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