X-Git-Url: https://git.tcpdump.org/libpcap/blobdiff_plain/6831a814896d7f1c31bfa68af85028e8d4182543..0177044b9f53b8d9d4fa80c5846a62f9d5849e8b:/pcap-linux.c diff --git a/pcap-linux.c b/pcap-linux.c index 08bd910e..61d4db07 100644 --- a/pcap-linux.c +++ b/pcap-linux.c @@ -122,6 +122,7 @@ static const char rcsid[] _U_ = #include #include #include +#include #include #include #include @@ -135,6 +136,12 @@ static const char rcsid[] _U_ = #include #include #include +#include + +#ifdef HAVE_LINUX_NET_TSTAMP_H +#include +#include +#endif /* * Got Wireless Extensions? @@ -156,6 +163,14 @@ static const char rcsid[] _U_ = #include #endif /* HAVE_LIBNL */ +/* + * Got ethtool support? + */ +#ifdef HAVE_LINUX_ETHTOOL_H +#include +#include +#endif /* HAVE_LINUX_ETHTOOL_H */ + #include "pcap-int.h" #include "pcap/sll.h" #include "pcap/vlan.h" @@ -168,6 +183,10 @@ static const char rcsid[] _U_ = #include "pcap-septel.h" #endif /* HAVE_SEPTEL_API */ +#ifdef HAVE_SNF_API +#include "pcap-snf.h" +#endif /* HAVE_SNF_API */ + #ifdef PCAP_SUPPORT_USB #include "pcap-usb-linux.h" #endif @@ -289,7 +308,7 @@ static short int map_packet_type_to_sll_type(short int); static int pcap_activate_linux(pcap_t *); static int activate_old(pcap_t *); static int activate_new(pcap_t *); -static int activate_mmap(pcap_t *); +static int activate_mmap(pcap_t *, int *); static int pcap_can_set_rfmon_linux(pcap_t *); static int pcap_read_linux(pcap_t *, int, pcap_handler, u_char *); static int pcap_read_packet(pcap_t *, pcap_handler, u_char *); @@ -309,7 +328,7 @@ union thdr { #define RING_GET_FRAME(h) (((union thdr **)h->buffer)[h->offset]) static void destroy_ring(pcap_t *handle); -static int create_ring(pcap_t *handle); +static int create_ring(pcap_t *handle, int *status); static int prepare_tpacket_socket(pcap_t *handle); static void pcap_cleanup_linux_mmap(pcap_t *); static int pcap_read_linux_mmap(pcap_t *, int, pcap_handler , u_char *); @@ -325,7 +344,7 @@ static void pcap_oneshot_mmap(u_char *user, const struct pcap_pkthdr *h, */ #ifdef HAVE_PF_PACKET_SOCKETS static int iface_get_id(int fd, const char *device, char *ebuf); -#endif +#endif /* HAVE_PF_PACKET_SOCKETS */ static int iface_get_mtu(int fd, const char *device, char *ebuf); static int iface_get_arptype(int fd, const char *device, char *ebuf); #ifdef HAVE_PF_PACKET_SOCKETS @@ -336,6 +355,7 @@ static int has_wext(int sock_fd, const char *device, char *ebuf); static int enter_rfmon_mode(pcap_t *handle, int sock_fd, const char *device); #endif /* HAVE_PF_PACKET_SOCKETS */ +static int iface_get_offload(pcap_t *handle); static int iface_bind_old(int fd, const char *device, char *ebuf); #ifdef SO_ATTACH_FILTER @@ -349,7 +369,7 @@ static struct sock_filter total_insn = BPF_STMT(BPF_RET | BPF_K, 0); static struct sock_fprog total_fcode = { 1, &total_insn }; -#endif +#endif /* SO_ATTACH_FILTER */ pcap_t * pcap_create(const char *device, char *ebuf) @@ -374,6 +394,13 @@ pcap_create(const char *device, char *ebuf) } #endif /* HAVE_SEPTEL_API */ +#ifdef HAVE_SNF_API + handle = snf_create(device, ebuf); + if (strstr(device, "snf") || handle != NULL) + return handle; + +#endif /* HAVE_SNF_API */ + #ifdef PCAP_SUPPORT_BT if (strstr(device, "bluetooth")) { return bt_create(device, ebuf); @@ -398,52 +425,73 @@ pcap_create(const char *device, char *ebuf) handle->activate_op = pcap_activate_linux; handle->can_set_rfmon_op = pcap_can_set_rfmon_linux; +#if defined(HAVE_LINUX_NET_TSTAMP_H) && defined(PACKET_TIMESTAMP) + /* + * We claim that we support: + * + * software time stamps, with no details about their precision; + * hardware time stamps, synced to the host time; + * hardware time stamps, not synced to the host time. + * + * XXX - we can't ask a device whether it supports + * hardware time stamps, so we just claim all devices do. + */ + handle->tstamp_type_count = 3; + handle->tstamp_type_list = malloc(3 * sizeof(u_int)); + if (handle->tstamp_type_list == NULL) { + free(handle); + return NULL; + } + handle->tstamp_type_list[0] = PCAP_TSTAMP_HOST; + handle->tstamp_type_list[1] = PCAP_TSTAMP_ADAPTER; + handle->tstamp_type_list[2] = PCAP_TSTAMP_ADAPTER_UNSYNCED; +#endif + return handle; } #ifdef HAVE_LIBNL /* - * - * If interface {if} is a mac80211 driver, the file - * /sys/class/net/{if}/phy80211 is a symlink to - * /sys/class/ieee80211/{phydev}, for some {phydev}. - * - * On Fedora 9, with a 2.6.26.3-29 kernel, my Zydas stick, at - * least, has a "wmaster0" device and a "wlan0" device; the - * latter is the one with the IP address. Both show up in - * "tcpdump -D" output. Capturing on the wmaster0 device - * captures with 802.11 headers. - * - * airmon-ng searches through /sys/class/net for devices named - * monN, starting with mon0; as soon as one *doesn't* exist, - * it chooses that as the monitor device name. If the "iw" - * command exists, it does "iw dev {if} interface add {monif} - * type monitor", where {monif} is the monitor device. It - * then (sigh) sleeps .1 second, and then configures the - * device up. Otherwise, if /sys/class/ieee80211/{phydev}/add_iface - * is a file, it writes {mondev}, without a newline, to that file, - * and again (sigh) sleeps .1 second, and then iwconfig's that - * device into monitor mode and configures it up. Otherwise, - * you can't do monitor mode. - * - * All these devices are "glued" together by having the - * /sys/class/net/{device}/phy80211 links pointing to the same - * place, so, given a wmaster, wlan, or mon device, you can - * find the other devices by looking for devices with - * the same phy80211 link. - * - * To turn monitor mode off, delete the monitor interface, - * either with "iw dev {monif} interface del" or by sending - * {monif}, with no NL, down /sys/class/ieee80211/{phydev}/remove_iface - * - * Note: if you try to create a monitor device named "monN", and - * there's already a "monN" device, it fails, as least with - * the netlink interface (which is what iw uses), with a return - * value of -ENFILE. (Return values are negative errnos.) We - * could probably use that to find an unused device. - * - * Yes, you can have multiple monitor devices for a given - * physical device. + * If interface {if} is a mac80211 driver, the file + * /sys/class/net/{if}/phy80211 is a symlink to + * /sys/class/ieee80211/{phydev}, for some {phydev}. + * + * On Fedora 9, with a 2.6.26.3-29 kernel, my Zydas stick, at + * least, has a "wmaster0" device and a "wlan0" device; the + * latter is the one with the IP address. Both show up in + * "tcpdump -D" output. Capturing on the wmaster0 device + * captures with 802.11 headers. + * + * airmon-ng searches through /sys/class/net for devices named + * monN, starting with mon0; as soon as one *doesn't* exist, + * it chooses that as the monitor device name. If the "iw" + * command exists, it does "iw dev {if} interface add {monif} + * type monitor", where {monif} is the monitor device. It + * then (sigh) sleeps .1 second, and then configures the + * device up. Otherwise, if /sys/class/ieee80211/{phydev}/add_iface + * is a file, it writes {mondev}, without a newline, to that file, + * and again (sigh) sleeps .1 second, and then iwconfig's that + * device into monitor mode and configures it up. Otherwise, + * you can't do monitor mode. + * + * All these devices are "glued" together by having the + * /sys/class/net/{device}/phy80211 links pointing to the same + * place, so, given a wmaster, wlan, or mon device, you can + * find the other devices by looking for devices with + * the same phy80211 link. + * + * To turn monitor mode off, delete the monitor interface, + * either with "iw dev {monif} interface del" or by sending + * {monif}, with no NL, down /sys/class/ieee80211/{phydev}/remove_iface + * + * Note: if you try to create a monitor device named "monN", and + * there's already a "monN" device, it fails, as least with + * the netlink interface (which is what iw uses), with a return + * value of -ENFILE. (Return values are negative errnos.) We + * could probably use that to find an unused device. + * + * Yes, you can have multiple monitor devices for a given + * physical device. */ /* @@ -488,8 +536,41 @@ get_mac80211_phydev(pcap_t *handle, const char *device, char *phydev_path, return 1; } +#ifdef HAVE_LIBNL_2_x +#define get_nl_errmsg nl_geterror +#else +/* libnl 2.x compatibility code */ + +#define nl_sock nl_handle + +static inline struct nl_handle * +nl_socket_alloc(void) +{ + return nl_handle_alloc(); +} + +static inline void +nl_socket_free(struct nl_handle *h) +{ + nl_handle_destroy(h); +} + +#define get_nl_errmsg strerror + +static inline int +__genl_ctrl_alloc_cache(struct nl_handle *h, struct nl_cache **cache) +{ + struct nl_cache *tmp = genl_ctrl_alloc_cache(h); + if (!tmp) + return -ENOMEM; + *cache = tmp; + return 0; +} +#define genl_ctrl_alloc_cache __genl_ctrl_alloc_cache +#endif /* !HAVE_LIBNL_2_x */ + struct nl80211_state { - struct nl_handle *nl_handle; + struct nl_sock *nl_sock; struct nl_cache *nl_cache; struct genl_family *nl80211; }; @@ -497,23 +578,26 @@ struct nl80211_state { static int nl80211_init(pcap_t *handle, struct nl80211_state *state, const char *device) { - state->nl_handle = nl_handle_alloc(); - if (!state->nl_handle) { + int err; + + state->nl_sock = nl_socket_alloc(); + if (!state->nl_sock) { snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "%s: failed to allocate netlink handle", device); return PCAP_ERROR; } - if (genl_connect(state->nl_handle)) { + if (genl_connect(state->nl_sock)) { snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "%s: failed to connect to generic netlink", device); goto out_handle_destroy; } - state->nl_cache = genl_ctrl_alloc_cache(state->nl_handle); - if (!state->nl_cache) { + err = genl_ctrl_alloc_cache(state->nl_sock, &state->nl_cache); + if (err < 0) { snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, - "%s: failed to allocate generic netlink cache", device); + "%s: failed to allocate generic netlink cache: %s", + device, get_nl_errmsg(-err)); goto out_handle_destroy; } @@ -529,7 +613,7 @@ nl80211_init(pcap_t *handle, struct nl80211_state *state, const char *device) out_cache_free: nl_cache_free(state->nl_cache); out_handle_destroy: - nl_handle_destroy(state->nl_handle); + nl_socket_free(state->nl_sock); return PCAP_ERROR; } @@ -538,7 +622,7 @@ nl80211_cleanup(struct nl80211_state *state) { genl_family_put(state->nl80211); nl_cache_free(state->nl_cache); - nl_handle_destroy(state->nl_handle); + nl_socket_free(state->nl_sock); } static int @@ -566,12 +650,19 @@ add_mon_if(pcap_t *handle, int sock_fd, struct nl80211_state *state, NLA_PUT_STRING(msg, NL80211_ATTR_IFNAME, mondevice); NLA_PUT_U32(msg, NL80211_ATTR_IFTYPE, NL80211_IFTYPE_MONITOR); - err = nl_send_auto_complete(state->nl_handle, msg); + err = nl_send_auto_complete(state->nl_sock, msg); if (err < 0) { +#ifdef HAVE_LIBNL_2_x + if (err == -NLE_FAILURE) { +#else if (err == -ENFILE) { +#endif /* * Device not available; our caller should just - * keep trying. + * keep trying. (libnl 2.x maps ENFILE to + * NLE_FAILURE; it can also map other errors + * to that, but there's not much we can do + * about that.) */ nlmsg_free(msg); return 0; @@ -582,17 +673,24 @@ add_mon_if(pcap_t *handle, int sock_fd, struct nl80211_state *state, */ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "%s: nl_send_auto_complete failed adding %s interface: %s", - device, mondevice, strerror(-err)); + device, mondevice, get_nl_errmsg(-err)); nlmsg_free(msg); return PCAP_ERROR; } } - err = nl_wait_for_ack(state->nl_handle); + err = nl_wait_for_ack(state->nl_sock); if (err < 0) { +#ifdef HAVE_LIBNL_2_x + if (err == -NLE_FAILURE) { +#else if (err == -ENFILE) { +#endif /* * Device not available; our caller should just - * keep trying. + * keep trying. (libnl 2.x maps ENFILE to + * NLE_FAILURE; it can also map other errors + * to that, but there's not much we can do + * about that.) */ nlmsg_free(msg); return 0; @@ -603,7 +701,7 @@ add_mon_if(pcap_t *handle, int sock_fd, struct nl80211_state *state, */ snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "%s: nl_wait_for_ack failed adding %s interface: %s", - device, mondevice, strerror(-err)); + device, mondevice, get_nl_errmsg(-err)); nlmsg_free(msg); return PCAP_ERROR; } @@ -646,47 +744,21 @@ del_mon_if(pcap_t *handle, int sock_fd, struct nl80211_state *state, 0, NL80211_CMD_DEL_INTERFACE, 0); NLA_PUT_U32(msg, NL80211_ATTR_IFINDEX, ifindex); - err = nl_send_auto_complete(state->nl_handle, msg); + err = nl_send_auto_complete(state->nl_sock, msg); if (err < 0) { - if (err == -ENFILE) { - /* - * Device not available; our caller should just - * keep trying. - */ - nlmsg_free(msg); - return 0; - } else { - /* - * Real failure, not just "that device is not - * available. - */ - snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, - "%s: nl_send_auto_complete failed deleting %s interface: %s", - device, mondevice, strerror(-err)); - nlmsg_free(msg); - return PCAP_ERROR; - } + snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, + "%s: nl_send_auto_complete failed deleting %s interface: %s", + device, mondevice, get_nl_errmsg(-err)); + nlmsg_free(msg); + return PCAP_ERROR; } - err = nl_wait_for_ack(state->nl_handle); + err = nl_wait_for_ack(state->nl_sock); if (err < 0) { - if (err == -ENFILE) { - /* - * Device not available; our caller should just - * keep trying. - */ - nlmsg_free(msg); - return 0; - } else { - /* - * Real failure, not just "that device is not - * available. - */ - snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, - "%s: nl_wait_for_ack failed adding %s interface: %s", - device, mondevice, strerror(-err)); - nlmsg_free(msg); - return PCAP_ERROR; - } + snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, + "%s: nl_wait_for_ack failed adding %s interface: %s", + device, mondevice, get_nl_errmsg(-err)); + nlmsg_free(msg); + return PCAP_ERROR; } /* @@ -1154,34 +1226,46 @@ pcap_activate_linux(pcap_t *handle) * to be compatible with older kernels for a while so we are * trying both methods with the newer method preferred. */ - - if ((status = activate_new(handle)) == 1) { + status = activate_new(handle); + if (status < 0) { + /* + * Fatal error with the new way; just fail. + * status has the error return; if it's PCAP_ERROR, + * handle->errbuf has been set appropriately. + */ + goto fail; + } + if (status == 1) { /* * Success. * Try to use memory-mapped access. */ - switch (activate_mmap(handle)) { + switch (activate_mmap(handle, &status)) { case 1: - /* we succeeded; nothing more to do */ - return 0; + /* + * We succeeded. status has been + * set to the status to return, + * which might be 0, or might be + * a PCAP_WARNING_ value. + */ + return status; case 0: /* * Kernel doesn't support it - just continue * with non-memory-mapped access. */ - status = 0; break; case -1: /* - * We failed to set up to use it, or kernel - * supports it, but we failed to enable it; - * return an error. handle->errbuf contains - * an error message. + * We failed to set up to use it, or the kernel + * supports it, but we failed to enable it. + * status has been set to the error status to + * return and, if it's PCAP_ERROR, handle->errbuf + * contains the error message. */ - status = PCAP_ERROR; goto fail; } } @@ -1195,18 +1279,12 @@ pcap_activate_linux(pcap_t *handle) */ goto fail; } - } else { - /* - * Fatal error with the new way; just fail. - * status has the error return; if it's PCAP_ERROR, - * handle->errbuf has been set appropriately. - */ - goto fail; } /* * We set up the socket, but not with memory-mapped access. */ + status = 0; if (handle->opt.buffer_size != 0) { /* * Set the socket buffer size to the specified value. @@ -1774,9 +1852,8 @@ pcap_stats_linux(pcap_t *handle, struct pcap_stat *stats) return 0; } -#ifdef HAVE_PROC_NET_DEV /* - * Get from "/proc/net/dev" all interfaces listed there; if they're + * 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. * @@ -1785,8 +1862,142 @@ pcap_stats_linux(pcap_t *handle, struct pcap_stat *stats) * 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 "/proc/net/dev"; we just leave - * the list of interfaces as is. + * 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 *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 && errno == ENOENT) + return (0); + + /* + * 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 directories (".", "..", and any subdirectories). + */ + if (ent->d_type == DT_DIR) + 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) @@ -1795,14 +2006,14 @@ scan_proc_net_dev(pcap_if_t **devlistp, char *errbuf) int fd; char linebuf[512]; int linenum; - unsigned char *p; + 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) + if (proc_net_f == NULL && errno == ENOENT) return (0); /* @@ -1812,6 +2023,7 @@ scan_proc_net_dev(pcap_if_t **devlistp, char *errbuf) if (fd < 0) { (void)snprintf(errbuf, PCAP_ERRBUF_SIZE, "socket: %s", pcap_strerror(errno)); + (void)fclose(proc_net_f); return (-1); } @@ -1828,7 +2040,7 @@ scan_proc_net_dev(pcap_if_t **devlistp, char *errbuf) /* * Skip leading white space. */ - while (*p != '\0' && isspace(*p)) + while (*p != '\0' && isascii(*p) && isspace(*p)) p++; if (*p == '\0' || *p == '\n') continue; /* blank line */ @@ -1837,7 +2049,7 @@ scan_proc_net_dev(pcap_if_t **devlistp, char *errbuf) * Get the interface name. */ q = &name[0]; - while (*p != '\0' && !isspace(*p)) { + while (*p != '\0' && isascii(*p) && !isspace(*p)) { if (*p == ':') { /* * This could be the separator between a @@ -1851,7 +2063,7 @@ scan_proc_net_dev(pcap_if_t **devlistp, char *errbuf) * next field. */ saveq = q; - while (isdigit(*p)) + while (isascii(*p) && isdigit(*p)) *q++ = *p++; if (*p != ':') { /* @@ -1914,7 +2126,6 @@ scan_proc_net_dev(pcap_if_t **devlistp, char *errbuf) (void)fclose(proc_net_f); return (ret); } -#endif /* HAVE_PROC_NET_DEV */ /* * Description string for the "any" device. @@ -1924,18 +2135,26 @@ static const char any_descr[] = "Pseudo-device that captures on all interfaces"; int pcap_platform_finddevs(pcap_if_t **alldevsp, char *errbuf) { -#ifdef HAVE_PROC_NET_DEV + int ret; + /* - * Read "/proc/net/dev", and add to the list of interfaces all + * 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 "/proc/net/dev" + * interfaces with no addresses, so you need to read "/sys/class/net" * to get the names of the rest of the interfaces. */ - if (scan_proc_net_dev(alldevsp, errbuf) == -1) - return (-1); -#endif + 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. @@ -1959,6 +2178,11 @@ pcap_platform_finddevs(pcap_if_t **alldevsp, char *errbuf) 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. @@ -2070,6 +2294,30 @@ pcap_setfilter_linux_common(pcap_t *handle, struct bpf_program *filter, } } + /* + * 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) { @@ -2145,7 +2393,6 @@ pcap_setdirection_linux(pcap_t *handle, pcap_direction_t d) return -1; } - #ifdef HAVE_PF_PACKET_SOCKETS /* * Map the PACKET_ value to a LINUX_SLL_ value; we @@ -2492,6 +2739,13 @@ static void map_arphrd_to_dlt(pcap_t *handle, int arptype, int cooked_ok) 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; @@ -2793,10 +3047,22 @@ activate_new(pcap_t *handle) #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) +activate_mmap(pcap_t *handle, int *status) { -#ifdef HAVE_PACKET_RING int ret; /* @@ -2808,7 +3074,8 @@ activate_mmap(pcap_t *handle) snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "can't allocate oneshot buffer: %s", pcap_strerror(errno)); - return PCAP_ERROR; + *status = PCAP_ERROR; + return -1; } if (handle->opt.buffer_size == 0) { @@ -2816,20 +3083,38 @@ activate_mmap(pcap_t *handle) handle->opt.buffer_size = 2*1024*1024; } ret = prepare_tpacket_socket(handle); - if (ret != 1) { + if (ret == -1) { free(handle->md.oneshot_buffer); + *status = PCAP_ERROR; return ret; } - ret = create_ring(handle); - if (ret != 1) { + 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 ret; + 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; } - /* 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 */ + /* + * 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; @@ -2838,12 +3123,21 @@ activate_mmap(pcap_t *handle) 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 */ } +#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) { @@ -2895,20 +3189,140 @@ prepare_tpacket_socket(pcap_t *handle) 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) +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; - /* Note that with large snapshot (say 64K) only a few frames - * will be available in the ring even with pretty large ring size - * (and a lot of memory will be unused). - * The snap len should be carefully chosen to achive best - * performance */ - req.tp_frame_size = TPACKET_ALIGN(handle->snapshot + - TPACKET_ALIGN(handle->md.tp_hdrlen) + - sizeof(struct sockaddr_ll)); + /* + * 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. @@ -2921,6 +3335,109 @@ create_ring(pcap_t *handle) 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; @@ -2955,6 +3472,7 @@ retry: snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, "can't create rx ring on packet socket: %s", pcap_strerror(errno)); + *status = PCAP_ERROR; return -1; } @@ -2968,6 +3486,7 @@ retry: /* clear the allocated ring on error*/ destroy_ring(handle); + *status = PCAP_ERROR; return -1; } @@ -2980,6 +3499,7 @@ retry: pcap_strerror(errno)); destroy_ring(handle); + *status = PCAP_ERROR; return -1; } @@ -4183,6 +4703,97 @@ enter_rfmon_mode(pcap_t *handle, int sock_fd, const char *device) return 0; } +/* + * Find out if we have any form of fragmentation/reassembly offloading. + */ +#ifdef SIOCETHTOOL +static int +iface_ethtool_ioctl(pcap_t *handle, int cmd, const char *cmdname) +{ + struct ifreq ifr; + struct ethtool_value eval; + + memset(&ifr, 0, sizeof(ifr)); + strncpy(ifr.ifr_name, handle->opt.source, sizeof(ifr.ifr_name)); + eval.cmd = cmd; + ifr.ifr_data = (caddr_t)&eval; + if (ioctl(handle->fd, SIOCETHTOOL, &ifr) == -1) { + if (errno == EOPNOTSUPP) { + /* + * OK, let's just return 0, which, in our + * case, either means "no, what we're asking + * about is not enabled" or "all the flags + * are clear (i.e., nothing is enabled)". + */ + return 0; + } + snprintf(handle->errbuf, PCAP_ERRBUF_SIZE, + "%s: SIOETHTOOL(%s) ioctl failed: %s", handle->opt.source, + cmdname, strerror(errno)); + return -1; + } + return eval.data; +} + +static int +iface_get_offload(pcap_t *handle) +{ + int ret; + + ret = iface_ethtool_ioctl(handle, ETHTOOL_GTSO, "ETHTOOL_GTSO"); + if (ret == -1) + return -1; + if (ret) + return 1; /* TCP segmentation offloading on */ + + ret = iface_ethtool_ioctl(handle, ETHTOOL_GUFO, "ETHTOOL_GUFO"); + if (ret == -1) + return -1; + if (ret) + return 1; /* UDP fragmentation offloading on */ + + /* + * XXX - will this cause large unsegmented packets to be + * handed to PF_PACKET sockets on transmission? If not, + * this need not be checked. + */ + ret = iface_ethtool_ioctl(handle, ETHTOOL_GGSO, "ETHTOOL_GGSO"); + if (ret == -1) + return -1; + if (ret) + return 1; /* generic segmentation offloading on */ + + ret = iface_ethtool_ioctl(handle, ETHTOOL_GFLAGS, "ETHTOOL_GFLAGS"); + if (ret == -1) + return -1; + if (ret & ETH_FLAG_LRO) + return 1; /* large receive offloading on */ + + /* + * XXX - will this cause large reassembled packets to be + * handed to PF_PACKET sockets on receipt? If not, + * this need not be checked. + */ + ret = iface_ethtool_ioctl(handle, ETHTOOL_GGRO, "ETHTOOL_GGRO"); + if (ret == -1) + return -1; + if (ret) + return 1; /* generic (large) receive offloading on */ + + return 0; +} +#else /* SIOCETHTOOL */ +static int +iface_get_offload(pcap_t *handle _U_) +{ + /* + * XXX - do we need to get this information if we don't + * have the ethtool ioctls? If so, how do we do that? + */ + return 0; +} +#endif /* SIOCETHTOOL */ + #endif /* HAVE_PF_PACKET_SOCKETS */ /* ===== Functions to interface to the older kernels ================== */