Add support for sending packets; includes contributions from Mark

[libpcap] / pcap-bpf.c

/*
 * Copyright (c) 1993, 1994, 1995, 1996, 1998
 *      The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that: (1) source code distributions
 * retain the above copyright notice and this paragraph in its entirety, (2)
 * distributions including binary code include the above copyright notice and
 * this paragraph in its entirety in the documentation or other materials
 * provided with the distribution, and (3) all advertising materials mentioning
 * features or use of this software display the following acknowledgement:
 * ``This product includes software developed by the University of California,
 * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
 * the University nor the names of its contributors may be used to endorse
 * or promote products derived from this software without specific prior
 * written permission.
 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
 * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
 */
#ifndef lint
static const char rcsid[] _U_ =
    "@(#) $Header: /tcpdump/master/libpcap/pcap-bpf.c,v 1.74 2004-03-23 19:18:04 guy Exp $ (LBL)";
#endif

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include <sys/param.h>                  /* optionally get BSD define */
#include <sys/time.h>
#include <sys/timeb.h>
#include <sys/socket.h>
#include <sys/file.h>
#include <sys/ioctl.h>
#include <sys/utsname.h>

#include <net/if.h>

#ifdef _AIX

/*
 * Make "pcap.h" not include "pcap-bpf.h"; we are going to include the
 * native OS version, as we need "struct bpf_config" from it.
 */
#define PCAP_DONT_INCLUDE_PCAP_BPF_H

#include <sys/types.h>

/*
 * Prevent bpf.h from redefining the DLT_ values to their
 * IFT_ values, as we're going to return the standard libpcap
 * values, not IBM's non-standard IFT_ values.
 */
#undef _AIX
#include <net/bpf.h>
#define _AIX

#include <net/if_types.h>               /* for IFT_ values */
#include <sys/sysconfig.h>
#include <sys/device.h>
#include <odmi.h>
#include <cf.h>

#ifdef __64BIT__
#define domakedev makedev64
#define getmajor major64
#define bpf_hdr bpf_hdr32
#else /* __64BIT__ */
#define domakedev makedev
#define getmajor major
#endif /* __64BIT__ */

#define BPF_NAME "bpf"
#define BPF_MINORS 4
#define DRIVER_PATH "/usr/lib/drivers"
#define BPF_NODE "/dev/bpf"
static int bpfloadedflag = 0;
static int odmlockid = 0;

#else /* _AIX */

#include <net/bpf.h>

#endif /* _AIX */

#include <ctype.h>
#include <errno.h>
#include <netdb.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>

#include "pcap-int.h"

#ifdef HAVE_DAG_API
#include "pcap-dag.h"
#endif /* HAVE_DAG_API */

#ifdef HAVE_OS_PROTO_H
#include "os-proto.h"
#endif

#include "gencode.h"    /* for "no_optimize" */

static int pcap_setfilter_bpf(pcap_t *p, struct bpf_program *fp);
static int pcap_set_datalink_bpf(pcap_t *p, int dlt);

static int
pcap_stats_bpf(pcap_t *p, struct pcap_stat *ps)
{
        struct bpf_stat s;

        /*
         * "ps_recv" counts packets handed to the filter, not packets
         * that passed the filter.  This includes packets later dropped
         * because we ran out of buffer space.
         *
         * "ps_drop" counts packets dropped inside the BPF device
         * because we ran out of buffer space.  It doesn't count
         * packets dropped by the interface driver.  It counts
         * only packets that passed the filter.
         *
         * Both statistics include packets not yet read from the kernel
         * by libpcap, and thus not yet seen by the application.
         */
        if (ioctl(p->fd, BIOCGSTATS, (caddr_t)&s) < 0) {
                snprintf(p->errbuf, PCAP_ERRBUF_SIZE, "BIOCGSTATS: %s",
                    pcap_strerror(errno));
                return (-1);
        }

        ps->ps_recv = s.bs_recv;
        ps->ps_drop = s.bs_drop;
        return (0);
}

static int
pcap_read_bpf(pcap_t *p, int cnt, pcap_handler callback, u_char *user)
{
        int cc;
        int n = 0;
        register u_char *bp, *ep;
        struct bpf_insn *fcode;

        fcode = p->md.use_bpf ? NULL : p->fcode.bf_insns;
 again:
        /*
         * Has "pcap_breakloop()" been called?
         */
        if (p->break_loop) {
                /*
                 * Yes - clear the flag that indicates that it
                 * has, and return -2 to indicate that we were
                 * told to break out of the loop.
                 */
                p->break_loop = 0;
                return (-2);
        }
        cc = p->cc;
        if (p->cc == 0) {
                cc = read(p->fd, (char *)p->buffer, p->bufsize);
                if (cc < 0) {
                        /* Don't choke when we get ptraced */
                        switch (errno) {

                        case EINTR:
                                goto again;

#ifdef _AIX
                        case EFAULT:
                                /*
                                 * Sigh.  More AIX wonderfulness.
                                 *
                                 * For some unknown reason the uiomove()
                                 * operation in the bpf kernel extension
                                 * used to copy the buffer into user 
                                 * space sometimes returns EFAULT. I have
                                 * no idea why this is the case given that
                                 * a kernel debugger shows the user buffer 
                                 * is correct. This problem appears to 
                                 * be mostly mitigated by the memset of 
                                 * the buffer before it is first used. 
                                 * Very strange.... Shaun Clowes
                                 *
                                 * In any case this means that we shouldn't 
                                 * treat EFAULT as a fatal error; as we
                                 * don't have an API for returning
                                 * a "some packets were dropped since
                                 * the last packet you saw" indication,
                                 * we just ignore EFAULT and keep reading.
                                 */
                                goto again;
#endif 
  
                        case EWOULDBLOCK:
                                return (0);
#if defined(sun) && !defined(BSD)
                        /*
                         * Due to a SunOS bug, after 2^31 bytes, the kernel
                         * file offset overflows and read fails with EINVAL.
                         * The lseek() to 0 will fix things.
                         */
                        case EINVAL:
                                if (lseek(p->fd, 0L, SEEK_CUR) +
                                    p->bufsize < 0) {
                                        (void)lseek(p->fd, 0L, SEEK_SET);
                                        goto again;
                                }
                                /* fall through */
#endif
                        }
                        snprintf(p->errbuf, PCAP_ERRBUF_SIZE, "read: %s",
                            pcap_strerror(errno));
                        return (-1);
                }
                bp = p->buffer;
        } else
                bp = p->bp;

        /*
         * Loop through each packet.
         */
#define bhp ((struct bpf_hdr *)bp)
        ep = bp + cc;
        while (bp < ep) {
                register int caplen, hdrlen;

                /*
                 * Has "pcap_breakloop()" been called?
                 * If so, return immediately - if we haven't read any
                 * packets, clear the flag and return -2 to indicate
                 * that we were told to break out of the loop, otherwise
                 * leave the flag set, so that the *next* call will break
                 * out of the loop without having read any packets, and
                 * return the number of packets we've processed so far.
                 */
                if (p->break_loop) {
                        if (n == 0) {
                                p->break_loop = 0;
                                return (-2);
                        } else {
                                p->bp = bp;
                                p->cc = ep - bp;
                                return (n);
                        }
                }

                caplen = bhp->bh_caplen;
                hdrlen = bhp->bh_hdrlen;
                /*
                 * Short-circuit evaluation: if using BPF filter
                 * in kernel, no need to do it now.
                 */
                if (fcode == NULL ||
                    bpf_filter(fcode, bp + hdrlen, bhp->bh_datalen, caplen)) {
#ifdef _AIX
                        /*
                         * AIX's BPF returns seconds/nanoseconds time
                         * stamps, not seconds/microseconds time stamps.
                         *
                         * XXX - I'm guessing here that it's a "struct
                         * timestamp"; if not, this code won't compile,
                         * but, if not, you want to send us a bug report
                         * and fall back on using DLPI.  It's not as if
                         * BPF used to work right on AIX before this
                         * change; this change attempts to fix the fact
                         * that it didn't....
                         */
                        bhp->bh_tstamp.tv_usec = bhp->bh_tstamp.tv_usec/1000;
#endif
                        /*
                         * XXX A bpf_hdr matches a pcap_pkthdr.
                         */
                        (*callback)(user, (struct pcap_pkthdr*)bp, bp + hdrlen);
                        bp += BPF_WORDALIGN(caplen + hdrlen);
                        if (++n >= cnt && cnt > 0) {
                                p->bp = bp;
                                p->cc = ep - bp;
                                return (n);
                        }
                } else {
                        /*
                         * Skip this packet.
                         */
                        bp += BPF_WORDALIGN(caplen + hdrlen);
                }
        }
#undef bhp
        p->cc = 0;
        return (n);
}

static int
pcap_inject_bpf(pcap_t *p, const void *buf, size_t size)
{
        int ret;

        /*
         * Do a BIOCSHDRCMPLT, if defined, to turn that flag on, so
         * the link-layer source address isn't forcibly overwritten?
         * (Ignore errors?  Return errors if not "sorry, that ioctl
         * isn't supported?)
         *
         * XXX - I seem to remember some packet-sending bug in some
         * BSDs - check CVS log for "bpf.c"?
         */
        ret = write(p->fd, buf, size);
        if (ret == -1) {
                snprintf(p->errbuf, PCAP_ERRBUF_SIZE, "send: %s",
                    pcap_strerror(errno));
                return (-1);
        }
        return (ret);
}

#ifdef _AIX
static int 
bpf_odminit(char *errbuf)
{
        char *errstr;

        if (odm_initialize() == -1) {
                if (odm_err_msg(odmerrno, &errstr) == -1)
                        errstr = "Unknown error";
                snprintf(errbuf, PCAP_ERRBUF_SIZE,
                    "bpf_load: odm_initialize failed: %s",
                    errstr);
                return (-1);
        }

        if ((odmlockid = odm_lock("/etc/objrepos/config_lock", ODM_WAIT)) == -1) {
                if (odm_err_msg(odmerrno, &errstr) == -1)
                        errstr = "Unknown error";
                snprintf(errbuf, PCAP_ERRBUF_SIZE,
                    "bpf_load: odm_lock of /etc/objrepos/config_lock failed: %s",
                    errstr);
                return (-1);
        }

        return (0);
}

static int 
bpf_odmcleanup(char *errbuf)
{
        char *errstr;

        if (odm_unlock(odmlockid) == -1) {
                if (odm_err_msg(odmerrno, &errstr) == -1)
                        errstr = "Unknown error";
                snprintf(errbuf, PCAP_ERRBUF_SIZE,
                    "bpf_load: odm_unlock failed: %s",
                    errstr);
                return (-1);
        }

        if (odm_terminate() == -1) {
                if (odm_err_msg(odmerrno, &errstr) == -1)
                        errstr = "Unknown error";
                snprintf(errbuf, PCAP_ERRBUF_SIZE,
                    "bpf_load: odm_terminate failed: %s",
                    errstr);
                return (-1);
        }

        return (0);
}

static int
bpf_load(char *errbuf)
{
        long major;
        int *minors;
        int numminors, i, rc;
        char buf[1024];
        struct stat sbuf;
        struct bpf_config cfg_bpf;
        struct cfg_load cfg_ld;
        struct cfg_kmod cfg_km;

        /*
         * This is very very close to what happens in the real implementation
         * but I've fixed some (unlikely) bug situations.
         */
        if (bpfloadedflag)
                return (0);

        if (bpf_odminit(errbuf) != 0)
                return (-1);

        major = genmajor(BPF_NAME);
        if (major == -1) {
                snprintf(errbuf, PCAP_ERRBUF_SIZE,
                    "bpf_load: genmajor failed: %s", pcap_strerror(errno));
                return (-1);
        }

        minors = getminor(major, &numminors, BPF_NAME);
        if (!minors) {
                minors = genminor("bpf", major, 0, BPF_MINORS, 1, 1);
                if (!minors) {
                        snprintf(errbuf, PCAP_ERRBUF_SIZE,
                            "bpf_load: genminor failed: %s",
                            pcap_strerror(errno));
                        return (-1);
                }
        }

        if (bpf_odmcleanup(errbuf))
                return (-1);

        rc = stat(BPF_NODE "0", &sbuf);
        if (rc == -1 && errno != ENOENT) {
                snprintf(errbuf, PCAP_ERRBUF_SIZE,
                    "bpf_load: can't stat %s: %s",
                    BPF_NODE "0", pcap_strerror(errno));
                return (-1);
        }

        if (rc == -1 || getmajor(sbuf.st_rdev) != major) {
                for (i = 0; i < BPF_MINORS; i++) {
                        sprintf(buf, "%s%d", BPF_NODE, i);
                        unlink(buf);
                        if (mknod(buf, S_IRUSR | S_IFCHR, domakedev(major, i)) == -1) {
                                snprintf(errbuf, PCAP_ERRBUF_SIZE,
                                    "bpf_load: can't mknod %s: %s",
                                    buf, pcap_strerror(errno));
                                return (-1);
                        }
                }
        }

        /* Check if the driver is loaded */
        memset(&cfg_ld, 0x0, sizeof(cfg_ld));
        cfg_ld.path = buf;
        sprintf(cfg_ld.path, "%s/%s", DRIVER_PATH, BPF_NAME);
        if ((sysconfig(SYS_QUERYLOAD, (void *)&cfg_ld, sizeof(cfg_ld)) == -1) ||
            (cfg_ld.kmid == 0)) {
                /* Driver isn't loaded, load it now */
                if (sysconfig(SYS_SINGLELOAD, (void *)&cfg_ld, sizeof(cfg_ld)) == -1) {
                        snprintf(errbuf, PCAP_ERRBUF_SIZE,
                            "bpf_load: could not load driver: %s",
                            strerror(errno));
                        return (-1);
                }
        }

        /* Configure the driver */
        cfg_km.cmd = CFG_INIT;
        cfg_km.kmid = cfg_ld.kmid;
        cfg_km.mdilen = sizeof(cfg_bpf);
        cfg_km.mdiptr = (void *)&cfg_bpf; 
        for (i = 0; i < BPF_MINORS; i++) {
                cfg_bpf.devno = domakedev(major, i);
                if (sysconfig(SYS_CFGKMOD, (void *)&cfg_km, sizeof(cfg_km)) == -1) {
                        snprintf(errbuf, PCAP_ERRBUF_SIZE,
                            "bpf_load: could not configure driver: %s",
                            strerror(errno));
                        return (-1);
                }
        }
        
        bpfloadedflag = 1;

        return (0);
}
#endif

static inline int
bpf_open(pcap_t *p, char *errbuf)
{
        int fd;
        int n = 0;
        char device[sizeof "/dev/bpf0000000000"];

#ifdef _AIX
        /*
         * Load the bpf driver, if it isn't already loaded,
         * and create the BPF device entries, if they don't
         * already exist.
         */
        if (bpf_load(errbuf) == -1)
                return (-1);
#endif

        /*
         * Go through all the minors and find one that isn't in use.
         */
        do {
                (void)snprintf(device, sizeof(device), "/dev/bpf%d", n++);
                /*
                 * Initially try a read/write open (to allow the inject
                 * method to work).  If that fails due to permission
                 * issues, fall back to read-only.  This allows a
                 * non-root user to be granted specific access to pcap
                 * capabilities via file permissions.
                 *
                 * XXX - we should have an API that has a flag that
                 * controls whether to open read-only or read-write,
                 * so that denial of permission to send (or inability
                 * to send, if sending packets isn't supported on
                 * the device in question) can be indicated at open
                 * time.
                 */
                fd = open(device, O_RDWR);
                if (fd == -1 && errno == EACCES)
                        fd = open(device, O_RDONLY);
        } while (fd < 0 && errno == EBUSY);

        /*
         * XXX better message for all minors used
         */
        if (fd < 0)
                snprintf(errbuf, PCAP_ERRBUF_SIZE, "(no devices found) %s: %s",
                    device, pcap_strerror(errno));

        return (fd);
}

static void
pcap_close_bpf(pcap_t *p)
{
        if (p->buffer != NULL)
                free(p->buffer);
        if (p->fd >= 0)
                close(p->fd);
}

/*
 * We include the OS's <net/bpf.h>, not our "pcap-bpf.h", so we probably
 * don't get DLT_DOCSIS defined.
 */
#ifndef DLT_DOCSIS
#define DLT_DOCSIS      143
#endif

/*
 * XXX - on AIX, IBM's tcpdump (and perhaps the incompatible-with-everybody-
 * else's libpcap in AIX 5.1) appears to forcibly load the BPF driver
 * if it's not already loaded, and to create the BPF devices if they
 * don't exist.
 *
 * It'd be nice if we could do the same, although the code to do so
 * might be version-dependent, alas (the way to do it isn't necessarily
 * documented).
 */
pcap_t *
pcap_open_live(const char *device, int snaplen, int promisc, int to_ms,
    char *ebuf)
{
        int fd;
        struct ifreq ifr;
        struct bpf_version bv;
#ifdef BIOCGDLTLIST
        struct bpf_dltlist bdl;
#endif
        u_int v;
        pcap_t *p;
        struct utsname osinfo;

#ifdef HAVE_DAG_API
        if (strstr(device, "dag")) {
                return dag_open_live(device, snaplen, promisc, to_ms, ebuf);
        }
#endif /* HAVE_DAG_API */

#ifdef BIOCGDLTLIST
        memset(&bdl, 0, sizeof(bdl));
#endif

        p = (pcap_t *)malloc(sizeof(*p));
        if (p == NULL) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE, "malloc: %s",
                    pcap_strerror(errno));
                return (NULL);
        }
        memset(p, 0, sizeof(*p));
        fd = bpf_open(p, ebuf);
        if (fd < 0)
                goto bad;

        p->fd = fd;
        p->snapshot = snaplen;

        if (ioctl(fd, BIOCVERSION, (caddr_t)&bv) < 0) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE, "BIOCVERSION: %s",
                    pcap_strerror(errno));
                goto bad;
        }
        if (bv.bv_major != BPF_MAJOR_VERSION ||
            bv.bv_minor < BPF_MINOR_VERSION) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE,
                    "kernel bpf filter out of date");
                goto bad;
        }

        /*
         * Try finding a good size for the buffer; 32768 may be too
         * big, so keep cutting it in half until we find a size
         * that works, or run out of sizes to try.  If the default
         * is larger, don't make it smaller.
         *
         * XXX - there should be a user-accessible hook to set the
         * initial buffer size.
         */
        if ((ioctl(fd, BIOCGBLEN, (caddr_t)&v) < 0) || v < 32768)
                v = 32768;
        for ( ; v != 0; v >>= 1) {
                /* Ignore the return value - this is because the call fails
                 * on BPF systems that don't have kernel malloc.  And if
                 * the call fails, it's no big deal, we just continue to
                 * use the standard buffer size.
                 */
                (void) ioctl(fd, BIOCSBLEN, (caddr_t)&v);

                (void)strncpy(ifr.ifr_name, device, sizeof(ifr.ifr_name));
                if (ioctl(fd, BIOCSETIF, (caddr_t)&ifr) >= 0)
                        break;  /* that size worked; we're done */

                if (errno != ENOBUFS) {
                        snprintf(ebuf, PCAP_ERRBUF_SIZE, "BIOCSETIF: %s: %s",
                            device, pcap_strerror(errno));
                        goto bad;
                }
        }

        if (v == 0) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE,
                         "BIOCSBLEN: %s: No buffer size worked", device);
                goto bad;
        }

        /* Get the data link layer type. */
        if (ioctl(fd, BIOCGDLT, (caddr_t)&v) < 0) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE, "BIOCGDLT: %s",
                    pcap_strerror(errno));
                goto bad;
        }
#ifdef _AIX
        /*
         * AIX's BPF returns IFF_ types, not DLT_ types, in BIOCGDLT.
         */
        switch (v) {

        case IFT_ETHER:
        case IFT_ISO88023:
                v = DLT_EN10MB;
                break;

        case IFT_FDDI:
                v = DLT_FDDI;
                break;

        case IFT_ISO88025:
                v = DLT_IEEE802;
                break;

        case IFT_LOOP:
                v = DLT_NULL;
                break;

        default:
                /*
                 * We don't know what to map this to yet.
                 */
                snprintf(ebuf, PCAP_ERRBUF_SIZE, "unknown interface type %u",
                    v);
                goto bad;
        }
#endif
#if _BSDI_VERSION - 0 >= 199510
        /* The SLIP and PPP link layer header changed in BSD/OS 2.1 */
        switch (v) {

        case DLT_SLIP:
                v = DLT_SLIP_BSDOS;
                break;

        case DLT_PPP:
                v = DLT_PPP_BSDOS;
                break;

        case 11:        /*DLT_FR*/
                v = DLT_FRELAY;
                break;

        case 12:        /*DLT_C_HDLC*/
                v = DLT_CHDLC;
                break;
        }
#endif
        p->linktype = v;

#ifdef BIOCGDLTLIST
        /*
         * We know the default link type -- now determine all the DLTs
         * this interface supports.  If this fails with EINVAL, it's
         * not fatal; we just don't get to use the feature later.
         */
        if (ioctl(fd, BIOCGDLTLIST, (caddr_t)&bdl) == 0) {
                u_int i;
                int is_ethernet;

                bdl.bfl_list = (u_int *) malloc(sizeof(u_int) * bdl.bfl_len + 1);
                if (bdl.bfl_list == NULL) {
                        (void)snprintf(ebuf, PCAP_ERRBUF_SIZE, "malloc: %s",
                            pcap_strerror(errno));
                        goto bad;
                }

                if (ioctl(fd, BIOCGDLTLIST, (caddr_t)&bdl) < 0) {
                        (void)snprintf(ebuf, PCAP_ERRBUF_SIZE,
                            "BIOCGDLTLIST: %s", pcap_strerror(errno));
                        free(bdl.bfl_list);
                        goto bad;
                }

                /*
                 * OK, for real Ethernet devices, add DLT_DOCSIS to the
                 * list, so that an application can let you choose it,
                 * in case you're capturing DOCSIS traffic that a Cisco
                 * Cable Modem Termination System is putting out onto
                 * an Ethernet (it doesn't put an Ethernet header onto
                 * the wire, it puts raw DOCSIS frames out on the wire
                 * inside the low-level Ethernet framing).
                 *
                 * A "real Ethernet device" is defined here as a device
                 * that has a link-layer type of DLT_EN10MB and that has
                 * no alternate link-layer types; that's done to exclude
                 * 802.11 interfaces (which might or might not be the
                 * right thing to do, but I suspect it is - Ethernet <->
                 * 802.11 bridges would probably badly mishandle frames
                 * that don't have Ethernet headers).
                 */
                if (p->linktype == DLT_EN10MB) {
                        is_ethernet = 1;
                        for (i = 0; i < bdl.bfl_len; i++) {
                                if (bdl.bfl_list != DLT_EN10MB) {
                                        is_ethernet = 0;
                                        break;
                                }
                        }
                        if (is_ethernet) {
                                /*
                                 * We reserved one more slot at the end of
                                 * the list.
                                 */
                                bdl.bfl_list[bdl.bfl_len] = DLT_DOCSIS;
                                bdl.bfl_len++;
                        }
                }
                p->dlt_count = bdl.bfl_len;
                p->dlt_list = bdl.bfl_list;
        } else {
                if (errno != EINVAL) {
                        (void)snprintf(ebuf, PCAP_ERRBUF_SIZE,
                            "BIOCGDLTLIST: %s", pcap_strerror(errno));
                        goto bad;
                }
        }
#endif

        /*
         * If this is an Ethernet device, and we don't have a DLT_ list,
         * give it a list with DLT_EN10MB and DLT_DOCSIS.  (That'd give
         * 802.11 interfaces DLT_DOCSIS, which isn't the right thing to
         * do, but there's not much we can do about that without finding
         * some other way of determining whether it's an Ethernet or 802.11
         * device.)
         */
        if (p->linktype == DLT_EN10MB && p->dlt_count == 0) {
                p->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
                /*
                 * If that fails, just leave the list empty.
                 */
                if (p->dlt_list != NULL) {
                        p->dlt_list[0] = DLT_EN10MB;
                        p->dlt_list[1] = DLT_DOCSIS;
                        p->dlt_count = 2;
                }
        }
                
        /* set timeout */
        if (to_ms != 0) {
                /*
                 * XXX - is this seconds/nanoseconds in AIX?
                 * (Treating it as such doesn't fix the timeout
                 * problem described below.)
                 */
                struct timeval to;
                to.tv_sec = to_ms / 1000;
                to.tv_usec = (to_ms * 1000) % 1000000;
                if (ioctl(p->fd, BIOCSRTIMEOUT, (caddr_t)&to) < 0) {
                        snprintf(ebuf, PCAP_ERRBUF_SIZE, "BIOCSRTIMEOUT: %s",
                            pcap_strerror(errno));
                        goto bad;
                }
        }

#ifdef _AIX
#ifdef  BIOCIMMEDIATE
        /*
         * Darren Reed notes that
         *
         *      On AIX (4.2 at least), if BIOCIMMEDIATE is not set, the
         *      timeout appears to be ignored and it waits until the buffer
         *      is filled before returning.  The result of not having it
         *      set is almost worse than useless if your BPF filter
         *      is reducing things to only a few packets (i.e. one every
         *      second or so).
         *
         * so we turn BIOCIMMEDIATE mode on if this is AIX.
         *
         * We don't turn it on for other platforms, as that means we
         * get woken up for every packet, which may not be what we want;
         * in the Winter 1993 USENIX paper on BPF, they say:
         *
         *      Since a process might want to look at every packet on a
         *      network and the time between packets can be only a few
         *      microseconds, it is not possible to do a read system call
         *      per packet and BPF must collect the data from several
         *      packets and return it as a unit when the monitoring
         *      application does a read.
         *
         * which I infer is the reason for the timeout - it means we
         * wait that amount of time, in the hopes that more packets
         * will arrive and we'll get them all with one read.
         *
         * Setting BIOCIMMEDIATE mode on FreeBSD (and probably other
         * BSDs) causes the timeout to be ignored.
         *
         * On the other hand, some platforms (e.g., Linux) don't support
         * timeouts, they just hand stuff to you as soon as it arrives;
         * if that doesn't cause a problem on those platforms, it may
         * be OK to have BIOCIMMEDIATE mode on BSD as well.
         *
         * (Note, though, that applications may depend on the read
         * completing, even if no packets have arrived, when the timeout
         * expires, e.g. GUI applications that have to check for input
         * while waiting for packets to arrive; a non-zero timeout
         * prevents "select()" from working right on FreeBSD and
         * possibly other BSDs, as the timer doesn't start until a
         * "read()" is done, so the timer isn't in effect if the
         * application is blocked on a "select()", and the "select()"
         * doesn't get woken up for a BPF device until the buffer
         * fills up.)
         */
        v = 1;
        if (ioctl(p->fd, BIOCIMMEDIATE, &v) < 0) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE, "BIOCIMMEDIATE: %s",
                    pcap_strerror(errno));
                goto bad;
        }
#endif  /* BIOCIMMEDIATE */
#endif  /* _AIX */

        if (promisc) {
                /* set promiscuous mode, okay if it fails */
                if (ioctl(p->fd, BIOCPROMISC, NULL) < 0) {
                        snprintf(ebuf, PCAP_ERRBUF_SIZE, "BIOCPROMISC: %s",
                            pcap_strerror(errno));
                }
        }

        if (ioctl(fd, BIOCGBLEN, (caddr_t)&v) < 0) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE, "BIOCGBLEN: %s",
                    pcap_strerror(errno));
                goto bad;
        }
        p->bufsize = v;
        p->buffer = (u_char *)malloc(p->bufsize);
        if (p->buffer == NULL) {
                snprintf(ebuf, PCAP_ERRBUF_SIZE, "malloc: %s",
                    pcap_strerror(errno));
                goto bad;
        }
#ifdef _AIX
        /* For some strange reason this seems to prevent the EFAULT 
         * problems we have experienced from AIX BPF. */
        memset(p->buffer, 0x0, p->bufsize);
#endif

        /*
         * On most BPF platforms, either you can do a "select()" or
         * "poll()" on a BPF file descriptor and it works correctly,
         * or you can do it and it will return "readable" if the
         * hold buffer is full but not if the timeout expires *and*
         * a non-blocking read will, if the hold buffer is empty
         * but the store buffer isn't empty, rotate the buffers
         * and return what packets are available.
         *
         * In the latter case, the fact that a non-blocking read
         * will give you the available packets means you can work
         * around the failure of "select()" and "poll()" to wake up
         * and return "readable" when the timeout expires by using
         * the timeout as the "select()" or "poll()" timeout, putting
         * the BPF descriptor into non-blocking mode, and read from
         * it regardless of whether "select()" reports it as readable
         * or not.
         *
         * However, in FreeBSD 4.3 and 4.4, "select()" and "poll()"
         * won't wake up and return "readable" if the timer expires
         * and non-blocking reads return EWOULDBLOCK if the hold
         * buffer is empty, even if the store buffer is non-empty.
         *
         * This means the workaround in question won't work.
         *
         * Therefore, on FreeBSD 4.3 and 4.4, we set "p->selectable_fd"
         * to -1, which means "sorry, you can't use 'select()' or 'poll()'
         * here".  On all other BPF platforms, we set it to the FD for
         * the BPF device; in NetBSD, OpenBSD, and Darwin, a non-blocking
         * read will, if the hold buffer is empty and the store buffer
         * isn't empty, rotate the buffers and return what packets are
         * there (and in sufficiently recent versions of OpenBSD
         * "select()" and "poll()" should work correctly).
         *
         * XXX - what about AIX?
         */
        if (uname(&osinfo) == 0) {
                /*
                 * We can check what OS this is.
                 */
                if (strcmp(osinfo.sysname, "FreeBSD") == 0 &&
                    (strcmp(osinfo.release, "4.3") == 0 ||
                     strcmp(osinfo.release, "4.4") == 0))
                        p->selectable_fd = -1;
                else
                        p->selectable_fd = p->fd;
        } else {
                /*
                 * We can't find out what OS this is, so assume we can
                 * do a "select()" or "poll()".
                 */
                p->selectable_fd = p->fd;
        }

        p->read_op = pcap_read_bpf;
        p->inject_op = pcap_inject_bpf;
        p->setfilter_op = pcap_setfilter_bpf;
        p->set_datalink_op = pcap_set_datalink_bpf;
        p->getnonblock_op = pcap_getnonblock_fd;
        p->setnonblock_op = pcap_setnonblock_fd;
        p->stats_op = pcap_stats_bpf;
        p->close_op = pcap_close_bpf;

        return (p);
 bad:
        (void)close(fd);
        if (p->dlt_list != NULL)
                free(p->dlt_list);
        free(p);
        return (NULL);
}

int
pcap_platform_finddevs(pcap_if_t **alldevsp, char *errbuf)
{
#ifdef HAVE_DAG_API
        if (dag_platform_finddevs(alldevsp, errbuf) < 0)
                return (-1);
#endif /* HAVE_DAG_API */

        return (0);
}

static int
pcap_setfilter_bpf(pcap_t *p, struct bpf_program *fp)
{
        /*
         * It looks that BPF code generated by gen_protochain() is not
         * compatible with some of kernel BPF code (for example BSD/OS 3.1).
         * Take a safer side for now.
         */
        if (no_optimize) {
                /*
                 * XXX - what if we already have a filter in the kernel?
                 */
                if (install_bpf_program(p, fp) < 0)
                        return (-1);
                p->md.use_bpf = 0;      /* filtering in userland */
                return (0);
        }

        /*
         * Free any user-mode filter we might happen to have installed.
         */
        pcap_freecode(&p->fcode);

        /*
         * Try to install the kernel filter.
         */
        if (ioctl(p->fd, BIOCSETF, (caddr_t)fp) < 0) {
                snprintf(p->errbuf, PCAP_ERRBUF_SIZE, "BIOCSETF: %s",
                    pcap_strerror(errno));
                return (-1);
        }
        p->md.use_bpf = 1;      /* filtering in the kernel */
        return (0);
}

static int
pcap_set_datalink_bpf(pcap_t *p, int dlt)
{
#ifdef BIOCSDLT
        if (ioctl(p->fd, BIOCSDLT, &dlt) == -1) {
                (void) snprintf(p->errbuf, sizeof(p->errbuf),
                    "Cannot set DLT %d: %s", dlt, strerror(errno));
                return (-1);
        }
#endif
        return (0);
}