+/*
+ * Get from "/sys/class/net" all interfaces listed there; if they're
+ * already in the list of interfaces we have, that won't add another
+ * instance, but if they're not, that'll add them.
+ *
+ * We don't bother getting any addresses for them; it appears you can't
+ * use SIOCGIFADDR on Linux to get IPv6 addresses for interfaces, and,
+ * although some other types of addresses can be fetched with SIOCGIFADDR,
+ * we don't bother with them for now.
+ *
+ * We also don't fail if we couldn't open "/sys/class/net"; we just leave
+ * the list of interfaces as is, and return 0, so that we can try
+ * scanning /proc/net/dev.
+ */
+static int
+scan_sys_class_net(pcap_if_t **devlistp, char *errbuf)
+{
+ DIR *sys_class_net_d;
+ int fd;
+ struct dirent *ent;
+ char subsystem_path[PATH_MAX+1];
+ struct stat statb;
+ char *p;
+ char name[512]; /* XXX - pick a size */
+ char *q, *saveq;
+ struct ifreq ifrflags;
+ int ret = 1;
+
+ sys_class_net_d = opendir("/sys/class/net");
+ if (sys_class_net_d == NULL) {
+ /*
+ * Don't fail if it doesn't exist at all.
+ */
+ if (errno == ENOENT)
+ return (0);
+
+ /*
+ * Fail if we got some other error.
+ */
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "Can't open /sys/class/net: %s", pcap_strerror(errno));
+ return (-1);
+ }
+
+ /*
+ * Create a socket from which to fetch interface information.
+ */
+ fd = socket(AF_INET, SOCK_DGRAM, 0);
+ if (fd < 0) {
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "socket: %s", pcap_strerror(errno));
+ (void)closedir(sys_class_net_d);
+ return (-1);
+ }
+
+ for (;;) {
+ errno = 0;
+ ent = readdir(sys_class_net_d);
+ if (ent == NULL) {
+ /*
+ * Error or EOF; if errno != 0, it's an error.
+ */
+ break;
+ }
+
+ /*
+ * Ignore "." and "..".
+ */
+ if (strcmp(ent->d_name, ".") == 0 ||
+ strcmp(ent->d_name, "..") == 0)
+ continue;
+
+ /*
+ * Ignore plain files; they do not have subdirectories
+ * and thus have no attributes.
+ */
+ if (ent->d_type == DT_REG)
+ continue;
+
+ /*
+ * Is there an "ifindex" file under that name?
+ * (We don't care whether it's a directory or
+ * a symlink; older kernels have directories
+ * for devices, newer kernels have symlinks to
+ * directories.)
+ */
+ snprintf(subsystem_path, sizeof subsystem_path,
+ "/sys/class/net/%s/ifindex", ent->d_name);
+ if (lstat(subsystem_path, &statb) != 0) {
+ /*
+ * Stat failed. Either there was an error
+ * other than ENOENT, and we don't know if
+ * this is an interface, or it's ENOENT,
+ * and either some part of "/sys/class/net/{if}"
+ * disappeared, in which case it probably means
+ * the interface disappeared, or there's no
+ * "ifindex" file, which means it's not a
+ * network interface.
+ */
+ continue;
+ }
+
+ /*
+ * Get the interface name.
+ */
+ p = &ent->d_name[0];
+ q = &name[0];
+ while (*p != '\0' && isascii(*p) && !isspace(*p)) {
+ if (*p == ':') {
+ /*
+ * This could be the separator between a
+ * name and an alias number, or it could be
+ * the separator between a name with no
+ * alias number and the next field.
+ *
+ * If there's a colon after digits, it
+ * separates the name and the alias number,
+ * otherwise it separates the name and the
+ * next field.
+ */
+ saveq = q;
+ while (isascii(*p) && isdigit(*p))
+ *q++ = *p++;
+ if (*p != ':') {
+ /*
+ * That was the next field,
+ * not the alias number.
+ */
+ q = saveq;
+ }
+ break;
+ } else
+ *q++ = *p++;
+ }
+ *q = '\0';
+
+ /*
+ * Get the flags for this interface, and skip it if
+ * it's not up.
+ */
+ strncpy(ifrflags.ifr_name, name, sizeof(ifrflags.ifr_name));
+ if (ioctl(fd, SIOCGIFFLAGS, (char *)&ifrflags) < 0) {
+ if (errno == ENXIO || errno == ENODEV)
+ continue;
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "SIOCGIFFLAGS: %.*s: %s",
+ (int)sizeof(ifrflags.ifr_name),
+ ifrflags.ifr_name,
+ pcap_strerror(errno));
+ ret = -1;
+ break;
+ }
+ if (!(ifrflags.ifr_flags & IFF_UP))
+ continue;
+
+ /*
+ * Add an entry for this interface, with no addresses.
+ */
+ if (pcap_add_if(devlistp, name, ifrflags.ifr_flags, NULL,
+ errbuf) == -1) {
+ /*
+ * Failure.
+ */
+ ret = -1;
+ break;
+ }
+ }
+ if (ret != -1) {
+ /*
+ * Well, we didn't fail for any other reason; did we
+ * fail due to an error reading the directory?
+ */
+ if (errno != 0) {
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "Error reading /sys/class/net: %s",
+ pcap_strerror(errno));
+ ret = -1;
+ }
+ }
+
+ (void)close(fd);
+ (void)closedir(sys_class_net_d);
+ return (ret);
+}
+
+/*
+ * Get from "/proc/net/dev" all interfaces listed there; if they're
+ * already in the list of interfaces we have, that won't add another
+ * instance, but if they're not, that'll add them.
+ *
+ * See comments from scan_sys_class_net().
+ */
+static int
+scan_proc_net_dev(pcap_if_t **devlistp, char *errbuf)
+{
+ FILE *proc_net_f;
+ int fd;
+ char linebuf[512];
+ int linenum;
+ char *p;
+ char name[512]; /* XXX - pick a size */
+ char *q, *saveq;
+ struct ifreq ifrflags;
+ int ret = 0;
+
+ proc_net_f = fopen("/proc/net/dev", "r");
+ if (proc_net_f == NULL) {
+ /*
+ * Don't fail if it doesn't exist at all.
+ */
+ if (errno == ENOENT)
+ return (0);
+
+ /*
+ * Fail if we got some other error.
+ */
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "Can't open /proc/net/dev: %s", pcap_strerror(errno));
+ return (-1);
+ }
+
+ /*
+ * Create a socket from which to fetch interface information.
+ */
+ fd = socket(AF_INET, SOCK_DGRAM, 0);
+ if (fd < 0) {
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "socket: %s", pcap_strerror(errno));
+ (void)fclose(proc_net_f);
+ return (-1);
+ }
+
+ for (linenum = 1;
+ fgets(linebuf, sizeof linebuf, proc_net_f) != NULL; linenum++) {
+ /*
+ * Skip the first two lines - they're headers.
+ */
+ if (linenum <= 2)
+ continue;
+
+ p = &linebuf[0];
+
+ /*
+ * Skip leading white space.
+ */
+ while (*p != '\0' && isascii(*p) && isspace(*p))
+ p++;
+ if (*p == '\0' || *p == '\n')
+ continue; /* blank line */
+
+ /*
+ * Get the interface name.
+ */
+ q = &name[0];
+ while (*p != '\0' && isascii(*p) && !isspace(*p)) {
+ if (*p == ':') {
+ /*
+ * This could be the separator between a
+ * name and an alias number, or it could be
+ * the separator between a name with no
+ * alias number and the next field.
+ *
+ * If there's a colon after digits, it
+ * separates the name and the alias number,
+ * otherwise it separates the name and the
+ * next field.
+ */
+ saveq = q;
+ while (isascii(*p) && isdigit(*p))
+ *q++ = *p++;
+ if (*p != ':') {
+ /*
+ * That was the next field,
+ * not the alias number.
+ */
+ q = saveq;
+ }
+ break;
+ } else
+ *q++ = *p++;
+ }
+ *q = '\0';
+
+ /*
+ * Get the flags for this interface, and skip it if
+ * it's not up.
+ */
+ strncpy(ifrflags.ifr_name, name, sizeof(ifrflags.ifr_name));
+ if (ioctl(fd, SIOCGIFFLAGS, (char *)&ifrflags) < 0) {
+ if (errno == ENXIO)
+ continue;
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "SIOCGIFFLAGS: %.*s: %s",
+ (int)sizeof(ifrflags.ifr_name),
+ ifrflags.ifr_name,
+ pcap_strerror(errno));
+ ret = -1;
+ break;
+ }
+ if (!(ifrflags.ifr_flags & IFF_UP))
+ continue;
+
+ /*
+ * Add an entry for this interface, with no addresses.
+ */
+ if (pcap_add_if(devlistp, name, ifrflags.ifr_flags, NULL,
+ errbuf) == -1) {
+ /*
+ * Failure.
+ */
+ ret = -1;
+ break;
+ }
+ }
+ if (ret != -1) {
+ /*
+ * Well, we didn't fail for any other reason; did we
+ * fail due to an error reading the file?
+ */
+ if (ferror(proc_net_f)) {
+ (void)snprintf(errbuf, PCAP_ERRBUF_SIZE,
+ "Error reading /proc/net/dev: %s",
+ pcap_strerror(errno));
+ ret = -1;
+ }
+ }
+
+ (void)close(fd);
+ (void)fclose(proc_net_f);
+ return (ret);
+}
+
+/*
+ * Description string for the "any" device.
+ */
+static const char any_descr[] = "Pseudo-device that captures on all interfaces";
+
+int
+pcap_platform_finddevs(pcap_if_t **alldevsp, char *errbuf)
+{
+ int ret;
+
+ /*
+ * Read "/sys/class/net", and add to the list of interfaces all
+ * interfaces listed there that we don't already have, because,
+ * on Linux, SIOCGIFCONF reports only interfaces with IPv4 addresses,
+ * and even getifaddrs() won't return information about
+ * interfaces with no addresses, so you need to read "/sys/class/net"
+ * to get the names of the rest of the interfaces.
+ */
+ ret = scan_sys_class_net(alldevsp, errbuf);
+ if (ret == -1)
+ return (-1); /* failed */
+ if (ret == 0) {
+ /*
+ * No /sys/class/net; try reading /proc/net/dev instead.
+ */
+ if (scan_proc_net_dev(alldevsp, errbuf) == -1)
+ return (-1);
+ }
+
+ /*
+ * Add the "any" device.
+ */
+ if (pcap_add_if(alldevsp, "any", 0, any_descr, errbuf) < 0)
+ return (-1);
+
+#ifdef HAVE_DAG_API
+ /*
+ * Add DAG devices.
+ */
+ if (dag_platform_finddevs(alldevsp, errbuf) < 0)
+ return (-1);
+#endif /* HAVE_DAG_API */
+
+#ifdef HAVE_SEPTEL_API
+ /*
+ * Add Septel devices.
+ */
+ if (septel_platform_finddevs(alldevsp, errbuf) < 0)
+ return (-1);
+#endif /* HAVE_SEPTEL_API */
+
+#ifdef HAVE_SNF_API
+ if (snf_platform_finddevs(alldevsp, errbuf) < 0)
+ return (-1);
+#endif /* HAVE_SNF_API */
+
+#ifdef PCAP_SUPPORT_BT
+ /*
+ * Add Bluetooth devices.
+ */
+ if (bt_platform_finddevs(alldevsp, errbuf) < 0)
+ return (-1);
+#endif
+
+#ifdef PCAP_SUPPORT_USB
+ /*
+ * Add USB devices.
+ */
+ if (usb_platform_finddevs(alldevsp, errbuf) < 0)
+ return (-1);
+#endif
+
+#ifdef PCAP_SUPPORT_NETFILTER
+ /*
+ * Add netfilter devices.
+ */
+ if (netfilter_platform_finddevs(alldevsp, errbuf) < 0)
+ return (-1);
+#endif
+
+#if PCAP_SUPPORT_CANUSB
+ if (canusb_platform_finddevs(alldevsp, errbuf) < 0)
+ return (-1);
+#endif
+
+ return (0);
+}
+
+/*
+ * Attach the given BPF code to the packet capture device.
+ */
+static int
+pcap_setfilter_linux_common(pcap_t *handle, struct bpf_program *filter,
+ int is_mmapped)
+{
+#ifdef SO_ATTACH_FILTER
+ struct sock_fprog fcode;
+ int can_filter_in_kernel;
+ int err = 0;
+#endif
+
+ if (!handle)
+ return -1;
+ if (!filter) {
+ strncpy(handle->errbuf, "setfilter: No filter specified",
+ PCAP_ERRBUF_SIZE);
+ return -1;
+ }
+
+ /* Make our private copy of the filter */
+
+ if (install_bpf_program(handle, filter) < 0)
+ /* install_bpf_program() filled in errbuf */
+ return -1;
+
+ /*
+ * Run user level packet filter by default. Will be overriden if
+ * installing a kernel filter succeeds.
+ */
+ handle->md.use_bpf = 0;
+
+ /* Install kernel level filter if possible */
+
+#ifdef SO_ATTACH_FILTER
+#ifdef USHRT_MAX
+ if (handle->fcode.bf_len > USHRT_MAX) {
+ /*
+ * fcode.len is an unsigned short for current kernel.
+ * I have yet to see BPF-Code with that much
+ * instructions but still it is possible. So for the
+ * sake of correctness I added this check.
+ */
+ fprintf(stderr, "Warning: Filter too complex for kernel\n");
+ fcode.len = 0;
+ fcode.filter = NULL;
+ can_filter_in_kernel = 0;
+ } else
+#endif /* USHRT_MAX */
+ {
+ /*
+ * Oh joy, the Linux kernel uses struct sock_fprog instead
+ * of struct bpf_program and of course the length field is
+ * of different size. Pointed out by Sebastian
+ *
+ * Oh, and we also need to fix it up so that all "ret"
+ * instructions with non-zero operands have 65535 as the
+ * operand if we're not capturing in memory-mapped modee,
+ * and so that, if we're in cooked mode, all memory-reference
+ * instructions use special magic offsets in references to
+ * the link-layer header and assume that the link-layer
+ * payload begins at 0; "fix_program()" will do that.
+ */
+ switch (fix_program(handle, &fcode, is_mmapped)) {
+
+ case -1:
+ default:
+ /*
+ * Fatal error; just quit.
+ * (The "default" case shouldn't happen; we
+ * return -1 for that reason.)
+ */
+ return -1;
+
+ case 0:
+ /*
+ * The program performed checks that we can't make
+ * work in the kernel.
+ */
+ can_filter_in_kernel = 0;
+ break;
+
+ case 1:
+ /*
+ * We have a filter that'll work in the kernel.
+ */
+ can_filter_in_kernel = 1;
+ break;
+ }
+ }
+
+ /*
+ * NOTE: at this point, we've set both the "len" and "filter"
+ * fields of "fcode". As of the 2.6.32.4 kernel, at least,
+ * those are the only members of the "sock_fprog" structure,
+ * so we initialize every member of that structure.
+ *
+ * If there is anything in "fcode" that is not initialized,
+ * it is either a field added in a later kernel, or it's
+ * padding.
+ *
+ * If a new field is added, this code needs to be updated
+ * to set it correctly.
+ *
+ * If there are no other fields, then:
+ *
+ * if the Linux kernel looks at the padding, it's
+ * buggy;
+ *
+ * if the Linux kernel doesn't look at the padding,
+ * then if some tool complains that we're passing
+ * uninitialized data to the kernel, then the tool
+ * is buggy and needs to understand that it's just
+ * padding.
+ */
+ if (can_filter_in_kernel) {
+ if ((err = set_kernel_filter(handle, &fcode)) == 0)
+ {
+ /* Installation succeded - using kernel filter. */
+ handle->md.use_bpf = 1;
+ }
+ else if (err == -1) /* Non-fatal error */
+ {
+ /*
+ * Print a warning if we weren't able to install
+ * the filter for a reason other than "this kernel
+ * isn't configured to support socket filters.
+ */
+ if (errno != ENOPROTOOPT && errno != EOPNOTSUPP) {
+ fprintf(stderr,
+ "Warning: Kernel filter failed: %s\n",
+ pcap_strerror(errno));
+ }
+ }
+ }
+
+ /*
+ * If we're not using the kernel filter, get rid of any kernel
+ * filter that might've been there before, e.g. because the
+ * previous filter could work in the kernel, or because some other
+ * code attached a filter to the socket by some means other than
+ * calling "pcap_setfilter()". Otherwise, the kernel filter may
+ * filter out packets that would pass the new userland filter.
+ */
+ if (!handle->md.use_bpf)
+ reset_kernel_filter(handle);
+
+ /*
+ * Free up the copy of the filter that was made by "fix_program()".
+ */
+ if (fcode.filter != NULL)
+ free(fcode.filter);
+
+ if (err == -2)
+ /* Fatal error */
+ return -1;
+#endif /* SO_ATTACH_FILTER */
+
+ return 0;
+}
+
+static int
+pcap_setfilter_linux(pcap_t *handle, struct bpf_program *filter)
+{
+ return pcap_setfilter_linux_common(handle, filter, 0);
+}
+
+
+/*
+ * Set direction flag: Which packets do we accept on a forwarding
+ * single device? IN, OUT or both?
+ */
+static int
+pcap_setdirection_linux(pcap_t *handle, pcap_direction_t d)
+{
+#ifdef HAVE_PF_PACKET_SOCKETS
+ if (!handle->md.sock_packet) {
+ handle->direction = d;
+ return 0;
+ }
+#endif
+ /*
+ * We're not using PF_PACKET sockets, so we can't determine
+ * the direction of the packet.
+ */
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "Setting direction is not supported on SOCK_PACKET sockets");
+ return -1;
+}
+
+#ifdef HAVE_PF_PACKET_SOCKETS
+/*
+ * Map the PACKET_ value to a LINUX_SLL_ value; we
+ * want the same numerical value to be used in
+ * the link-layer header even if the numerical values
+ * for the PACKET_ #defines change, so that programs
+ * that look at the packet type field will always be
+ * able to handle DLT_LINUX_SLL captures.
+ */
+static short int
+map_packet_type_to_sll_type(short int sll_pkttype)
+{
+ switch (sll_pkttype) {
+
+ case PACKET_HOST:
+ return htons(LINUX_SLL_HOST);
+
+ case PACKET_BROADCAST:
+ return htons(LINUX_SLL_BROADCAST);
+
+ case PACKET_MULTICAST:
+ return htons(LINUX_SLL_MULTICAST);
+
+ case PACKET_OTHERHOST:
+ return htons(LINUX_SLL_OTHERHOST);
+
+ case PACKET_OUTGOING:
+ return htons(LINUX_SLL_OUTGOING);
+
+ default:
+ return -1;
+ }
+}
+#endif
+
+static int
+is_wifi(int sock_fd
+#ifndef IW_MODE_MONITOR
+_U_
+#endif
+, const char *device)
+{
+ char *pathstr;
+ struct stat statb;
+#ifdef IW_MODE_MONITOR
+ char errbuf[PCAP_ERRBUF_SIZE];
+#endif
+
+ /*
+ * See if there's a sysfs wireless directory for it.
+ * If so, it's a wireless interface.
+ */
+ if (asprintf(&pathstr, "/sys/class/net/%s/wireless", device) == -1) {
+ /*
+ * Just give up here.
+ */
+ return 0;
+ }
+ if (stat(pathstr, &statb) == 0) {
+ free(pathstr);
+ return 1;
+ }
+ free(pathstr);
+
+#ifdef IW_MODE_MONITOR
+ /*
+ * OK, maybe it's not wireless, or maybe this kernel doesn't
+ * support sysfs. Try the wireless extensions.
+ */
+ if (has_wext(sock_fd, device, errbuf) == 1) {
+ /*
+ * It supports the wireless extensions, so it's a Wi-Fi
+ * device.
+ */
+ return 1;
+ }
+#endif
+ return 0;
+}
+
+/*
+ * Linux uses the ARP hardware type to identify the type of an
+ * interface. pcap uses the DLT_xxx constants for this. This
+ * function takes a pointer to a "pcap_t", and an ARPHRD_xxx
+ * constant, as arguments, and sets "handle->linktype" to the
+ * appropriate DLT_XXX constant and sets "handle->offset" to
+ * the appropriate value (to make "handle->offset" plus link-layer
+ * header length be a multiple of 4, so that the link-layer payload
+ * will be aligned on a 4-byte boundary when capturing packets).
+ * (If the offset isn't set here, it'll be 0; add code as appropriate
+ * for cases where it shouldn't be 0.)
+ *
+ * If "cooked_ok" is non-zero, we can use DLT_LINUX_SLL and capture
+ * in cooked mode; otherwise, we can't use cooked mode, so we have
+ * to pick some type that works in raw mode, or fail.
+ *
+ * Sets the link type to -1 if unable to map the type.
+ */
+static void map_arphrd_to_dlt(pcap_t *handle, int sock_fd, int arptype,
+ const char *device, int cooked_ok)
+{
+ static const char cdma_rmnet[] = "cdma_rmnet";
+
+ switch (arptype) {
+
+ case ARPHRD_ETHER:
+ /*
+ * For various annoying reasons having to do with DHCP
+ * software, some versions of Android give the mobile-
+ * phone-network interface an ARPHRD_ value of
+ * ARPHRD_ETHER, even though the packets supplied by
+ * that interface have no link-layer header, and begin
+ * with an IP header, so that the ARPHRD_ value should
+ * be ARPHRD_NONE.
+ *
+ * Detect those devices by checking the device name, and
+ * use DLT_RAW for them.
+ */
+ if (strncmp(device, cdma_rmnet, sizeof cdma_rmnet - 1) == 0) {
+ handle->linktype = DLT_RAW;
+ return;
+ }
+
+ /*
+ * Is this a real Ethernet device? If so, give it a
+ * link-layer-type list with DLT_EN10MB and DLT_DOCSIS, 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).
+ *
+ * XXX - are there any other sorts of "fake Ethernet" that
+ * have ARPHRD_ETHER but that shouldn't offer DLT_DOCSIS as
+ * a Cisco CMTS won't put traffic onto it or get traffic
+ * bridged onto it? ISDN is handled in "activate_new()",
+ * as we fall back on cooked mode there, and we use
+ * is_wifi() to check for 802.11 devices; are there any
+ * others?
+ */
+ if (!is_wifi(sock_fd, device)) {
+ /*
+ * It's not a Wi-Fi device; offer DOCSIS.
+ */
+ handle->dlt_list = (u_int *) malloc(sizeof(u_int) * 2);
+ /*
+ * If that fails, just leave the list empty.
+ */
+ if (handle->dlt_list != NULL) {
+ handle->dlt_list[0] = DLT_EN10MB;
+ handle->dlt_list[1] = DLT_DOCSIS;
+ handle->dlt_count = 2;
+ }
+ }
+ /* FALLTHROUGH */
+
+ case ARPHRD_METRICOM:
+ case ARPHRD_LOOPBACK:
+ handle->linktype = DLT_EN10MB;
+ handle->offset = 2;
+ break;
+
+ case ARPHRD_EETHER:
+ handle->linktype = DLT_EN3MB;
+ break;
+
+ case ARPHRD_AX25:
+ handle->linktype = DLT_AX25_KISS;
+ break;
+
+ case ARPHRD_PRONET:
+ handle->linktype = DLT_PRONET;
+ break;
+
+ case ARPHRD_CHAOS:
+ handle->linktype = DLT_CHAOS;
+ break;
+#ifndef ARPHRD_CAN
+#define ARPHRD_CAN 280
+#endif
+ case ARPHRD_CAN:
+ handle->linktype = DLT_CAN_SOCKETCAN;
+ break;
+
+#ifndef ARPHRD_IEEE802_TR
+#define ARPHRD_IEEE802_TR 800 /* From Linux 2.4 */
+#endif
+ case ARPHRD_IEEE802_TR:
+ case ARPHRD_IEEE802:
+ handle->linktype = DLT_IEEE802;
+ handle->offset = 2;
+ break;
+
+ case ARPHRD_ARCNET:
+ handle->linktype = DLT_ARCNET_LINUX;
+ break;
+
+#ifndef ARPHRD_FDDI /* From Linux 2.2.13 */
+#define ARPHRD_FDDI 774
+#endif
+ case ARPHRD_FDDI:
+ handle->linktype = DLT_FDDI;
+ handle->offset = 3;
+ break;
+
+#ifndef ARPHRD_ATM /* FIXME: How to #include this? */
+#define ARPHRD_ATM 19
+#endif
+ case ARPHRD_ATM:
+ /*
+ * The Classical IP implementation in ATM for Linux
+ * supports both what RFC 1483 calls "LLC Encapsulation",
+ * in which each packet has an LLC header, possibly
+ * with a SNAP header as well, prepended to it, and
+ * what RFC 1483 calls "VC Based Multiplexing", in which
+ * different virtual circuits carry different network
+ * layer protocols, and no header is prepended to packets.
+ *
+ * They both have an ARPHRD_ type of ARPHRD_ATM, so
+ * you can't use the ARPHRD_ type to find out whether
+ * captured packets will have an LLC header, and,
+ * while there's a socket ioctl to *set* the encapsulation
+ * type, there's no ioctl to *get* the encapsulation type.
+ *
+ * This means that
+ *
+ * programs that dissect Linux Classical IP frames
+ * would have to check for an LLC header and,
+ * depending on whether they see one or not, dissect
+ * the frame as LLC-encapsulated or as raw IP (I
+ * don't know whether there's any traffic other than
+ * IP that would show up on the socket, or whether
+ * there's any support for IPv6 in the Linux
+ * Classical IP code);
+ *
+ * filter expressions would have to compile into
+ * code that checks for an LLC header and does
+ * the right thing.
+ *
+ * Both of those are a nuisance - and, at least on systems
+ * that support PF_PACKET sockets, we don't have to put
+ * up with those nuisances; instead, we can just capture
+ * in cooked mode. That's what we'll do, if we can.
+ * Otherwise, we'll just fail.
+ */
+ if (cooked_ok)
+ handle->linktype = DLT_LINUX_SLL;
+ else
+ handle->linktype = -1;
+ break;
+
+#ifndef ARPHRD_IEEE80211 /* From Linux 2.4.6 */
+#define ARPHRD_IEEE80211 801
+#endif
+ case ARPHRD_IEEE80211:
+ handle->linktype = DLT_IEEE802_11;
+ break;
+
+#ifndef ARPHRD_IEEE80211_PRISM /* From Linux 2.4.18 */
+#define ARPHRD_IEEE80211_PRISM 802
+#endif
+ case ARPHRD_IEEE80211_PRISM:
+ handle->linktype = DLT_PRISM_HEADER;
+ break;
+
+#ifndef ARPHRD_IEEE80211_RADIOTAP /* new */
+#define ARPHRD_IEEE80211_RADIOTAP 803
+#endif
+ case ARPHRD_IEEE80211_RADIOTAP:
+ handle->linktype = DLT_IEEE802_11_RADIO;
+ break;
+
+ case ARPHRD_PPP:
+ /*
+ * Some PPP code in the kernel supplies no link-layer
+ * header whatsoever to PF_PACKET sockets; other PPP
+ * code supplies PPP link-layer headers ("syncppp.c");
+ * some PPP code might supply random link-layer
+ * headers (PPP over ISDN - there's code in Ethereal,
+ * for example, to cope with PPP-over-ISDN captures
+ * with which the Ethereal developers have had to cope,
+ * heuristically trying to determine which of the
+ * oddball link-layer headers particular packets have).
+ *
+ * As such, we just punt, and run all PPP interfaces
+ * in cooked mode, if we can; otherwise, we just treat
+ * it as DLT_RAW, for now - if somebody needs to capture,
+ * on a 2.0[.x] kernel, on PPP devices that supply a
+ * link-layer header, they'll have to add code here to
+ * map to the appropriate DLT_ type (possibly adding a
+ * new DLT_ type, if necessary).
+ */
+ if (cooked_ok)
+ handle->linktype = DLT_LINUX_SLL;
+ else {
+ /*
+ * XXX - handle ISDN types here? We can't fall
+ * back on cooked sockets, so we'd have to
+ * figure out from the device name what type of
+ * link-layer encapsulation it's using, and map
+ * that to an appropriate DLT_ value, meaning
+ * we'd map "isdnN" devices to DLT_RAW (they
+ * supply raw IP packets with no link-layer
+ * header) and "isdY" devices to a new DLT_I4L_IP
+ * type that has only an Ethernet packet type as
+ * a link-layer header.
+ *
+ * But sometimes we seem to get random crap
+ * in the link-layer header when capturing on
+ * ISDN devices....
+ */
+ handle->linktype = DLT_RAW;
+ }
+ break;
+
+#ifndef ARPHRD_CISCO
+#define ARPHRD_CISCO 513 /* previously ARPHRD_HDLC */
+#endif
+ case ARPHRD_CISCO:
+ handle->linktype = DLT_C_HDLC;
+ break;
+
+ /* Not sure if this is correct for all tunnels, but it
+ * works for CIPE */
+ case ARPHRD_TUNNEL:
+#ifndef ARPHRD_SIT
+#define ARPHRD_SIT 776 /* From Linux 2.2.13 */
+#endif
+ case ARPHRD_SIT:
+ case ARPHRD_CSLIP:
+ case ARPHRD_SLIP6:
+ case ARPHRD_CSLIP6:
+ case ARPHRD_ADAPT:
+ case ARPHRD_SLIP:
+#ifndef ARPHRD_RAWHDLC
+#define ARPHRD_RAWHDLC 518
+#endif
+ case ARPHRD_RAWHDLC:
+#ifndef ARPHRD_DLCI
+#define ARPHRD_DLCI 15
+#endif
+ case ARPHRD_DLCI:
+ /*
+ * XXX - should some of those be mapped to DLT_LINUX_SLL
+ * instead? Should we just map all of them to DLT_LINUX_SLL?
+ */
+ handle->linktype = DLT_RAW;
+ break;
+
+#ifndef ARPHRD_FRAD
+#define ARPHRD_FRAD 770
+#endif
+ case ARPHRD_FRAD:
+ handle->linktype = DLT_FRELAY;
+ break;
+
+ case ARPHRD_LOCALTLK:
+ handle->linktype = DLT_LTALK;
+ break;
+
+#ifndef ARPHRD_FCPP
+#define ARPHRD_FCPP 784
+#endif
+ case ARPHRD_FCPP:
+#ifndef ARPHRD_FCAL
+#define ARPHRD_FCAL 785
+#endif
+ case ARPHRD_FCAL:
+#ifndef ARPHRD_FCPL
+#define ARPHRD_FCPL 786
+#endif
+ case ARPHRD_FCPL:
+#ifndef ARPHRD_FCFABRIC
+#define ARPHRD_FCFABRIC 787
+#endif
+ case ARPHRD_FCFABRIC:
+ /*
+ * We assume that those all mean RFC 2625 IP-over-
+ * Fibre Channel, with the RFC 2625 header at
+ * the beginning of the packet.
+ */
+ handle->linktype = DLT_IP_OVER_FC;
+ break;
+
+#ifndef ARPHRD_IRDA
+#define ARPHRD_IRDA 783
+#endif
+ case ARPHRD_IRDA:
+ /* Don't expect IP packet out of this interfaces... */
+ handle->linktype = DLT_LINUX_IRDA;
+ /* We need to save packet direction for IrDA decoding,
+ * so let's use "Linux-cooked" mode. Jean II */
+ //handle->md.cooked = 1;
+ break;
+
+ /* ARPHRD_LAPD is unofficial and randomly allocated, if reallocation
+#ifndef ARPHRD_LAPD
+#define ARPHRD_LAPD 8445
+#endif
+ case ARPHRD_LAPD:
+ /* Don't expect IP packet out of this interfaces... */
+ handle->linktype = DLT_LINUX_LAPD;
+ break;
+
+#ifndef ARPHRD_NONE
+#define ARPHRD_NONE 0xFFFE
+#endif
+ case ARPHRD_NONE:
+ /*
+ * No link-layer header; packets are just IP
+ * packets, so use DLT_RAW.
+ */
+ handle->linktype = DLT_RAW;
+ break;
+
+#ifndef ARPHRD_IEEE802154
+#define ARPHRD_IEEE802154 804
+#endif
+ case ARPHRD_IEEE802154:
+ handle->linktype = DLT_IEEE802_15_4_NOFCS;
+ break;
+
+ default:
+ handle->linktype = -1;
+ break;
+ }
+}
+
+/* ===== Functions to interface to the newer kernels ================== */
+
+/*
+ * Try to open a packet socket using the new kernel PF_PACKET interface.
+ * Returns 1 on success, 0 on an error that means the new interface isn't
+ * present (so the old SOCK_PACKET interface should be tried), and a
+ * PCAP_ERROR_ value on an error that means that the old mechanism won't
+ * work either (so it shouldn't be tried).
+ */
+static int
+activate_new(pcap_t *handle)
+{
+#ifdef HAVE_PF_PACKET_SOCKETS
+ const char *device = handle->opt.source;
+ int is_any_device = (strcmp(device, "any") == 0);
+ int sock_fd = -1, arptype;
+#ifdef HAVE_PACKET_AUXDATA
+ int val;
+#endif
+ int err = 0;
+ struct packet_mreq mr;
+
+ /*
+ * Open a socket with protocol family packet. If the
+ * "any" device was specified, we open a SOCK_DGRAM
+ * socket for the cooked interface, otherwise we first
+ * try a SOCK_RAW socket for the raw interface.
+ */
+ sock_fd = is_any_device ?
+ socket(PF_PACKET, SOCK_DGRAM, htons(ETH_P_ALL)) :
+ socket(PF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
+
+ if (sock_fd == -1) {
+ if (errno == EINVAL || errno == EAFNOSUPPORT) {
+ /*
+ * We don't support PF_PACKET/SOCK_whatever
+ * sockets; try the old mechanism.
+ */
+ return 0;
+ }
+
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "socket: %s",
+ pcap_strerror(errno) );
+ if (errno == EPERM || errno == EACCES) {
+ /*
+ * You don't have permission to open the
+ * socket.
+ */
+ return PCAP_ERROR_PERM_DENIED;
+ } else {
+ /*
+ * Other error.
+ */
+ return PCAP_ERROR;
+ }
+ }
+
+ /* It seems the kernel supports the new interface. */
+ handle->md.sock_packet = 0;
+
+ /*
+ * Get the interface index of the loopback device.
+ * If the attempt fails, don't fail, just set the
+ * "md.lo_ifindex" to -1.
+ *
+ * XXX - can there be more than one device that loops
+ * packets back, i.e. devices other than "lo"? If so,
+ * we'd need to find them all, and have an array of
+ * indices for them, and check all of them in
+ * "pcap_read_packet()".
+ */
+ handle->md.lo_ifindex = iface_get_id(sock_fd, "lo", handle->errbuf);
+
+ /*
+ * Default value for offset to align link-layer payload
+ * on a 4-byte boundary.
+ */
+ handle->offset = 0;
+
+ /*
+ * What kind of frames do we have to deal with? Fall back
+ * to cooked mode if we have an unknown interface type
+ * or a type we know doesn't work well in raw mode.
+ */
+ if (!is_any_device) {
+ /* Assume for now we don't need cooked mode. */
+ handle->md.cooked = 0;
+
+ if (handle->opt.rfmon) {
+ /*
+ * We were asked to turn on monitor mode.
+ * Do so before we get the link-layer type,
+ * because entering monitor mode could change
+ * the link-layer type.
+ */
+ err = enter_rfmon_mode(handle, sock_fd, device);
+ if (err < 0) {
+ /* Hard failure */
+ close(sock_fd);
+ return err;
+ }
+ if (err == 0) {
+ /*
+ * Nothing worked for turning monitor mode
+ * on.
+ */
+ close(sock_fd);
+ return PCAP_ERROR_RFMON_NOTSUP;
+ }
+
+ /*
+ * Either monitor mode has been turned on for
+ * the device, or we've been given a different
+ * device to open for monitor mode. If we've
+ * been given a different device, use it.
+ */
+ if (handle->md.mondevice != NULL)
+ device = handle->md.mondevice;
+ }
+ arptype = iface_get_arptype(sock_fd, device, handle->errbuf);
+ if (arptype < 0) {
+ close(sock_fd);
+ return arptype;
+ }
+ map_arphrd_to_dlt(handle, sock_fd, arptype, device, 1);
+ if (handle->linktype == -1 ||
+ handle->linktype == DLT_LINUX_SLL ||
+ handle->linktype == DLT_LINUX_IRDA ||
+ handle->linktype == DLT_LINUX_LAPD ||
+ (handle->linktype == DLT_EN10MB &&
+ (strncmp("isdn", device, 4) == 0 ||
+ strncmp("isdY", device, 4) == 0))) {
+ /*
+ * Unknown interface type (-1), or a
+ * device we explicitly chose to run
+ * in cooked mode (e.g., PPP devices),
+ * or an ISDN device (whose link-layer
+ * type we can only determine by using
+ * APIs that may be different on different
+ * kernels) - reopen in cooked mode.
+ */
+ if (close(sock_fd) == -1) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "close: %s", pcap_strerror(errno));
+ return PCAP_ERROR;
+ }
+ sock_fd = socket(PF_PACKET, SOCK_DGRAM,
+ htons(ETH_P_ALL));
+ if (sock_fd == -1) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "socket: %s", pcap_strerror(errno));
+ if (errno == EPERM || errno == EACCES) {
+ /*
+ * You don't have permission to
+ * open the socket.
+ */
+ return PCAP_ERROR_PERM_DENIED;
+ } else {
+ /*
+ * Other error.
+ */
+ return PCAP_ERROR;
+ }
+ }
+ handle->md.cooked = 1;
+
+ /*
+ * Get rid of any link-layer type list
+ * we allocated - this only supports cooked
+ * capture.
+ */
+ if (handle->dlt_list != NULL) {
+ free(handle->dlt_list);
+ handle->dlt_list = NULL;
+ handle->dlt_count = 0;
+ }
+
+ if (handle->linktype == -1) {
+ /*
+ * Warn that we're falling back on
+ * cooked mode; we may want to
+ * update "map_arphrd_to_dlt()"
+ * to handle the new type.
+ */
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "arptype %d not "
+ "supported by libpcap - "
+ "falling back to cooked "
+ "socket",
+ arptype);
+ }
+
+ /*
+ * IrDA capture is not a real "cooked" capture,
+ * it's IrLAP frames, not IP packets. The
+ * same applies to LAPD capture.
+ */
+ if (handle->linktype != DLT_LINUX_IRDA &&
+ handle->linktype != DLT_LINUX_LAPD)
+ handle->linktype = DLT_LINUX_SLL;
+ }
+
+ handle->md.ifindex = iface_get_id(sock_fd, device,
+ handle->errbuf);
+ if (handle->md.ifindex == -1) {
+ close(sock_fd);
+ return PCAP_ERROR;
+ }
+
+ if ((err = iface_bind(sock_fd, handle->md.ifindex,
+ handle->errbuf)) != 1) {
+ close(sock_fd);
+ if (err < 0)
+ return err;
+ else
+ return 0; /* try old mechanism */
+ }
+ } else {
+ /*
+ * The "any" device.
+ */
+ if (handle->opt.rfmon) {
+ /*
+ * It doesn't support monitor mode.
+ */
+ return PCAP_ERROR_RFMON_NOTSUP;
+ }
+
+ /*
+ * It uses cooked mode.
+ */
+ handle->md.cooked = 1;
+ handle->linktype = DLT_LINUX_SLL;
+
+ /*
+ * We're not bound to a device.
+ * For now, we're using this as an indication
+ * that we can't transmit; stop doing that only
+ * if we figure out how to transmit in cooked
+ * mode.
+ */
+ handle->md.ifindex = -1;
+ }
+
+ /*
+ * Select promiscuous mode on if "promisc" is set.
+ *
+ * Do not turn allmulti mode on if we don't select
+ * promiscuous mode - on some devices (e.g., Orinoco
+ * wireless interfaces), allmulti mode isn't supported
+ * and the driver implements it by turning promiscuous
+ * mode on, and that screws up the operation of the
+ * card as a normal networking interface, and on no
+ * other platform I know of does starting a non-
+ * promiscuous capture affect which multicast packets
+ * are received by the interface.
+ */
+
+ /*
+ * Hmm, how can we set promiscuous mode on all interfaces?
+ * I am not sure if that is possible at all. For now, we
+ * silently ignore attempts to turn promiscuous mode on
+ * for the "any" device (so you don't have to explicitly
+ * disable it in programs such as tcpdump).
+ */
+
+ if (!is_any_device && handle->opt.promisc) {
+ memset(&mr, 0, sizeof(mr));
+ mr.mr_ifindex = handle->md.ifindex;
+ mr.mr_type = PACKET_MR_PROMISC;
+ if (setsockopt(sock_fd, SOL_PACKET, PACKET_ADD_MEMBERSHIP,
+ &mr, sizeof(mr)) == -1) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "setsockopt: %s", pcap_strerror(errno));
+ close(sock_fd);
+ return PCAP_ERROR;
+ }
+ }
+
+ /* Enable auxillary data if supported and reserve room for
+ * reconstructing VLAN headers. */
+#ifdef HAVE_PACKET_AUXDATA
+ val = 1;
+ if (setsockopt(sock_fd, SOL_PACKET, PACKET_AUXDATA, &val,
+ sizeof(val)) == -1 && errno != ENOPROTOOPT) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "setsockopt: %s", pcap_strerror(errno));
+ close(sock_fd);
+ return PCAP_ERROR;
+ }
+ handle->offset += VLAN_TAG_LEN;
+#endif /* HAVE_PACKET_AUXDATA */
+
+ /*
+ * This is a 2.2[.x] or later kernel (we know that
+ * because we're not using a SOCK_PACKET socket -
+ * PF_PACKET is supported only in 2.2 and later
+ * kernels).
+ *
+ * We can safely pass "recvfrom()" a byte count
+ * based on the snapshot length.
+ *
+ * If we're in cooked mode, make the snapshot length
+ * large enough to hold a "cooked mode" header plus
+ * 1 byte of packet data (so we don't pass a byte
+ * count of 0 to "recvfrom()").
+ */
+ if (handle->md.cooked) {
+ if (handle->snapshot < SLL_HDR_LEN + 1)
+ handle->snapshot = SLL_HDR_LEN + 1;
+ }
+ handle->bufsize = handle->snapshot;
+
+ /*
+ * Set the offset at which to insert VLAN tags.
+ */
+ switch (handle->linktype) {
+
+ case DLT_EN10MB:
+ handle->md.vlan_offset = 2 * ETH_ALEN;
+ break;
+
+ case DLT_LINUX_SLL:
+ handle->md.vlan_offset = 14;
+ break;
+
+ default:
+ handle->md.vlan_offset = -1; /* unknown */
+ break;
+ }
+
+ /* Save the socket FD in the pcap structure */
+ handle->fd = sock_fd;
+
+ return 1;
+#else
+ strncpy(ebuf,
+ "New packet capturing interface not supported by build "
+ "environment", PCAP_ERRBUF_SIZE);
+ return 0;
+#endif
+}
+
+#ifdef HAVE_PACKET_RING
+/*
+ * Attempt to activate with memory-mapped access.
+ *
+ * On success, returns 1, and sets *status to 0 if there are no warnings
+ * or to a PCAP_WARNING_ code if there is a warning.
+ *
+ * On failure due to lack of support for memory-mapped capture, returns
+ * 0.
+ *
+ * On error, returns -1, and sets *status to the appropriate error code;
+ * if that is PCAP_ERROR, sets handle->errbuf to the appropriate message.
+ */
+static int
+activate_mmap(pcap_t *handle, int *status)
+{
+ int ret;
+
+ /*
+ * Attempt to allocate a buffer to hold the contents of one
+ * packet, for use by the oneshot callback.
+ */
+ handle->md.oneshot_buffer = malloc(handle->snapshot);
+ if (handle->md.oneshot_buffer == NULL) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't allocate oneshot buffer: %s",
+ pcap_strerror(errno));
+ *status = PCAP_ERROR;
+ return -1;
+ }
+
+ if (handle->opt.buffer_size == 0) {
+ /* by default request 2M for the ring buffer */
+ handle->opt.buffer_size = 2*1024*1024;
+ }
+ ret = prepare_tpacket_socket(handle);
+ if (ret == -1) {
+ free(handle->md.oneshot_buffer);
+ *status = PCAP_ERROR;
+ return ret;
+ }
+ ret = create_ring(handle, status);
+ if (ret == 0) {
+ /*
+ * We don't support memory-mapped capture; our caller
+ * will fall back on reading from the socket.
+ */
+ free(handle->md.oneshot_buffer);
+ return 0;
+ }
+ if (ret == -1) {
+ /*
+ * Error attempting to enable memory-mapped capture;
+ * fail. create_ring() has set *status.
+ */
+ free(handle->md.oneshot_buffer);
+ return -1;
+ }
+
+ /*
+ * Success. *status has been set either to 0 if there are no
+ * warnings or to a PCAP_WARNING_ value if there is a warning.
+ *
+ * Override some defaults and inherit the other fields from
+ * activate_new.
+ * handle->offset is used to get the current position into the rx ring.
+ * handle->cc is used to store the ring size.
+ */
+ handle->read_op = pcap_read_linux_mmap;
+ handle->cleanup_op = pcap_cleanup_linux_mmap;
+ handle->setfilter_op = pcap_setfilter_linux_mmap;
+ handle->setnonblock_op = pcap_setnonblock_mmap;
+ handle->getnonblock_op = pcap_getnonblock_mmap;
+ handle->oneshot_callback = pcap_oneshot_mmap;
+ handle->selectable_fd = handle->fd;
+ return 1;
+}
+#else /* HAVE_PACKET_RING */
+static int
+activate_mmap(pcap_t *handle _U_, int *status _U_)
+{
+ return 0;
+}
+#endif /* HAVE_PACKET_RING */
+
+#ifdef HAVE_PACKET_RING
+/*
+ * Attempt to set the socket to version 2 of the memory-mapped header.
+ * Return 1 if we succeed or if we fail because version 2 isn't
+ * supported; return -1 on any other error, and set handle->errbuf.
+ */
+static int
+prepare_tpacket_socket(pcap_t *handle)
+{
+#ifdef HAVE_TPACKET2
+ socklen_t len;
+ int val;
+#endif
+
+ handle->md.tp_version = TPACKET_V1;
+ handle->md.tp_hdrlen = sizeof(struct tpacket_hdr);
+
+#ifdef HAVE_TPACKET2
+ /* Probe whether kernel supports TPACKET_V2 */
+ val = TPACKET_V2;
+ len = sizeof(val);
+ if (getsockopt(handle->fd, SOL_PACKET, PACKET_HDRLEN, &val, &len) < 0) {
+ if (errno == ENOPROTOOPT)
+ return 1; /* no - just drive on */
+
+ /* Yes - treat as a failure. */
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't get TPACKET_V2 header len on packet socket: %s",
+ pcap_strerror(errno));
+ return -1;
+ }
+ handle->md.tp_hdrlen = val;
+
+ val = TPACKET_V2;
+ if (setsockopt(handle->fd, SOL_PACKET, PACKET_VERSION, &val,
+ sizeof(val)) < 0) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't activate TPACKET_V2 on packet socket: %s",
+ pcap_strerror(errno));
+ return -1;
+ }
+ handle->md.tp_version = TPACKET_V2;
+
+ /* Reserve space for VLAN tag reconstruction */
+ val = VLAN_TAG_LEN;
+ if (setsockopt(handle->fd, SOL_PACKET, PACKET_RESERVE, &val,
+ sizeof(val)) < 0) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't set up reserve on packet socket: %s",
+ pcap_strerror(errno));
+ return -1;
+ }
+
+#endif /* HAVE_TPACKET2 */
+ return 1;
+}
+
+/*
+ * Attempt to set up memory-mapped access.
+ *
+ * On success, returns 1, and sets *status to 0 if there are no warnings
+ * or to a PCAP_WARNING_ code if there is a warning.
+ *
+ * On failure due to lack of support for memory-mapped capture, returns
+ * 0.
+ *
+ * On error, returns -1, and sets *status to the appropriate error code;
+ * if that is PCAP_ERROR, sets handle->errbuf to the appropriate message.
+ */
+static int
+create_ring(pcap_t *handle, int *status)
+{
+ unsigned i, j, frames_per_block;
+ struct tpacket_req req;
+ socklen_t len;
+ unsigned int sk_type, tp_reserve, maclen, tp_hdrlen, netoff, macoff;
+ unsigned int frame_size;
+
+ /*
+ * Start out assuming no warnings or errors.
+ */
+ *status = 0;
+
+ /* Note that with large snapshot length (say 64K, which is the default
+ * for recent versions of tcpdump, the value that "-s 0" has given
+ * for a long time with tcpdump, and the default in Wireshark/TShark),
+ * if we use the snapshot length to calculate the frame length,
+ * only a few frames will be available in the ring even with pretty
+ * large ring size (and a lot of memory will be unused).
+ *
+ * Ideally, we should choose a frame length based on the
+ * minimum of the specified snapshot length and the maximum
+ * packet size. That's not as easy as it sounds; consider, for
+ * example, an 802.11 interface in monitor mode, where the
+ * frame would include a radiotap header, where the maximum
+ * radiotap header length is device-dependent.
+ *
+ * So, for now, we just do this for Ethernet devices, where
+ * there's no metadata header, and the link-layer header is
+ * fixed length. We can get the maximum packet size by
+ * adding 18, the Ethernet header length plus the CRC length
+ * (just in case we happen to get the CRC in the packet), to
+ * the MTU of the interface; we fetch the MTU in the hopes
+ * that it reflects support for jumbo frames. (Even if the
+ * interface is just being used for passive snooping, the driver
+ * might set the size of buffers in the receive ring based on
+ * the MTU, so that the MTU limits the maximum size of packets
+ * that we can receive.)
+ *
+ * We don't do that if segmentation/fragmentation or receive
+ * offload are enabled, so we don't get rudely surprised by
+ * "packets" bigger than the MTU. */
+ frame_size = handle->snapshot;
+ if (handle->linktype == DLT_EN10MB) {
+ int mtu;
+ int offload;
+
+ offload = iface_get_offload(handle);
+ if (offload == -1) {
+ *status = PCAP_ERROR;
+ return -1;
+ }
+ if (!offload) {
+ mtu = iface_get_mtu(handle->fd, handle->opt.source,
+ handle->errbuf);
+ if (mtu == -1) {
+ *status = PCAP_ERROR;
+ return -1;
+ }
+ if (frame_size > mtu + 18)
+ frame_size = mtu + 18;
+ }
+ }
+
+ /* NOTE: calculus matching those in tpacket_rcv()
+ * in linux-2.6/net/packet/af_packet.c
+ */
+ len = sizeof(sk_type);
+ if (getsockopt(handle->fd, SOL_SOCKET, SO_TYPE, &sk_type, &len) < 0) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "getsockopt: %s", pcap_strerror(errno));
+ *status = PCAP_ERROR;
+ return -1;
+ }
+#ifdef PACKET_RESERVE
+ len = sizeof(tp_reserve);
+ if (getsockopt(handle->fd, SOL_PACKET, PACKET_RESERVE, &tp_reserve, &len) < 0) {
+ if (errno != ENOPROTOOPT) {
+ /*
+ * ENOPROTOOPT means "kernel doesn't support
+ * PACKET_RESERVE", in which case we fall back
+ * as best we can.
+ */
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "getsockopt: %s", pcap_strerror(errno));
+ *status = PCAP_ERROR;
+ return -1;
+ }
+ tp_reserve = 0; /* older kernel, reserve not supported */
+ }
+#else
+ tp_reserve = 0; /* older kernel, reserve not supported */
+#endif
+ maclen = (sk_type == SOCK_DGRAM) ? 0 : MAX_LINKHEADER_SIZE;
+ /* XXX: in the kernel maclen is calculated from
+ * LL_ALLOCATED_SPACE(dev) and vnet_hdr.hdr_len
+ * in: packet_snd() in linux-2.6/net/packet/af_packet.c
+ * then packet_alloc_skb() in linux-2.6/net/packet/af_packet.c
+ * then sock_alloc_send_pskb() in linux-2.6/net/core/sock.c
+ * but I see no way to get those sizes in userspace,
+ * like for instance with an ifreq ioctl();
+ * the best thing I've found so far is MAX_HEADER in the kernel
+ * part of linux-2.6/include/linux/netdevice.h
+ * which goes up to 128+48=176; since pcap-linux.c defines
+ * a MAX_LINKHEADER_SIZE of 256 which is greater than that,
+ * let's use it.. maybe is it even large enough to directly
+ * replace macoff..
+ */
+ tp_hdrlen = TPACKET_ALIGN(handle->md.tp_hdrlen) + sizeof(struct sockaddr_ll) ;
+ netoff = TPACKET_ALIGN(tp_hdrlen + (maclen < 16 ? 16 : maclen)) + tp_reserve;
+ /* NOTE: AFAICS tp_reserve may break the TPACKET_ALIGN of
+ * netoff, which contradicts
+ * linux-2.6/Documentation/networking/packet_mmap.txt
+ * documenting that:
+ * "- Gap, chosen so that packet data (Start+tp_net)
+ * aligns to TPACKET_ALIGNMENT=16"
+ */
+ /* NOTE: in linux-2.6/include/linux/skbuff.h:
+ * "CPUs often take a performance hit
+ * when accessing unaligned memory locations"
+ */
+ macoff = netoff - maclen;
+ req.tp_frame_size = TPACKET_ALIGN(macoff + frame_size);
+ req.tp_frame_nr = handle->opt.buffer_size/req.tp_frame_size;
+
+ /* compute the minumum block size that will handle this frame.
+ * The block has to be page size aligned.
+ * The max block size allowed by the kernel is arch-dependent and
+ * it's not explicitly checked here. */
+ req.tp_block_size = getpagesize();
+ while (req.tp_block_size < req.tp_frame_size)
+ req.tp_block_size <<= 1;
+
+ frames_per_block = req.tp_block_size/req.tp_frame_size;
+
+ /*
+ * PACKET_TIMESTAMP was added after linux/net_tstamp.h was,
+ * so we check for PACKET_TIMESTAMP. We check for
+ * linux/net_tstamp.h just in case a system somehow has
+ * PACKET_TIMESTAMP but not linux/net_tstamp.h; that might
+ * be unnecessary.
+ *
+ * SIOCSHWTSTAMP was introduced in the patch that introduced
+ * linux/net_tstamp.h, so we don't bother checking whether
+ * SIOCSHWTSTAMP is defined (if your Linux system has
+ * linux/net_tstamp.h but doesn't define SIOCSHWTSTAMP, your
+ * Linux system is badly broken).
+ */
+#if defined(HAVE_LINUX_NET_TSTAMP_H) && defined(PACKET_TIMESTAMP)
+ /*
+ * If we were told to do so, ask the kernel and the driver
+ * to use hardware timestamps.
+ *
+ * Hardware timestamps are only supported with mmapped
+ * captures.
+ */
+ if (handle->opt.tstamp_type == PCAP_TSTAMP_ADAPTER ||
+ handle->opt.tstamp_type == PCAP_TSTAMP_ADAPTER_UNSYNCED) {
+ struct hwtstamp_config hwconfig;
+ struct ifreq ifr;
+ int timesource;
+
+ /*
+ * Ask for hardware time stamps on all packets,
+ * including transmitted packets.
+ */
+ memset(&hwconfig, 0, sizeof(hwconfig));
+ hwconfig.tx_type = HWTSTAMP_TX_ON;
+ hwconfig.rx_filter = HWTSTAMP_FILTER_ALL;
+
+ memset(&ifr, 0, sizeof(ifr));
+ strcpy(ifr.ifr_name, handle->opt.source);
+ ifr.ifr_data = (void *)&hwconfig;
+
+ if (ioctl(handle->fd, SIOCSHWTSTAMP, &ifr) < 0) {
+ switch (errno) {
+
+ case EPERM:
+ /*
+ * Treat this as an error, as the
+ * user should try to run this
+ * with the appropriate privileges -
+ * and, if they can't, shouldn't
+ * try requesting hardware time stamps.
+ */
+ *status = PCAP_ERROR_PERM_DENIED;
+ return -1;
+
+ case EOPNOTSUPP:
+ /*
+ * Treat this as a warning, as the
+ * only way to fix the warning is to
+ * get an adapter that supports hardware
+ * time stamps. We'll just fall back
+ * on the standard host time stamps.
+ */
+ *status = PCAP_WARNING_TSTAMP_TYPE_NOTSUP;
+ break;
+
+ default:
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "SIOCSHWTSTAMP failed: %s",
+ pcap_strerror(errno));
+ *status = PCAP_ERROR;
+ return -1;
+ }
+ } else {
+ /*
+ * Well, that worked. Now specify the type of
+ * hardware time stamp we want for this
+ * socket.
+ */
+ if (handle->opt.tstamp_type == PCAP_TSTAMP_ADAPTER) {
+ /*
+ * Hardware timestamp, synchronized
+ * with the system clock.
+ */
+ timesource = SOF_TIMESTAMPING_SYS_HARDWARE;
+ } else {
+ /*
+ * PCAP_TSTAMP_ADAPTER_UNSYNCED - hardware
+ * timestamp, not synchronized with the
+ * system clock.
+ */
+ timesource = SOF_TIMESTAMPING_RAW_HARDWARE;
+ }
+ if (setsockopt(handle->fd, SOL_PACKET, PACKET_TIMESTAMP,
+ (void *)×ource, sizeof(timesource))) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't set PACKET_TIMESTAMP: %s",
+ pcap_strerror(errno));
+ *status = PCAP_ERROR;
+ return -1;
+ }
+ }
+ }
+#endif /* HAVE_LINUX_NET_TSTAMP_H && PACKET_TIMESTAMP */
+
+ /* ask the kernel to create the ring */
+retry:
+ req.tp_block_nr = req.tp_frame_nr / frames_per_block;
+
+ /* req.tp_frame_nr is requested to match frames_per_block*req.tp_block_nr */
+ req.tp_frame_nr = req.tp_block_nr * frames_per_block;
+
+ if (setsockopt(handle->fd, SOL_PACKET, PACKET_RX_RING,
+ (void *) &req, sizeof(req))) {
+ if ((errno == ENOMEM) && (req.tp_block_nr > 1)) {
+ /*
+ * Memory failure; try to reduce the requested ring
+ * size.
+ *
+ * We used to reduce this by half -- do 5% instead.
+ * That may result in more iterations and a longer
+ * startup, but the user will be much happier with
+ * the resulting buffer size.
+ */
+ if (req.tp_frame_nr < 20)
+ req.tp_frame_nr -= 1;
+ else
+ req.tp_frame_nr -= req.tp_frame_nr/20;
+ goto retry;
+ }
+ if (errno == ENOPROTOOPT) {
+ /*
+ * We don't have ring buffer support in this kernel.
+ */
+ return 0;
+ }
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't create rx ring on packet socket: %s",
+ pcap_strerror(errno));
+ *status = PCAP_ERROR;
+ return -1;
+ }
+
+ /* memory map the rx ring */
+ handle->md.mmapbuflen = req.tp_block_nr * req.tp_block_size;
+ handle->md.mmapbuf = mmap(0, handle->md.mmapbuflen,
+ PROT_READ|PROT_WRITE, MAP_SHARED, handle->fd, 0);
+ if (handle->md.mmapbuf == MAP_FAILED) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't mmap rx ring: %s", pcap_strerror(errno));
+
+ /* clear the allocated ring on error*/
+ destroy_ring(handle);
+ *status = PCAP_ERROR;
+ return -1;
+ }
+
+ /* allocate a ring for each frame header pointer*/
+ handle->cc = req.tp_frame_nr;
+ handle->buffer = malloc(handle->cc * sizeof(union thdr *));
+ if (!handle->buffer) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't allocate ring of frame headers: %s",
+ pcap_strerror(errno));
+
+ destroy_ring(handle);
+ *status = PCAP_ERROR;
+ return -1;
+ }
+
+ /* fill the header ring with proper frame ptr*/
+ handle->offset = 0;
+ for (i=0; i<req.tp_block_nr; ++i) {
+ void *base = &handle->md.mmapbuf[i*req.tp_block_size];
+ for (j=0; j<frames_per_block; ++j, ++handle->offset) {
+ RING_GET_FRAME(handle) = base;
+ base += req.tp_frame_size;
+ }
+ }
+
+ handle->bufsize = req.tp_frame_size;
+ handle->offset = 0;
+ return 1;
+}
+
+/* free all ring related resources*/
+static void
+destroy_ring(pcap_t *handle)
+{
+ /* tell the kernel to destroy the ring*/
+ struct tpacket_req req;
+ memset(&req, 0, sizeof(req));
+ setsockopt(handle->fd, SOL_PACKET, PACKET_RX_RING,
+ (void *) &req, sizeof(req));
+
+ /* if ring is mapped, unmap it*/
+ if (handle->md.mmapbuf) {
+ /* do not test for mmap failure, as we can't recover from any error */
+ munmap(handle->md.mmapbuf, handle->md.mmapbuflen);
+ handle->md.mmapbuf = NULL;
+ }
+}
+
+/*
+ * Special one-shot callback, used for pcap_next() and pcap_next_ex(),
+ * for Linux mmapped capture.
+ *
+ * The problem is that pcap_next() and pcap_next_ex() expect the packet
+ * data handed to the callback to be valid after the callback returns,
+ * but pcap_read_linux_mmap() has to release that packet as soon as
+ * the callback returns (otherwise, the kernel thinks there's still
+ * at least one unprocessed packet available in the ring, so a select()
+ * will immediately return indicating that there's data to process), so,
+ * in the callback, we have to make a copy of the packet.
+ *
+ * Yes, this means that, if the capture is using the ring buffer, using
+ * pcap_next() or pcap_next_ex() requires more copies than using
+ * pcap_loop() or pcap_dispatch(). If that bothers you, don't use
+ * pcap_next() or pcap_next_ex().
+ */
+static void
+pcap_oneshot_mmap(u_char *user, const struct pcap_pkthdr *h,
+ const u_char *bytes)
+{
+ struct oneshot_userdata *sp = (struct oneshot_userdata *)user;
+
+ *sp->hdr = *h;
+ memcpy(sp->pd->md.oneshot_buffer, bytes, h->caplen);
+ *sp->pkt = sp->pd->md.oneshot_buffer;
+}
+
+static void
+pcap_cleanup_linux_mmap( pcap_t *handle )
+{
+ destroy_ring(handle);
+ if (handle->md.oneshot_buffer != NULL) {
+ free(handle->md.oneshot_buffer);
+ handle->md.oneshot_buffer = NULL;
+ }
+ pcap_cleanup_linux(handle);
+}
+
+
+static int
+pcap_getnonblock_mmap(pcap_t *p, char *errbuf)
+{
+ /* use negative value of timeout to indicate non blocking ops */
+ return (p->md.timeout<0);
+}
+
+static int
+pcap_setnonblock_mmap(pcap_t *p, int nonblock, char *errbuf)
+{
+ /* map each value to the corresponding 2's complement, to
+ * preserve the timeout value provided with pcap_set_timeout */
+ if (nonblock) {
+ if (p->md.timeout >= 0) {
+ /*
+ * Timeout is non-negative, so we're not already
+ * in non-blocking mode; set it to the 2's
+ * complement, to make it negative, as an
+ * indication that we're in non-blocking mode.
+ */
+ p->md.timeout = p->md.timeout*-1 - 1;
+ }
+ } else {
+ if (p->md.timeout < 0) {
+ /*
+ * Timeout is negative, so we're not already
+ * in blocking mode; reverse the previous
+ * operation, to make the timeout non-negative
+ * again.
+ */
+ p->md.timeout = (p->md.timeout+1)*-1;
+ }
+ }
+ return 0;
+}
+
+static inline union thdr *
+pcap_get_ring_frame(pcap_t *handle, int status)
+{
+ union thdr h;
+
+ h.raw = RING_GET_FRAME(handle);
+ switch (handle->md.tp_version) {
+ case TPACKET_V1:
+ if (status != (h.h1->tp_status ? TP_STATUS_USER :
+ TP_STATUS_KERNEL))
+ return NULL;
+ break;
+#ifdef HAVE_TPACKET2
+ case TPACKET_V2:
+ if (status != (h.h2->tp_status ? TP_STATUS_USER :
+ TP_STATUS_KERNEL))
+ return NULL;
+ break;
+#endif
+ }
+ return h.raw;
+}
+
+#ifndef POLLRDHUP
+#define POLLRDHUP 0
+#endif
+
+static int
+pcap_read_linux_mmap(pcap_t *handle, int max_packets, pcap_handler callback,
+ u_char *user)
+{
+ int timeout;
+ int pkts = 0;
+ char c;
+
+ /* wait for frames availability.*/
+ if (!pcap_get_ring_frame(handle, TP_STATUS_USER)) {
+ struct pollfd pollinfo;
+ int ret;
+
+ pollinfo.fd = handle->fd;
+ pollinfo.events = POLLIN;
+
+ if (handle->md.timeout == 0)
+ timeout = -1; /* block forever */
+ else if (handle->md.timeout > 0)
+ timeout = handle->md.timeout; /* block for that amount of time */
+ else
+ timeout = 0; /* non-blocking mode - poll to pick up errors */
+ do {
+ ret = poll(&pollinfo, 1, timeout);
+ if (ret < 0 && errno != EINTR) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "can't poll on packet socket: %s",
+ pcap_strerror(errno));
+ return PCAP_ERROR;
+ } else if (ret > 0 &&
+ (pollinfo.revents & (POLLHUP|POLLRDHUP|POLLERR|POLLNVAL))) {
+ /*
+ * There's some indication other than
+ * "you can read on this descriptor" on
+ * the descriptor.
+ */
+ if (pollinfo.revents & (POLLHUP | POLLRDHUP)) {
+ snprintf(handle->errbuf,
+ PCAP_ERRBUF_SIZE,
+ "Hangup on packet socket");
+ return PCAP_ERROR;
+ }
+ if (pollinfo.revents & POLLERR) {
+ /*
+ * A recv() will give us the
+ * actual error code.
+ *
+ * XXX - make the socket non-blocking?
+ */
+ if (recv(handle->fd, &c, sizeof c,
+ MSG_PEEK) != -1)
+ continue; /* what, no error? */
+ if (errno == ENETDOWN) {
+ /*
+ * The device on which we're
+ * capturing went away.
+ *
+ * XXX - we should really return
+ * PCAP_ERROR_IFACE_NOT_UP,
+ * but pcap_dispatch() etc.
+ * aren't defined to return
+ * that.
+ */
+ snprintf(handle->errbuf,
+ PCAP_ERRBUF_SIZE,
+ "The interface went down");
+ } else {
+ snprintf(handle->errbuf,
+ PCAP_ERRBUF_SIZE,
+ "Error condition on packet socket: %s",
+ strerror(errno));
+ }
+ return PCAP_ERROR;
+ }
+ if (pollinfo.revents & POLLNVAL) {
+ snprintf(handle->errbuf,
+ PCAP_ERRBUF_SIZE,
+ "Invalid polling request on packet socket");
+ return PCAP_ERROR;
+ }
+ }
+ /* check for break loop condition on interrupted syscall*/
+ if (handle->break_loop) {
+ handle->break_loop = 0;
+ return PCAP_ERROR_BREAK;
+ }
+ } while (ret < 0);
+ }
+
+ /* non-positive values of max_packets are used to require all
+ * packets currently available in the ring */
+ while ((pkts < max_packets) || (max_packets <= 0)) {
+ int run_bpf;
+ struct sockaddr_ll *sll;
+ struct pcap_pkthdr pcaphdr;
+ unsigned char *bp;
+ union thdr h;
+ unsigned int tp_len;
+ unsigned int tp_mac;
+ unsigned int tp_snaplen;
+ unsigned int tp_sec;
+ unsigned int tp_usec;
+
+ h.raw = pcap_get_ring_frame(handle, TP_STATUS_USER);
+ if (!h.raw)
+ break;
+
+ switch (handle->md.tp_version) {
+ case TPACKET_V1:
+ tp_len = h.h1->tp_len;
+ tp_mac = h.h1->tp_mac;
+ tp_snaplen = h.h1->tp_snaplen;
+ tp_sec = h.h1->tp_sec;
+ tp_usec = h.h1->tp_usec;
+ break;
+#ifdef HAVE_TPACKET2
+ case TPACKET_V2:
+ tp_len = h.h2->tp_len;
+ tp_mac = h.h2->tp_mac;
+ tp_snaplen = h.h2->tp_snaplen;
+ tp_sec = h.h2->tp_sec;
+ tp_usec = h.h2->tp_nsec / 1000;
+ break;
+#endif
+ default:
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "unsupported tpacket version %d",
+ handle->md.tp_version);
+ return -1;
+ }
+ /* perform sanity check on internal offset. */
+ if (tp_mac + tp_snaplen > handle->bufsize) {
+ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE,
+ "corrupted frame on kernel ring mac "
+ "offset %u + caplen %u > frame len %d",
+ tp_mac, tp_snaplen, handle->bufsize);
+ return -1;
+ }
+
+ /* run filter on received packet
+ * If the kernel filtering is enabled we need to run the
+ * filter until all the frames present into the ring
+ * at filter creation time are processed.
+ * In such case md.use_bpf is used as a counter for the
+ * packet we need to filter.
+ * Note: alternatively it could be possible to stop applying
+ * the filter when the ring became empty, but it can possibly
+ * happen a lot later... */
+ bp = (unsigned char*)h.raw + tp_mac;
+ run_bpf = (!handle->md.use_bpf) ||
+ ((handle->md.use_bpf>1) && handle->md.use_bpf--);
+ if (run_bpf && handle->fcode.bf_insns &&
+ (bpf_filter(handle->fcode.bf_insns, bp,
+ tp_len, tp_snaplen) == 0))
+ goto skip;
+
+ /*
+ * Do checks based on packet direction.
+ */
+ sll = (void *)h.raw + TPACKET_ALIGN(handle->md.tp_hdrlen);
+ if (sll->sll_pkttype == PACKET_OUTGOING) {
+ /*
+ * Outgoing packet.
+ * If this is from the loopback device, reject it;
+ * we'll see the packet as an incoming packet as well,
+ * and we don't want to see it twice.
+ */
+ if (sll->sll_ifindex == handle->md.lo_ifindex)
+ goto skip;
+
+ /*
+ * If the user only wants incoming packets, reject it.
+ */
+ if (handle->direction == PCAP_D_IN)
+ goto skip;
+ } else {
+ /*
+ * Incoming packet.
+ * If the user only wants outgoing packets, reject it.
+ */
+ if (handle->direction == PCAP_D_OUT)
+ goto skip;
+ }
+
+ /* get required packet info from ring header */
+ pcaphdr.ts.tv_sec = tp_sec;
+ pcaphdr.ts.tv_usec = tp_usec;
+ pcaphdr.caplen = tp_snaplen;
+ pcaphdr.len = tp_len;
+
+ /* if required build in place the sll header*/
+ if (handle->md.cooked) {
+ struct sll_header *hdrp;
projects / libpcap / blobdiff