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 #include <config.h>
  28
  29 #include <poll.h>
  30 #include <errno.h>
  31 #include <netdb.h>
  32 #include <stdio.h>
  33 #include <stdlib.h>
  34 #include <string.h>
  35 #include <unistd.h>
  36 #include <sys/ioctl.h>
  37
  38 #define NETMAP_WITH_LIBS
  39 #include <net/netmap_user.h>
  40
  41 #include "pcap-int.h"
  42 #include "pcap-netmap.h"
  43
  44 #ifndef __FreeBSD__
  45   /*
  46    * On FreeBSD we use IFF_PPROMISC which is in ifr_flagshigh.
  47    * Remap to IFF_PROMISC on other platforms.
  48    *
  49    * XXX - DragonFly BSD?
  50    */
  51   #define IFF_PPROMISC  IFF_PROMISC
  52 #endif /* __FreeBSD__ */
  53
  54 struct pcap_netmap {
  55         struct nm_desc *d;      /* pointer returned by nm_open() */
  56         pcap_handler cb;        /* callback and argument */
  57         u_char *cb_arg;
  58         int must_clear_promisc; /* flag */
  59         uint64_t rx_pkts;       /* # of pkts received before the filter */
  60 };
  61
  62
  63 static int
  64 pcap_netmap_stats(pcap_t *p, struct pcap_stat *ps)
  65 {
  66         struct pcap_netmap *pn = p->priv;
  67
  68         ps->ps_recv = (u_int)pn->rx_pkts;
  69         ps->ps_drop = 0;
  70         ps->ps_ifdrop = 0;
  71         return 0;
  72 }
  73
  74
  75 static void
  76 pcap_netmap_filter(u_char *arg, struct pcap_pkthdr *h, const u_char *buf)
  77 {
  78         pcap_t *p = (pcap_t *)arg;
  79         struct pcap_netmap *pn = p->priv;
  80         const struct bpf_insn *pc = p->fcode.bf_insns;
  81
  82         ++pn->rx_pkts;
  83         if (pc == NULL || pcapint_filter(pc, buf, h->len, h->caplen))
  84                 pn->cb(pn->cb_arg, h, buf);
  85 }
  86
  87
  88 static int
  89 pcap_netmap_dispatch(pcap_t *p, int cnt, pcap_handler cb, u_char *user)
  90 {
  91         int ret;
  92         struct pcap_netmap *pn = p->priv;
  93         struct nm_desc *d = pn->d;
  94         struct pollfd pfd = { .fd = p->fd, .events = POLLIN, .revents = 0 };
  95
  96         pn->cb = cb;
  97         pn->cb_arg = user;
  98
  99         for (;;) {
 100                 if (p->break_loop) {
 101                         p->break_loop = 0;
 102                         return PCAP_ERROR_BREAK;
 103                 }
 104                 /* nm_dispatch won't run forever */
 105
 106                 ret = nm_dispatch((void *)d, cnt, (void *)pcap_netmap_filter, (void *)p);
 107                 if (ret != 0)
 108                         break;
 109                 errno = 0;
 110                 ret = poll(&pfd, 1, p->opt.timeout);
 111         }
 112         return ret;
 113 }
 114
 115
 116 /* XXX need to check the NIOCTXSYNC/poll */
 117 static int
 118 pcap_netmap_inject(pcap_t *p, const void *buf, int size)
 119 {
 120         struct pcap_netmap *pn = p->priv;
 121         struct nm_desc *d = pn->d;
 122
 123         return nm_inject(d, buf, size);
 124 }
 125
 126
 127 static int
 128 pcap_netmap_ioctl(pcap_t *p, u_long what, uint32_t *if_flags)
 129 {
 130         struct pcap_netmap *pn = p->priv;
 131         struct nm_desc *d = pn->d;
 132         struct ifreq ifr;
 133         int error, fd = d->fd;
 134
 135 #ifdef __linux__
 136         fd = socket(AF_INET, SOCK_DGRAM, 0);
 137         if (fd < 0) {
 138                 fprintf(stderr, "Error: cannot get device control socket.\n");
 139                 return -1;
 140         }
 141 #endif /* __linux__ */
 142         bzero(&ifr, sizeof(ifr));
 143         /*
 144          * ifreq.ifr_name and nmreq.nr_name have the same size and both
 145          * contain a NUL-terminated string.
 146          */
 147         (void)pcapint_strlcpy(ifr.ifr_name, d->req.nr_name, sizeof(ifr.ifr_name));
 148         switch (what) {
 149         case SIOCSIFFLAGS:
 150                 /*
 151                  * The flags we pass in are 32-bit and unsigned.
 152                  *
 153                  * On most if not all UN*Xes, ifr_flags is 16-bit and
 154                  * signed, and the result of assigning a longer
 155                  * unsigned value to a shorter signed value is
 156                  * implementation-defined (even if, in practice, it'll
 157                  * do what's intended on all platforms we support
 158                  * result of assigning a 32-bit unsigned value).
 159                  * So we mask out the upper 16 bits.
 160                  */
 161                 ifr.ifr_flags = *if_flags & 0xffff;
 162 #ifdef __FreeBSD__
 163                 /*
 164                  * In FreeBSD, we need to set the high-order flags,
 165                  * as we're using IFF_PPROMISC, which is in those bits.
 166                  *
 167                  * XXX - DragonFly BSD?
 168                  */
 169                 ifr.ifr_flagshigh = *if_flags >> 16;
 170 #endif /* __FreeBSD__ */
 171                 break;
 172         }
 173         error = ioctl(fd, what, &ifr);
 174         if (!error) {
 175                 switch (what) {
 176                 case SIOCGIFFLAGS:
 177                         /*
 178                          * The flags we return are 32-bit.
 179                          *
 180                          * On most if not all UN*Xes, ifr_flags is
 181                          * 16-bit and signed, and will get sign-
 182                          * extended, so that the upper 16 bits of
 183                          * those flags will be forced on.  So we
 184                          * mask out the upper 16 bits of the
 185                          * sign-extended value.
 186                          */
 187                         *if_flags = ifr.ifr_flags & 0xffff;
 188 #ifdef __FreeBSD__
 189                         /*
 190                          * In FreeBSD, we need to return the
 191                          * high-order flags, as we're using
 192                          * IFF_PPROMISC, which is in those bits.
 193                          *
 194                          * XXX - DragonFly BSD?
 195                          */
 196                         *if_flags |= (ifr.ifr_flagshigh << 16);
 197 #endif /* __FreeBSD__ */
 198                 }
 199         }
 200 #ifdef __linux__
 201         close(fd);
 202 #endif /* __linux__ */
 203         return error ? -1 : 0;
 204 }
 205
 206
 207 static void
 208 pcap_netmap_close(pcap_t *p)
 209 {
 210         struct pcap_netmap *pn = p->priv;
 211         struct nm_desc *d = pn->d;
 212         uint32_t if_flags = 0;
 213
 214         if (pn->must_clear_promisc) {
 215                 pcap_netmap_ioctl(p, SIOCGIFFLAGS, &if_flags); /* fetch flags */
 216                 if (if_flags & IFF_PPROMISC) {
 217                         if_flags &= ~IFF_PPROMISC;
 218                         pcap_netmap_ioctl(p, SIOCSIFFLAGS, &if_flags);
 219                 }
 220         }
 221         nm_close(d);
 222         pcapint_cleanup_live_common(p);
 223 }
 224
 225
 226 static int
 227 pcap_netmap_activate(pcap_t *p)
 228 {
 229         struct pcap_netmap *pn = p->priv;
 230         struct nm_desc *d;
 231         uint32_t if_flags = 0;
 232
 233         d = nm_open(p->opt.device, NULL, 0, NULL);
 234         if (d == NULL) {
 235                 pcapint_fmt_errmsg_for_errno(p->errbuf, PCAP_ERRBUF_SIZE,
 236                     errno, "netmap open: cannot access %s",
 237                     p->opt.device);
 238                 pcapint_cleanup_live_common(p);
 239                 return (PCAP_ERROR);
 240         }
 241 #if 0
 242         fprintf(stderr, "%s device %s priv %p fd %d ports %d..%d\n",
 243             __func__, p->opt.device, d, d->fd,
 244             d->first_rx_ring, d->last_rx_ring);
 245 #endif
 246         pn->d = d;
 247         p->fd = d->fd;
 248
 249         /*
 250          * Turn a negative snapshot value (invalid), a snapshot value of
 251          * 0 (unspecified), or a value bigger than the normal maximum
 252          * value, into the maximum allowed value.
 253          *
 254          * If some application really *needs* a bigger snapshot
 255          * length, we should just increase MAXIMUM_SNAPLEN.
 256          */
 257         if (p->snapshot <= 0 || p->snapshot > MAXIMUM_SNAPLEN)
 258                 p->snapshot = MAXIMUM_SNAPLEN;
 259
 260         if (p->opt.promisc && !(d->req.nr_ringid & NETMAP_SW_RING)) {
 261                 pcap_netmap_ioctl(p, SIOCGIFFLAGS, &if_flags); /* fetch flags */
 262                 if (!(if_flags & IFF_PPROMISC)) {
 263                         pn->must_clear_promisc = 1;
 264                         if_flags |= IFF_PPROMISC;
 265                         pcap_netmap_ioctl(p, SIOCSIFFLAGS, &if_flags);
 266                 }
 267         }
 268         p->linktype = DLT_EN10MB;
 269         p->selectable_fd = p->fd;
 270         p->read_op = pcap_netmap_dispatch;
 271         p->inject_op = pcap_netmap_inject;
 272         p->setfilter_op = pcapint_install_bpf_program;
 273         p->setdirection_op = NULL;
 274         p->set_datalink_op = NULL;
 275         p->getnonblock_op = pcapint_getnonblock_fd;
 276         p->setnonblock_op = pcapint_setnonblock_fd;
 277         p->stats_op = pcap_netmap_stats;
 278         p->cleanup_op = pcap_netmap_close;
 279
 280         return (0);
 281 }
 282
 283
 284 pcap_t *
 285 pcap_netmap_create(const char *device, char *ebuf, int *is_ours)
 286 {
 287         pcap_t *p;
 288
 289         *is_ours = (!strncmp(device, "netmap:", 7) || !strncmp(device, "vale", 4));
 290         if (! *is_ours)
 291                 return NULL;
 292         p = PCAP_CREATE_COMMON(ebuf, struct pcap_netmap);
 293         if (p == NULL)
 294                 return (NULL);
 295         p->activate_op = pcap_netmap_activate;
 296         return (p);
 297 }
 298
 299 /*
 300  * The "device name" for netmap devices isn't a name for a device, it's
 301  * an expression that indicates how the device should be set up, so
 302  * there's no way to enumerate them.
 303  */
 304 int
 305 pcap_netmap_findalldevs(pcap_if_list_t *devlistp _U_, char *err_str _U_)
 306 {
 307         return 0;
 308 }