]> The Tcpdump Group git mirrors - libpcap/blobdiff - pcap-netfilter-linux.c
CI: Call print_so_deps() on rpcapd in remote enabled build
[libpcap] / pcap-netfilter-linux.c
index 9e61e455dbfefd34f5fd6d898ab5bb0f97def081..90296be71b781dc9a3be4dcf31b9876275a5a960 100644 (file)
@@ -11,8 +11,8 @@
  * 2. Redistributions in binary form must reproduce the above copyright
  * notice, this list of conditions and the following disclaimer in the
  * documentation and/or other materials provided with the distribution.
- * 3. The name of the author may not be used to endorse or promote 
- * products derived from this software without specific prior written 
+ * 3. The name of the author may not be used to endorse or promote
+ * products derived from this software without specific prior written
  * permission.
  *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#ifdef HAVE_CONFIG_H
-#include "config.h"
-#endif
+#include <config.h>
 
 #include "pcap-int.h"
-
-#ifdef NEED_STRERROR_H
-#include "strerror.h"
-#endif
+#include "diag-control.h"
 
 #include <errno.h>
 #include <stdlib.h>
 #include <linux/netfilter/nfnetlink_log.h>
 #include <linux/netfilter/nfnetlink_queue.h>
 
-/* NOTE: if your program drops privilages after pcap_activate() it WON'T work with nfqueue.
- *       It took me quite some time to debug ;/ 
+/* NOTE: if your program drops privileges after pcap_activate() it WON'T work with nfqueue.
+ *       It took me quite some time to debug ;/
  *
- *       Sending any data to nfnetlink socket requires CAP_NET_ADMIN privilages,
- *       and in nfqueue we need to send verdict reply after recving packet.
+ *       Sending any data to nfnetlink socket requires CAP_NET_ADMIN privileges,
+ *       and in nfqueue we need to send verdict reply after receiving packet.
  *
- *       In tcpdump you can disable dropping privilages with -Z root
+ *       In tcpdump you can disable dropping privileges with -Z root
  */
+
 #include "pcap-netfilter-linux.h"
 
 #define HDR_LENGTH (NLMSG_LENGTH(NLMSG_ALIGN(sizeof(struct nfgenmsg))))
@@ -79,47 +74,130 @@ typedef enum { OTHER = -1, NFLOG, NFQUEUE } nftype_t;
  */
 struct pcap_netfilter {
        u_int   packets_read;   /* count of packets read with recvfrom() */
+       u_int   packets_nobufs; /* ENOBUFS counter */
 };
 
-static int nfqueue_send_verdict(const pcap_t *handle, u_int16_t group_id, u_int32_t id, u_int32_t verdict);
+static int nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict);
+
 
 static int
 netfilter_read_linux(pcap_t *handle, int max_packets, pcap_handler callback, u_char *user)
 {
        struct pcap_netfilter *handlep = handle->priv;
-       const unsigned char *buf;
+       register u_char *bp, *ep;
        int count = 0;
-       int len;
+       u_int cc;
 
-       /* ignore interrupt system call error */
-       do {
-               len = recv(handle->fd, handle->buffer, handle->bufsize, 0);
-               if (handle->break_loop) {
-                       handle->break_loop = 0;
-                       return -2;
+       /*
+        * Has "pcap_breakloop()" been called?
+        */
+       if (handle->break_loop) {
+               /*
+                * Yes - clear the flag that indicates that it
+                * has, and return PCAP_ERROR_BREAK to indicate
+                * that we were told to break out of the loop.
+                */
+               handle->break_loop = 0;
+               return PCAP_ERROR_BREAK;
+       }
+       cc = handle->cc;
+       if (cc == 0) {
+               /*
+                * The buffer is empty; refill it.
+                *
+                * We ignore EINTR, as that might just be due to a signal
+                * being delivered - if the signal should interrupt the
+                * loop, the signal handler should call pcap_breakloop()
+                * to set handle->break_loop (we ignore it on other
+                * platforms as well).
+                */
+               ssize_t read_ret;
+
+               do {
+                       read_ret = recv(handle->fd, handle->buffer, handle->bufsize, 0);
+                       if (handle->break_loop) {
+                               handle->break_loop = 0;
+                               return PCAP_ERROR_BREAK;
+                       }
+                       if (read_ret == -1 && errno == ENOBUFS)
+                               handlep->packets_nobufs++;
+               } while ((read_ret == -1) && (errno == EINTR || errno == ENOBUFS));
+
+               if (read_ret < 0) {
+                       pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                           PCAP_ERRBUF_SIZE, errno, "Can't receive packet");
+                       return PCAP_ERROR;
                }
-       } while ((len == -1) && (errno == EINTR));
 
-       if (len < 0) {
-               snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Can't receive packet %d:%s", errno, pcap_strerror(errno));
-               return -1;
-       }
+               /*
+                * At this point, read_ret is guaranteed to be
+                * >= 0 and < p->bufsize; p->bufsize is a u_int,
+                * so its value is guaranteed to fit in cc, which
+                * is also a u_int.
+                */
+               cc = (u_int)read_ret;
+               bp = (unsigned char *)handle->buffer;
+       } else
+               bp = handle->bp;
 
-       buf = handle->buffer;
-       while (len >= NLMSG_SPACE(0)) {
-               const struct nlmsghdr *nlh = (const struct nlmsghdr *) buf;
-               u_int32_t msg_len;
+       /*
+        * Loop through each message.
+        *
+        * This assumes that a single buffer of message will have
+        * <= INT_MAX packets, so the message count doesn't overflow.
+        */
+       ep = bp + cc;
+       while (bp < ep) {
+               const struct nlmsghdr *nlh = (const struct nlmsghdr *) bp;
+               uint32_t msg_len;
                nftype_t type = OTHER;
+               /*
+                * Has "pcap_breakloop()" been called?
+                * If so, return immediately - if we haven't read any
+                * packets, clear the flag and return PCAP_ERROR_BREAK
+                * 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 (handle->break_loop) {
+                       handle->bp = bp;
+                       handle->cc = (u_int)(ep - bp);
+                       if (count == 0) {
+                               handle->break_loop = 0;
+                               return PCAP_ERROR_BREAK;
+                       } else
+                               return count;
+               }
+               /*
+                * NLMSG_SPACE(0) might be signed or might be unsigned,
+                * depending on whether the kernel defines NLMSG_ALIGNTO
+                * as 4, which older kernels do, or as 4U, which newer
+                * kernels do.
+                *
+                * ep - bp is of type ptrdiff_t, which is signed.
+                *
+                * To squelch warnings, we cast both to size_t, which
+                * is unsigned; ep >= bp, so the cast is safe.
+                */
+               if ((size_t)(ep - bp) < (size_t)NLMSG_SPACE(0)) {
+                       /*
+                        * There's less than one netlink message left
+                        * in the buffer.  Give up.
+                        */
+                       break;
+               }
 
-               if (nlh->nlmsg_len < sizeof(struct nlmsghdr) || len < nlh->nlmsg_len) {
-                       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Message truncated: (got: %d) (nlmsg_len: %u)", len, nlh->nlmsg_len);
+               if (nlh->nlmsg_len < sizeof(struct nlmsghdr) || cc < nlh->nlmsg_len) {
+                       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Message truncated: (got: %u) (nlmsg_len: %u)", cc, nlh->nlmsg_len);
                        return -1;
                }
 
-               if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_ULOG && 
-                   NFNL_MSG_TYPE(nlh->nlmsg_type) == NFULNL_MSG_PACKET) 
+               if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_ULOG &&
+                   NFNL_MSG_TYPE(nlh->nlmsg_type) == NFULNL_MSG_PACKET)
                        type = NFLOG;
-               else if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_QUEUE && 
+               else if (NFNL_SUBSYS_ID(nlh->nlmsg_type) == NFNL_SUBSYS_QUEUE &&
                         NFNL_MSG_TYPE(nlh->nlmsg_type) == NFQNL_MSG_PACKET)
                        type = NFQUEUE;
 
@@ -184,7 +262,7 @@ netfilter_read_linux(pcap_t *handle, int max_packets, pcap_handler callback, u_c
 
                                gettimeofday(&pkth.ts, NULL);
                                if (handle->fcode.bf_insns == NULL ||
-                                               bpf_filter(handle->fcode.bf_insns, payload, pkth.len, pkth.caplen)) 
+                                               pcapint_filter(handle->fcode.bf_insns, payload, pkth.len, pkth.caplen))
                                {
                                        handlep->packets_read++;
                                        callback(user, &pkth, payload);
@@ -202,12 +280,30 @@ netfilter_read_linux(pcap_t *handle, int max_packets, pcap_handler callback, u_c
                }
 
                msg_len = NLMSG_ALIGN(nlh->nlmsg_len);
-               if (msg_len > len)
-                       msg_len = len;
-
-               len -= msg_len;
-               buf += msg_len;
+               /*
+                * If the message length would run past the end of the
+                * buffer, truncate it to the remaining space in the
+                * buffer.
+                *
+                * To squelch warnings, we cast ep - bp to uint32_t, which
+                * is unsigned and is the type of msg_len; ep >= bp, and
+                * len should fit in 32 bits (either it's set from an int
+                * or it's set from a recv() call with a buffer size that's
+                * an int, and we're assuming either ILP32 or LP64), so
+                * the cast is safe.
+                */
+               if (msg_len > (uint32_t)(ep - bp))
+                       msg_len = (uint32_t)(ep - bp);
+
+               bp += msg_len;
+               if (count >= max_packets && !PACKET_COUNT_IS_UNLIMITED(max_packets)) {
+                       handle->bp = bp;
+                       handle->cc = (u_int)(ep - bp);
+                       return count;
+               }
        }
+
+       handle->cc = 0;
        return count;
 }
 
@@ -224,37 +320,41 @@ netfilter_stats_linux(pcap_t *handle, struct pcap_stat *stats)
        struct pcap_netfilter *handlep = handle->priv;
 
        stats->ps_recv = handlep->packets_read;
-       stats->ps_drop = 0;
+       stats->ps_drop = handlep->packets_nobufs;
        stats->ps_ifdrop = 0;
        return 0;
 }
 
 static int
-netfilter_inject_linux(pcap_t *handle, const void *buf, size_t size)
+netfilter_inject_linux(pcap_t *handle, const void *buf _U_, int size _U_)
 {
-       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "inject not supported on netfilter devices");
+       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+           "Packet injection is not supported on netfilter devices");
        return (-1);
-}                           
+}
 
 struct my_nfattr {
-       u_int16_t nfa_len;
-       u_int16_t nfa_type;
+       uint16_t nfa_len;
+       uint16_t nfa_type;
        void *data;
 };
 
 static int
-netfilter_send_config_msg(const pcap_t *handle, u_int16_t msg_type, int ack, u_int8_t family, u_int16_t res_id, const struct my_nfattr *mynfa)
+netfilter_send_config_msg(const pcap_t *handle, uint16_t msg_type, int ack, u_int8_t family, u_int16_t res_id, const struct my_nfattr *mynfa)
 {
        char buf[1024] __attribute__ ((aligned));
+       memset(buf, 0, sizeof(buf));
 
        struct nlmsghdr *nlh = (struct nlmsghdr *) buf;
        struct nfgenmsg *nfg = (struct nfgenmsg *) (buf + sizeof(struct nlmsghdr));
 
        struct sockaddr_nl snl;
        static unsigned int seq_id;
-       
+
        if (!seq_id)
+DIAG_OFF_NARROWING
                seq_id = time(NULL);
+DIAG_ON_NARROWING
        ++seq_id;
 
        nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nfgenmsg));
@@ -292,7 +392,11 @@ netfilter_send_config_msg(const pcap_t *handle, u_int16_t msg_type, int ack, u_i
 
                /* ignore interrupt system call error */
                do {
-                       len = recvfrom(handle->fd, buf, sizeof(buf), 0, (struct sockaddr *) &snl, &addrlen);
+                       /*
+                        * The buffer is not so big that its size won't
+                        * fit into an int.
+                        */
+                       len = (int)recvfrom(handle->fd, buf, sizeof(buf), 0, (struct sockaddr *) &snl, &addrlen);
                } while ((len == -1) && (errno == EINTR));
 
                if (len <= 0)
@@ -307,7 +411,7 @@ netfilter_send_config_msg(const pcap_t *handle, u_int16_t msg_type, int ack, u_i
                if (snl.nl_pid != 0 || seq_id != nlh->nlmsg_seq)        /* if not from kernel or wrong sequence skip */
                        continue;
 
-               while (len >= NLMSG_SPACE(0) && NLMSG_OK(nlh, len)) {
+               while ((u_int)len >= NLMSG_SPACE(0) && NLMSG_OK(nlh, (u_int)len)) {
                        if (nlh->nlmsg_type == NLMSG_ERROR || (nlh->nlmsg_type == NLMSG_DONE && nlh->nlmsg_flags & NLM_F_MULTI)) {
                                if (nlh->nlmsg_len < NLMSG_ALIGN(sizeof(struct nlmsgerr))) {
                                        errno = EBADMSG;
@@ -324,13 +428,13 @@ netfilter_send_config_msg(const pcap_t *handle, u_int16_t msg_type, int ack, u_i
 }
 
 static int
-nflog_send_config_msg(const pcap_t *handle, u_int8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
+nflog_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
 {
        return netfilter_send_config_msg(handle, (NFNL_SUBSYS_ULOG << 8) | NFULNL_MSG_CONFIG, 1, family, group_id, mynfa);
 }
 
 static int
-nflog_send_config_cmd(const pcap_t *handle, u_int16_t group_id, u_int8_t cmd, u_int8_t family)
+nflog_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int8_t family)
 {
        struct nfulnl_msg_config_cmd msg;
        struct my_nfattr nfa;
@@ -344,8 +448,8 @@ nflog_send_config_cmd(const pcap_t *handle, u_int16_t group_id, u_int8_t cmd, u_
        return nflog_send_config_msg(handle, family, group_id, &nfa);
 }
 
-static int 
-nflog_send_config_mode(const pcap_t *handle, u_int16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
+static int
+nflog_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
 {
        struct nfulnl_msg_config_mode msg;
        struct my_nfattr nfa;
@@ -361,7 +465,7 @@ nflog_send_config_mode(const pcap_t *handle, u_int16_t group_id, u_int8_t copy_m
 }
 
 static int
-nfqueue_send_verdict(const pcap_t *handle, u_int16_t group_id, u_int32_t id, u_int32_t verdict)
+nfqueue_send_verdict(const pcap_t *handle, uint16_t group_id, u_int32_t id, u_int32_t verdict)
 {
        struct nfqnl_msg_verdict_hdr msg;
        struct my_nfattr nfa;
@@ -377,13 +481,13 @@ nfqueue_send_verdict(const pcap_t *handle, u_int16_t group_id, u_int32_t id, u_i
 }
 
 static int
-nfqueue_send_config_msg(const pcap_t *handle, u_int8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
+nfqueue_send_config_msg(const pcap_t *handle, uint8_t family, u_int16_t group_id, const struct my_nfattr *mynfa)
 {
        return netfilter_send_config_msg(handle, (NFNL_SUBSYS_QUEUE << 8) | NFQNL_MSG_CONFIG, 1, family, group_id, mynfa);
 }
 
 static int
-nfqueue_send_config_cmd(const pcap_t *handle, u_int16_t group_id, u_int8_t cmd, u_int16_t pf)
+nfqueue_send_config_cmd(const pcap_t *handle, uint16_t group_id, u_int8_t cmd, u_int16_t pf)
 {
        struct nfqnl_msg_config_cmd msg;
        struct my_nfattr nfa;
@@ -398,8 +502,8 @@ nfqueue_send_config_cmd(const pcap_t *handle, u_int16_t group_id, u_int8_t cmd,
        return nfqueue_send_config_msg(handle, AF_UNSPEC, group_id, &nfa);
 }
 
-static int 
-nfqueue_send_config_mode(const pcap_t *handle, u_int16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
+static int
+nfqueue_send_config_mode(const pcap_t *handle, uint16_t group_id, u_int8_t copy_mode, u_int32_t copy_range)
 {
        struct nfqnl_msg_config_params msg;
        struct my_nfattr nfa;
@@ -417,21 +521,21 @@ nfqueue_send_config_mode(const pcap_t *handle, u_int16_t group_id, u_int8_t copy
 static int
 netfilter_activate(pcap_t* handle)
 {
-       const char *dev = handle->opt.source;
+       const char *dev = handle->opt.device;
        unsigned short groups[32];
        int group_count = 0;
        nftype_t type = OTHER;
        int i;
 
-       if (strncmp(dev, NFLOG_IFACE, strlen(NFLOG_IFACE)) == 0) {
-               dev += strlen(NFLOG_IFACE);
+       if (strncmp(dev, NFLOG_IFACE, strlen(NFLOG_IFACE)) == 0) {
+               dev += strlen(NFLOG_IFACE);
                type = NFLOG;
 
        } else if (strncmp(dev, NFQUEUE_IFACE, strlen(NFQUEUE_IFACE)) == 0) {
                dev += strlen(NFQUEUE_IFACE);
                type = NFQUEUE;
        }
+
        if (type != OTHER && *dev == ':') {
                dev++;
                while (*dev) {
@@ -440,8 +544,8 @@ netfilter_activate(pcap_t* handle)
 
                        if (group_count == 32) {
                                snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
-                                               "Maximum 32 netfilter groups! dev: %s", 
-                                               handle->opt.source);
+                                               "Maximum 32 netfilter groups! dev: %s",
+                                               handle->opt.device);
                                return PCAP_ERROR;
                        }
 
@@ -465,8 +569,8 @@ netfilter_activate(pcap_t* handle)
 
        if (type == OTHER || *dev) {
                snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
-                               "Can't get netfilter group(s) index from %s", 
-                               handle->opt.source);
+                               "Can't get netfilter group(s) index from %s",
+                               handle->opt.device);
                return PCAP_ERROR;
        }
 
@@ -476,87 +580,116 @@ netfilter_activate(pcap_t* handle)
                group_count = 1;
        }
 
+       /*
+        * Turn a negative snapshot value (invalid), a snapshot value of
+        * 0 (unspecified), or a value bigger than the normal maximum
+        * value, into the maximum allowed value.
+        *
+        * If some application really *needs* a bigger snapshot
+        * length, we should just increase MAXIMUM_SNAPLEN.
+        */
+       if (handle->snapshot <= 0 || handle->snapshot > MAXIMUM_SNAPLEN)
+               handle->snapshot = MAXIMUM_SNAPLEN;
+
        /* Initialize some components of the pcap structure. */
        handle->bufsize = 128 + handle->snapshot;
        handle->offset = 0;
        handle->read_op = netfilter_read_linux;
        handle->inject_op = netfilter_inject_linux;
-       handle->setfilter_op = install_bpf_program; /* no kernel filtering */
+       handle->setfilter_op = pcapint_install_bpf_program; /* no kernel filtering */
        handle->setdirection_op = NULL;
        handle->set_datalink_op = netfilter_set_datalink;
-       handle->getnonblock_op = pcap_getnonblock_fd;
-       handle->setnonblock_op = pcap_setnonblock_fd;
+       handle->getnonblock_op = pcapint_getnonblock_fd;
+       handle->setnonblock_op = pcapint_setnonblock_fd;
        handle->stats_op = netfilter_stats_linux;
 
        /* Create netlink socket */
        handle->fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
        if (handle->fd < 0) {
-               snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Can't create raw socket %d:%s", errno, pcap_strerror(errno));
+               pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
+                   errno, "Can't create raw socket");
                return PCAP_ERROR;
        }
 
        if (type == NFLOG) {
                handle->linktype = DLT_NFLOG;
                handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
-               if (handle->dlt_list != NULL) {
-                       handle->dlt_list[0] = DLT_NFLOG;
-                       handle->dlt_list[1] = DLT_IPV4;
-                       handle->dlt_count = 2;
+               if (handle->dlt_list == NULL) {
+                       pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                           PCAP_ERRBUF_SIZE, errno,
+                           "Can't allocate DLT list");
+                       goto close_fail;
                }
-
+               handle->dlt_list[0] = DLT_NFLOG;
+               handle->dlt_list[1] = DLT_IPV4;
+               handle->dlt_count = 2;
        } else
                handle->linktype = DLT_IPV4;
 
        handle->buffer = malloc(handle->bufsize);
        if (!handle->buffer) {
-               snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Can't allocate dump buffer: %s", pcap_strerror(errno));
+               pcapint_fmt_errmsg_for_errno(handle->errbuf, PCAP_ERRBUF_SIZE,
+                   errno, "Can't allocate dump buffer");
                goto close_fail;
        }
 
        if (type == NFLOG) {
                if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
-                       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "NFULNL_CFG_CMD_PF_UNBIND: %s", pcap_strerror(errno));
+                       pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                           PCAP_ERRBUF_SIZE, errno,
+                           "NFULNL_CFG_CMD_PF_UNBIND");
                        goto close_fail;
                }
 
                if (nflog_send_config_cmd(handle, 0, NFULNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
-                       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "NFULNL_CFG_CMD_PF_BIND: %s", pcap_strerror(errno));
+                       pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                           PCAP_ERRBUF_SIZE, errno, "NFULNL_CFG_CMD_PF_BIND");
                        goto close_fail;
                }
 
                /* Bind socket to the nflog groups */
                for (i = 0; i < group_count; i++) {
                        if (nflog_send_config_cmd(handle, groups[i], NFULNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
-                               snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Can't listen on group group index: %s", pcap_strerror(errno));
+                               pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                                   PCAP_ERRBUF_SIZE, errno,
+                                   "Can't listen on group index");
                                goto close_fail;
                        }
 
                        if (nflog_send_config_mode(handle, groups[i], NFULNL_COPY_PACKET, handle->snapshot) < 0) {
-                               snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "NFULNL_COPY_PACKET: %s", pcap_strerror(errno));
+                               pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                                   PCAP_ERRBUF_SIZE, errno,
+                                   "NFULNL_COPY_PACKET");
                                goto close_fail;
                        }
                }
 
        } else {
                if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_UNBIND, AF_INET) < 0) {
-                       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "NFQNL_CFG_CMD_PF_UNBIND: %s", pcap_strerror(errno));
+                       pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                           PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_UNBIND");
                        goto close_fail;
                }
 
                if (nfqueue_send_config_cmd(handle, 0, NFQNL_CFG_CMD_PF_BIND, AF_INET) < 0) {
-                       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "NFQNL_CFG_CMD_PF_BIND: %s", pcap_strerror(errno));
+                       pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                           PCAP_ERRBUF_SIZE, errno, "NFQNL_CFG_CMD_PF_BIND");
                        goto close_fail;
                }
 
                /* Bind socket to the nfqueue groups */
                for (i = 0; i < group_count; i++) {
                        if (nfqueue_send_config_cmd(handle, groups[i], NFQNL_CFG_CMD_BIND, AF_UNSPEC) < 0) {
-                               snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "Can't listen on group group index: %s", pcap_strerror(errno));
+                               pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                                   PCAP_ERRBUF_SIZE, errno,
+                                   "Can't listen on group index");
                                goto close_fail;
                        }
 
                        if (nfqueue_send_config_mode(handle, groups[i], NFQNL_COPY_PACKET, handle->snapshot) < 0) {
-                               snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "NFQNL_COPY_PACKET: %s", pcap_strerror(errno));
+                               pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                                   PCAP_ERRBUF_SIZE, errno,
+                                   "NFQNL_COPY_PACKET");
                                goto close_fail;
                        }
                }
@@ -566,7 +699,7 @@ netfilter_activate(pcap_t* handle)
                /*
                 * Monitor mode doesn't apply to netfilter devices.
                 */
-               pcap_cleanup_live_common(handle);
+               pcapint_cleanup_live_common(handle);
                return PCAP_ERROR_RFMON_NOTSUP;
        }
 
@@ -575,7 +708,8 @@ netfilter_activate(pcap_t* handle)
                 * Set the socket buffer size to the specified value.
                 */
                if (setsockopt(handle->fd, SOL_SOCKET, SO_RCVBUF, &handle->opt.buffer_size, sizeof(handle->opt.buffer_size)) == -1) {
-                       snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "SO_RCVBUF: %s", pcap_strerror(errno));
+                       pcapint_fmt_errmsg_for_errno(handle->errbuf,
+                           PCAP_ERRBUF_SIZE, errno, "SO_RCVBUF");
                        goto close_fail;
                }
        }
@@ -584,7 +718,7 @@ netfilter_activate(pcap_t* handle)
        return 0;
 
 close_fail:
-       pcap_cleanup_live_common(handle);
+       pcapint_cleanup_live_common(handle);
        return PCAP_ERROR;
 }
 
@@ -595,14 +729,12 @@ netfilter_create(const char *device, char *ebuf, int *is_ours)
        pcap_t *p;
 
        /* Does this look like an netfilter device? */
-       cp = strrchr(device, '/');
-       if (cp == NULL)
-               cp = device;
+       cp = device;
 
        /* Does it begin with NFLOG_IFACE or NFQUEUE_IFACE? */
        if (strncmp(cp, NFLOG_IFACE, sizeof NFLOG_IFACE - 1) == 0)
                cp += sizeof NFLOG_IFACE - 1;
-       else if (strncmp(cp, NFQUEUE_IFACE, sizeof NFQUEUE_IFACE - 1) == 0) 
+       else if (strncmp(cp, NFQUEUE_IFACE, sizeof NFQUEUE_IFACE - 1) == 0)
                cp += sizeof NFQUEUE_IFACE - 1;
        else {
                /* Nope, doesn't begin with NFLOG_IFACE nor NFQUEUE_IFACE */
@@ -623,7 +755,7 @@ netfilter_create(const char *device, char *ebuf, int *is_ours)
        /* OK, it's probably ours. */
        *is_ours = 1;
 
-       p = pcap_create_common(device, ebuf, sizeof (struct pcap_netfilter));
+       p = PCAP_CREATE_COMMON(ebuf, struct pcap_netfilter);
        if (p == NULL)
                return (NULL);
 
@@ -631,25 +763,33 @@ netfilter_create(const char *device, char *ebuf, int *is_ours)
        return (p);
 }
 
-int 
-netfilter_findalldevs(pcap_if_t **alldevsp, char *err_str)
+int
+netfilter_findalldevs(pcap_if_list_t *devlistp, char *err_str)
 {
        int sock;
-       
+
        sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_NETFILTER);
        if (sock < 0) {
                /* if netlink is not supported this is not fatal */
                if (errno == EAFNOSUPPORT || errno == EPROTONOSUPPORT)
                        return 0;
-               snprintf(err_str, PCAP_ERRBUF_SIZE, "Can't open netlink socket %d:%s",
-                       errno, pcap_strerror(errno));
+               pcapint_fmt_errmsg_for_errno(err_str, PCAP_ERRBUF_SIZE,
+                   errno, "Can't open netlink socket");
                return -1;
        }
        close(sock);
 
-       if (pcap_add_if(alldevsp, NFLOG_IFACE, 0, "Linux netfilter log (NFLOG) interface", err_str) < 0)
+       /*
+        * The notion of "connected" vs. "disconnected" doesn't apply.
+        * XXX - what about "up" and "running"?
+        */
+       if (pcapint_add_dev(devlistp, NFLOG_IFACE,
+           PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
+           "Linux netfilter log (NFLOG) interface", err_str) == NULL)
                return -1;
-       if (pcap_add_if(alldevsp, NFQUEUE_IFACE, 0, "Linux netfilter queue (NFQUEUE) interface", err_str) < 0)
+       if (pcapint_add_dev(devlistp, NFQUEUE_IFACE,
+           PCAP_IF_CONNECTION_STATUS_NOT_APPLICABLE,
+           "Linux netfilter queue (NFQUEUE) interface", err_str) == NULL)
                return -1;
        return 0;
 }