;
398    return write(socket, data, len) == len;
399}
400
401
402#if USE_TUNTAP
403
404EtherTap::EtherTap(const Params *p) : EtherTapBase(p)
405{
406    int fd = open(p->tun_clone_device.c_str(), O_RDWR | O_NONBLOCK);
407    if (fd < 0)
408        panic("Couldn't open %s.\n", p->tun_clone_device);
409
410    struct ifreq ifr;
411    memset(&ifr, 0, sizeof(ifr));
412    ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
413    strncpy(ifr.ifr_name, p->tap_device_name.c_str(), IFNAMSIZ - 1);
414
415    if (ioctl(fd, TUNSETIFF, (void *)&ifr) < 0)
416        panic("Failed to access tap device %s.\n", ifr.ifr_name);
417    // fd now refers to the tap device.
418    tap = fd;
419    pollFd(tap);
420}
421
422EtherTap::~EtherTap()
423{
424    stopPolling();
425    close(tap);
426    tap = -1;
427}
428
429void
430EtherTap::recvReal(int revent)
431{
432    if (revent & POLLERR)
433        panic("Error polling for tap data.\n");
434
435    if (!(revent & POLLIN))
436        return;
437
438    ssize_t ret;
439    while ((ret = read(tap, buffer, buflen))) {
440        if (ret < 0) {
441            if (errno == EAGAIN)
442                break;
443            panic("Failed to read from tap device.\n");
444        }
445
446        sendSimulated(buffer, ret);
447    }
448}
449
450bool
451EtherTap::sendReal(const void *data, size_t len)
452{
453    int n;
454    pollfd pfd[1];
455    pfd->fd = tap;
456    pfd->events = POLLOUT;
457
458    // `tap` is a nonblock fd. Here we try to write until success, and use
459    // poll to make a blocking wait.
460    while ((n = write(tap, data, len)) != len) {
461        if (errno != EAGAIN)
462            panic("Failed to write data to tap device.\n");
463        pfd->revents = 0;
464        int ret = poll(pfd, 1, -1);
465        // timeout is set to inf, we shouldn't get 0 in any case.
466        assert(ret != 0);
467        if (ret == -1 || (ret == 1 && (pfd->revents & POLLERR))) {
468            panic("Failed when polling to write data to tap device.\n");
469        }
470    }
471    return true;
472}
473
474EtherTap *
475EtherTapParams::create()
476{
477    return new EtherTap(this);
478}
479
480#endif
481
482EtherTapStub *
483EtherTapStubParams::create()
484{
485    return new EtherTapStub(this);
486}
487