The Tcpdump Group git mirrors - libpcap/blob - pcap-netmap.c

   1 /*
   2  * Copyright (C) 2014 Luigi Rizzo. All rights reserved.
   3  *
   4  * Redistribution and use in source and binary forms, with or without
   5  * modification, are permitted provided that the following conditions
   6  * are met:
   7  *
   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.
  13  *
  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
  24  * SUCH DAMAGE.
  25  */
  26
  27 #ifdef HAVE_CONFIG_H
  28 #include <config.h>
  29 #endif
  30
  31 #include <poll.h>
  32 #include <ctype.h>
  33 #include <errno.h>
  34 #include <netdb.h>
  35 #include <stdio.h>
  36 #include <stdlib.h>
  37 #include <string.h>
  38 #include <unistd.h>
  39
  40 #define NETMAP_WITH_LIBS
  41 #include <net/netmap_user.h>
  42
  43 #include "pcap-int.h"
  44 #include "pcap-netmap.h"
  45
  46 #ifndef __FreeBSD__
  47   /*
  48    * On FreeBSD we use IFF_PPROMISC which is in ifr_flagshigh.
  49    * Remap to IFF_PROMISC on other platforms.
  50    *
  51    * XXX - DragonFly BSD?
  52    */
  53   #define IFF_PPROMISC  IFF_PROMISC
  54 #endif /* __FreeBSD__ */
  55
  56 struct pcap_netmap {
  57         struct nm_desc *d;      /* pointer returned by nm_open() */
  58         pcap_handler cb;        /* callback and argument */
  59         u_char *cb_arg;
  60         int must_clear_promisc; /* flag */
  61         uint64_t rx_pkts;       /* # of pkts received before the filter */
  62 };
  63
  64
  65 static int
  66 pcap_netmap_stats(pcap_t *p, struct pcap_stat *ps)
  67 {
  68         struct pcap_netmap *pn = p->priv;
  69
  70         ps->ps_recv = pn->rx_pkts;
  71         ps->ps_drop = 0;
  72         ps->ps_ifdrop = 0;
  73         return 0;
  74 }
  75
  76
  77 static void
  78 pcap_netmap_filter(u_char *arg, struct pcap_pkthdr *h, const u_char *buf)
  79 {
  80         pcap_t *p = (pcap_t *)arg;
  81         struct pcap_netmap *pn = p->priv;
  82         const struct bpf_insn *pc = p->fcode.bf_insns;
  83
  84         ++pn->rx_pkts;
  85         if (pc == NULL || bpf_filter(pc, buf, h->len, h->caplen))
  86                 pn->cb(pn->cb_arg, h, buf);
  87 }
  88
  89
  90 static int
  91 pcap_netmap_dispatch(pcap_t *p, int cnt, pcap_handler cb, u_char *user)
  92 {
  93         int ret;
  94         struct pcap_netmap *pn = p->priv;
  95         struct nm_desc *d = pn->d;
  96         struct pollfd pfd = { .fd = p->fd, .events = POLLIN, .revents = 0 };
  97
  98         pn->cb = cb;
  99         pn->cb_arg = user;
 100
 101         for (;;) {
 102                 if (p->break_loop) {
 103                         p->break_loop = 0;
 104                         return PCAP_ERROR_BREAK;
 105                 }
 106                 /* nm_dispatch won't run forever */
 107
 108                 ret = nm_dispatch((void *)d, cnt, (void *)pcap_netmap_filter, (void *)p);
 109                 if (ret != 0)
 110                         break;
 111                 errno = 0;
 112                 ret = poll(&pfd, 1, p->opt.timeout);
 113         }
 114         return ret;
 115 }
 116
 117
 118 /* XXX need to check the NIOCTXSYNC/poll */
 119 static int
 120 pcap_netmap_inject(pcap_t *p, const void *buf, size_t size)
 121 {
 122         struct pcap_netmap *pn = p->priv;
 123         struct nm_desc *d = pn->d;
 124
 125         return nm_inject(d, buf, size);
 126 }
 127
 128
 129 static int
 130 pcap_netmap_ioctl(pcap_t *p, u_long what, uint32_t *if_flags)
 131 {
 132         struct pcap_netmap *pn = p->priv;
 133         struct nm_desc *d = pn->d;
 134         struct ifreq ifr;
 135         int error, fd = d->fd;
 136
 137 #ifdef linux
 138         fd = socket(AF_INET, SOCK_DGRAM, 0);
 139         if (fd < 0) {
 140                 fprintf(stderr, "Error: cannot get device control socket.\n");
 141                 return -1;
 142         }
 143 #endif /* linux */
 144         bzero(&ifr, sizeof(ifr));
 145         strncpy(ifr.ifr_name, d->req.nr_name, sizeof(ifr.ifr_name));
 146         switch (what) {
 147         case SIOCSIFFLAGS:
 148                 /*
 149                  * The flags we pass in are 32-bit and unsigned.
 150                  *
 151                  * On most if not all UN*Xes, ifr_flags is 16-bit and
 152                  * signed, and the result of assigning a longer
 153                  * unsigned value to a shorter signed value is
 154                  * implementation-defined (even if, in practice, it'll
 155                  * do what's intended on all platforms we support
 156                  * result of assigning a 32-bit unsigned value).
 157                  * So we mask out the upper 16 bits.
 158                  */
 159                 ifr.ifr_flags = *if_flags & 0xffff;
 160 #ifdef __FreeBSD__
 161                 /*
 162                  * In FreeBSD, we need to set the high-order flags,
 163                  * as we're using IFF_PPROMISC, which is in those bits.
 164                  *
 165                  * XXX - DragonFly BSD?
 166                  */
 167                 ifr.ifr_flagshigh = *if_flags >> 16;
 168 #endif /* __FreeBSD__ */
 169                 break;
 170         }
 171         error = ioctl(fd, what, &ifr);
 172         if (!error) {
 173                 switch (what) {
 174                 case SIOCGIFFLAGS:
 175                         /*
 176                          * The flags we return are 32-bit.
 177                          *
 178                          * On most if not all UN*Xes, ifr_flags is
 179                          * 16-bit and signed, and will get sign-
 180                          * extended, so that the upper 16 bits of
 181                          * those flags will be forced on.  So we
 182                          * mask out the upper 16 bits of the
 183                          * sign-extended value.
 184                          */
 185                         *if_flags = ifr.ifr_flags & 0xffff;
 186 #ifdef __FreeBSD__
 187                         /*
 188                          * In FreeBSD, we need to return the
 189                          * high-order flags, as we're using
 190                          * IFF_PPROMISC, which is in those bits.
 191                          *
 192                          * XXX - DragonFly BSD?
 193                          */
 194                         *if_flags |= (ifr.ifr_flagshigh << 16);
 195 #endif /* __FreeBSD__ */
 196                 }
 197         }
 198 #ifdef linux
 199         close(fd);
 200 #endif /* linux */
 201         return error ? -1 : 0;
 202 }
 203
 204
 205 static void
 206 pcap_netmap_close(pcap_t *p)
 207 {
 208         struct pcap_netmap *pn = p->priv;
 209         struct nm_desc *d = pn->d;
 210         uint32_t if_flags = 0;
 211
 212         if (pn->must_clear_promisc) {
 213                 pcap_netmap_ioctl(p, SIOCGIFFLAGS, &if_flags); /* fetch flags */
 214                 if (if_flags & IFF_PPROMISC) {
 215                         if_flags &= ~IFF_PPROMISC;
 216                         pcap_netmap_ioctl(p, SIOCSIFFLAGS, &if_flags);
 217                 }
 218         }
 219         nm_close(d);
 220         pcap_cleanup_live_common(p);
 221 }
 222
 223
 224 static int
 225 pcap_netmap_activate(pcap_t *p)
 226 {
 227         struct pcap_netmap *pn = p->priv;
 228         struct nm_desc *d;
 229         uint32_t if_flags = 0;
 230
 231         d = nm_open(p->opt.device, NULL, 0, NULL);
 232         if (d == NULL) {
 233                 pcap_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
 234                     errno, "netmap open: cannot access %s",
 235                     p->opt.device);
 236                 pcap_cleanup_live_common(p);
 237                 return (PCAP_ERROR);
 238         }
 239 #if 0
 240         fprintf(stderr, "%s device %s priv %p fd %d ports %d..%d\n",
 241             __FUNCTION__, p->opt.device, d, d->fd,
 242             d->first_rx_ring, d->last_rx_ring);
 243 #endif
 244         pn->d = d;
 245         p->fd = d->fd;
 246
 247         /*
 248          * Turn a negative snapshot value (invalid), a snapshot value of
 249          * 0 (unspecified), or a value bigger than the normal maximum
 250          * value, into the maximum allowed value.
 251          *
 252          * If some application really *needs* a bigger snapshot
 253          * length, we should just increase MAXIMUM_SNAPLEN.
 254          */
 255         if (p->snapshot <= 0 || p->snapshot > MAXIMUM_SNAPLEN)
 256                 p->snapshot = MAXIMUM_SNAPLEN;
 257
 258         if (p->opt.promisc && !(d->req.nr_ringid & NETMAP_SW_RING)) {
 259                 pcap_netmap_ioctl(p, SIOCGIFFLAGS, &if_flags); /* fetch flags */
 260                 if (!(if_flags & IFF_PPROMISC)) {
 261                         pn->must_clear_promisc = 1;
 262                         if_flags |= IFF_PPROMISC;
 263                         pcap_netmap_ioctl(p, SIOCSIFFLAGS, &if_flags);
 264                 }
 265         }
 266         p->linktype = DLT_EN10MB;
 267         p->selectable_fd = p->fd;
 268         p->read_op = pcap_netmap_dispatch;
 269         p->inject_op = pcap_netmap_inject;
 270         p->setfilter_op = install_bpf_program;
 271         p->setdirection_op = NULL;
 272         p->set_datalink_op = NULL;
 273         p->getnonblock_op = pcap_getnonblock_fd;
 274         p->setnonblock_op = pcap_setnonblock_fd;
 275         p->stats_op = pcap_netmap_stats;
 276         p->cleanup_op = pcap_netmap_close;
 277
 278         return (0);
 279 }
 280
 281
 282 pcap_t *
 283 pcap_netmap_create(const char *device, char *ebuf, int *is_ours)
 284 {
 285         pcap_t *p;
 286
 287         *is_ours = (!strncmp(device, "netmap:", 7) || !strncmp(device, "vale", 4));
 288         if (! *is_ours)
 289                 return NULL;
 290         p = pcap_create_common(ebuf, sizeof (struct pcap_netmap));
 291         if (p == NULL)
 292                 return (NULL);
 293         p->activate_op = pcap_netmap_activate;
 294         return (p);
 295 }
 296
 297 /*
 298  * The "device name" for netmap devices isn't a name for a device, it's
 299  * an expression that indicates how the device should be set up, so
 300  * there's no way to enumerate them.
 301  */
 302 int
 303 pcap_netmap_findalldevs(pcap_if_list_t *devlistp _U_, char *err_str _U_)
 304 {
 305         return 0;
 306 }