/*
- * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
+ * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include <config.h>
#include "socket-util.h"
+#include <sys/types.h>
+#include <netinet/in.h>
#include <arpa/inet.h>
+#include <ctype.h>
#include <errno.h>
#include <fcntl.h>
#include <net/if.h>
#include <netdb.h>
+#include <netinet/tcp.h>
#include <poll.h>
#include <stddef.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#include <sys/ioctl.h>
-#include <sys/resource.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/uio.h>
#include <sys/un.h>
#include <unistd.h>
-#include "dynamic-string.h"
-#include "fatal-signal.h"
+#include "openvswitch/dynamic-string.h"
#include "ovs-thread.h"
#include "packets.h"
-#include "poll-loop.h"
+#include "openvswitch/poll-loop.h"
#include "util.h"
-#include "vlog.h"
-#if AF_PACKET && LINUX_DATAPATH
+#include "openvswitch/vlog.h"
+#ifdef __linux__
#include <linux/if_packet.h>
#endif
#ifdef HAVE_NETLINK
#include "netlink-protocol.h"
#include "netlink-socket.h"
#endif
+#include "dns-resolve.h"
VLOG_DEFINE_THIS_MODULE(socket_util);
-/* #ifdefs make it a pain to maintain code: you have to try to build both ways.
- * Thus, this file compiles all of the code regardless of the target, by
- * writing "if (LINUX_DATAPATH)" instead of "#ifdef __linux__". */
-#ifndef LINUX_DATAPATH
-#define LINUX_DATAPATH 0
-#endif
-
-#ifndef O_DIRECTORY
-#define O_DIRECTORY 0
-#endif
-
-/* Maximum length of the sun_path member in a struct sockaddr_un, excluding
- * space for a null terminator. */
-#define MAX_UN_LEN (sizeof(((struct sockaddr_un *) 0)->sun_path) - 1)
-
static int getsockopt_int(int fd, int level, int option, const char *optname,
int *valuep);
+static struct sockaddr_in *sin_cast(const struct sockaddr *);
+static struct sockaddr_in6 *sin6_cast(const struct sockaddr *);
+static const struct sockaddr *sa_cast(const struct sockaddr_storage *);
+static bool parse_sockaddr_components(struct sockaddr_storage *ss,
+ char *host_s,
+ const char *port_s,
+ uint16_t default_port,
+ const char *s,
+ bool resolve_host);
/* Sets 'fd' to non-blocking mode. Returns 0 if successful, otherwise a
* positive errno value. */
}
}
+void
+setsockopt_tcp_nodelay(int fd)
+{
+ int on = 1;
+ int retval;
+
+ retval = setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &on, sizeof on);
+ if (retval) {
+ retval = sock_errno();
+ VLOG_ERR("setsockopt(TCP_NODELAY): %s", sock_strerror(retval));
+ }
+}
+
+/* Sets the DSCP value of socket 'fd' to 'dscp', which must be 63 or less.
+ * 'family' must indicate the socket's address family (AF_INET or AF_INET6, to
+ * do anything useful). */
int
-set_dscp(int fd, uint8_t dscp)
+set_dscp(int fd, int family, uint8_t dscp)
{
+ int retval;
int val;
+#ifdef _WIN32
+ /* XXX: Consider using QoS2 APIs for Windows to set dscp. */
+ return 0;
+#endif
+
if (dscp > 63) {
return EINVAL;
}
-
val = dscp << 2;
- if (setsockopt(fd, IPPROTO_IP, IP_TOS, &val, sizeof val)) {
- return sock_errno();
- }
- return 0;
-}
-
-static bool
-rlim_is_finite(rlim_t limit)
-{
- if (limit == RLIM_INFINITY) {
- return false;
- }
+ switch (family) {
+ case AF_INET:
+ retval = setsockopt(fd, IPPROTO_IP, IP_TOS, &val, sizeof val);
+ break;
-#ifdef RLIM_SAVED_CUR /* FreeBSD 8.0 lacks RLIM_SAVED_CUR. */
- if (limit == RLIM_SAVED_CUR) {
- return false;
- }
-#endif
+ case AF_INET6:
+ retval = setsockopt(fd, IPPROTO_IPV6, IPV6_TCLASS, &val, sizeof val);
+ break;
-#ifdef RLIM_SAVED_MAX /* FreeBSD 8.0 lacks RLIM_SAVED_MAX. */
- if (limit == RLIM_SAVED_MAX) {
- return false;
+ default:
+ return ENOPROTOOPT;
}
-#endif
- return true;
+ return retval ? sock_errno() : 0;
}
-/* Returns the maximum valid FD value, plus 1. */
-int
-get_max_fds(void)
+/* Checks whether 'host_name' is an IPv4 or IPv6 address. It is assumed
+ * that 'host_name' is valid. Returns false if it is IPv4 address, true if
+ * it is IPv6 address. */
+bool
+addr_is_ipv6(const char *host_name)
{
- static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
- static int max_fds;
-
- if (ovsthread_once_start(&once)) {
- struct rlimit r;
- if (!getrlimit(RLIMIT_NOFILE, &r) && rlim_is_finite(r.rlim_cur)) {
- max_fds = r.rlim_cur;
- } else {
- VLOG_WARN("failed to obtain fd limit, defaulting to 1024");
- max_fds = 1024;
- }
- ovsthread_once_done(&once);
- }
-
- return max_fds;
+ return strchr(host_name, ':') != NULL;
}
/* Translates 'host_name', which must be a string representation of an IP
int
lookup_ip(const char *host_name, struct in_addr *addr)
{
- if (!inet_aton(host_name, addr)) {
+ if (!ip_parse(host_name, &addr->s_addr)) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
VLOG_ERR_RL(&rl, "\"%s\" is not a valid IP address", host_name);
return ENOENT;
int
lookup_ipv6(const char *host_name, struct in6_addr *addr)
{
- if (inet_pton(AF_INET6, host_name, addr) != 1) {
+ if (!ipv6_parse(host_name, addr)) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
VLOG_ERR_RL(&rl, "\"%s\" is not a valid IPv6 address", host_name);
return ENOENT;
struct addrinfo *result;
struct addrinfo hints;
- if (inet_aton(host_name, addr)) {
+ if (ip_parse(host_name, &addr->s_addr)) {
return 0;
}
pfd.fd = fd;
pfd.events = POLLOUT;
+
+#ifndef _WIN32
do {
retval = poll(&pfd, 1, 0);
} while (retval < 0 && errno == EINTR);
+#else
+ fd_set wrset, exset;
+ FD_ZERO(&wrset);
+ FD_ZERO(&exset);
+ FD_SET(fd, &exset);
+ FD_SET(fd, &wrset);
+ pfd.revents = 0;
+ struct timeval tv = { 0, 0 };
+ /* WSAPoll is broken on Windows, instead do a select */
+ retval = select(0, NULL, &wrset, &exset, &tv);
if (retval == 1) {
- if (pfd.revents & POLLERR) {
- ssize_t n = send(fd, "", 1, MSG_DONTWAIT);
+ if (FD_ISSET(fd, &wrset)) {
+ pfd.revents |= pfd.events;
+ }
+ if (FD_ISSET(fd, &exset)) {
+ pfd.revents |= POLLERR;
+ }
+ }
+#endif
+ if (retval == 1) {
+ if (pfd.revents & (POLLERR | POLLHUP)) {
+ ssize_t n = send(fd, "", 1, 0);
if (n < 0) {
- return errno;
+ return sock_errno();
} else {
VLOG_ERR_RL(&rl, "poll return POLLERR but send succeeded");
return EPROTO;
}
return 0;
} else if (retval < 0) {
- VLOG_ERR_RL(&rl, "poll: %s", ovs_strerror(errno));
+ VLOG_ERR_RL(&rl, "poll: %s", sock_strerror(sock_errno()));
return errno;
} else {
return EAGAIN;
}
}
-/* Drain all the data currently in the receive queue of a datagram socket (and
- * possibly additional data). There is no way to know how many packets are in
- * the receive queue, but we do know that the total number of bytes queued does
- * not exceed the receive buffer size, so we pull packets until none are left
- * or we've read that many bytes. */
-int
-drain_rcvbuf(int fd)
-{
- int rcvbuf;
-
- rcvbuf = get_socket_rcvbuf(fd);
- if (rcvbuf < 0) {
- return -rcvbuf;
- }
-
- while (rcvbuf > 0) {
- /* In Linux, specifying MSG_TRUNC in the flags argument causes the
- * datagram length to be returned, even if that is longer than the
- * buffer provided. Thus, we can use a 1-byte buffer to discard the
- * incoming datagram and still be able to account how many bytes were
- * removed from the receive buffer.
- *
- * On other Unix-like OSes, MSG_TRUNC has no effect in the flags
- * argument. */
- char buffer[LINUX_DATAPATH ? 1 : 2048];
- ssize_t n_bytes = recv(fd, buffer, sizeof buffer,
- MSG_TRUNC | MSG_DONTWAIT);
- if (n_bytes <= 0 || n_bytes >= rcvbuf) {
- break;
- }
- rcvbuf -= n_bytes;
- }
- return 0;
-}
-
/* Returns the size of socket 'sock''s receive buffer (SO_RCVBUF), or a
* negative errno value if an error occurs. */
int
}
}
-/* Attempts to shorten 'name' by opening a file descriptor for the directory
- * part of the name and indirecting through /proc/self/fd/<dirfd>/<basename>.
- * On systems with Linux-like /proc, this works as long as <basename> isn't too
- * long.
- *
- * On success, returns 0 and stores the short name in 'short_name' and a
- * directory file descriptor to eventually be closed in '*dirfpd'. */
-static int
-shorten_name_via_proc(const char *name, char short_name[MAX_UN_LEN + 1],
- int *dirfdp)
-{
- char *dir, *base;
- int dirfd;
- int len;
-
- if (!LINUX_DATAPATH) {
- return ENAMETOOLONG;
- }
-
- dir = dir_name(name);
- dirfd = open(dir, O_DIRECTORY | O_RDONLY);
- if (dirfd < 0) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
- int error = errno;
-
- VLOG_WARN_RL(&rl, "%s: open failed (%s)", dir, ovs_strerror(error));
- free(dir);
-
- return error;
- }
- free(dir);
-
- base = base_name(name);
- len = snprintf(short_name, MAX_UN_LEN + 1,
- "/proc/self/fd/%d/%s", dirfd, base);
- free(base);
-
- if (len >= 0 && len <= MAX_UN_LEN) {
- *dirfdp = dirfd;
- return 0;
- } else {
- close(dirfd);
- return ENAMETOOLONG;
- }
-}
-
-/* Attempts to shorten 'name' by creating a symlink for the directory part of
- * the name and indirecting through <symlink>/<basename>. This works on
- * systems that support symlinks, as long as <basename> isn't too long.
- *
- * On success, returns 0 and stores the short name in 'short_name' and the
- * symbolic link to eventually delete in 'linkname'. */
-static int
-shorten_name_via_symlink(const char *name, char short_name[MAX_UN_LEN + 1],
- char linkname[MAX_UN_LEN + 1])
+ovs_be32
+guess_netmask(ovs_be32 ip_)
{
- char *abs, *dir, *base;
- const char *tmpdir;
- int error;
- int i;
-
- abs = abs_file_name(NULL, name);
- dir = dir_name(abs);
- base = base_name(abs);
- free(abs);
-
- tmpdir = getenv("TMPDIR");
- if (tmpdir == NULL) {
- tmpdir = "/tmp";
- }
-
- for (i = 0; i < 1000; i++) {
- int len;
-
- len = snprintf(linkname, MAX_UN_LEN + 1,
- "%s/ovs-un-c-%"PRIu32, tmpdir, random_uint32());
- error = (len < 0 || len > MAX_UN_LEN ? ENAMETOOLONG
- : symlink(dir, linkname) ? errno
- : 0);
- if (error != EEXIST) {
- break;
- }
- }
-
- if (!error) {
- int len;
-
- fatal_signal_add_file_to_unlink(linkname);
-
- len = snprintf(short_name, MAX_UN_LEN + 1, "%s/%s", linkname, base);
- if (len < 0 || len > MAX_UN_LEN) {
- fatal_signal_unlink_file_now(linkname);
- error = ENAMETOOLONG;
- }
- }
-
- if (error) {
- linkname[0] = '\0';
- }
- free(dir);
- free(base);
-
- return error;
+ uint32_t ip = ntohl(ip_);
+ return ((ip >> 31) == 0 ? htonl(0xff000000) /* Class A */
+ : (ip >> 30) == 2 ? htonl(0xffff0000) /* Class B */
+ : (ip >> 29) == 6 ? htonl(0xffffff00) /* Class C */
+ : htonl(0)); /* ??? */
}
-/* Stores in '*un' a sockaddr_un that refers to file 'name'. Stores in
- * '*un_len' the size of the sockaddr_un.
- *
- * Returns 0 on success, otherwise a positive errno value.
- *
- * Uses '*dirfdp' and 'linkname' to store references to data when the caller no
- * longer needs to use 'un'. On success, freeing these references with
- * free_sockaddr_un() is mandatory to avoid a leak; on failure, freeing them is
- * unnecessary but harmless. */
-static int
-make_sockaddr_un(const char *name, struct sockaddr_un *un, socklen_t *un_len,
- int *dirfdp, char linkname[MAX_UN_LEN + 1])
+static char *
+unbracket(char *s)
{
- char short_name[MAX_UN_LEN + 1];
-
- *dirfdp = -1;
- linkname[0] = '\0';
- if (strlen(name) > MAX_UN_LEN) {
- /* 'name' is too long to fit in a sockaddr_un. Try a workaround. */
- int error = shorten_name_via_proc(name, short_name, dirfdp);
- if (error == ENAMETOOLONG) {
- error = shorten_name_via_symlink(name, short_name, linkname);
- }
- if (error) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
+ if (*s == '[') {
+ s++;
- VLOG_WARN_RL(&rl, "Unix socket name %s is longer than maximum "
- "%"PRIuSIZE" bytes", name, MAX_UN_LEN);
- return error;
+ char *end = strchr(s, '\0');
+ if (end[-1] == ']') {
+ end[-1] = '\0';
}
-
- name = short_name;
}
-
- un->sun_family = AF_UNIX;
- ovs_strzcpy(un->sun_path, name, sizeof un->sun_path);
- *un_len = (offsetof(struct sockaddr_un, sun_path)
- + strlen (un->sun_path) + 1);
- return 0;
+ return s;
}
-/* Clean up after make_sockaddr_un(). */
+/* 'host_index' is 0 if the host precedes the port within 's', 1 otherwise. */
static void
-free_sockaddr_un(int dirfd, const char *linkname)
+inet_parse_tokens__(char *s, int host_index, char **hostp, char **portp)
{
- if (dirfd >= 0) {
- close(dirfd);
- }
- if (linkname[0]) {
- fatal_signal_unlink_file_now(linkname);
- }
-}
-
-/* Binds Unix domain socket 'fd' to a file with permissions 0700. */
-static int
-bind_unix_socket(int fd, struct sockaddr *sun, socklen_t sun_len)
-{
- /* According to _Unix Network Programming_, umask should affect bind(). */
- mode_t old_umask = umask(0077);
- int error = bind(fd, sun, sun_len) ? errno : 0;
- umask(old_umask);
- return error;
-}
-
-/* Creates a Unix domain socket in the given 'style' (either SOCK_DGRAM or
- * SOCK_STREAM) that is bound to '*bind_path' (if 'bind_path' is non-null) and
- * connected to '*connect_path' (if 'connect_path' is non-null). If 'nonblock'
- * is true, the socket is made non-blocking.
- *
- * Returns the socket's fd if successful, otherwise a negative errno value. */
-int
-make_unix_socket(int style, bool nonblock,
- const char *bind_path, const char *connect_path)
-{
- int error;
- int fd;
-
- fd = socket(PF_UNIX, style, 0);
- if (fd < 0) {
- return -errno;
- }
-
- /* Set nonblocking mode right away, if we want it. This prevents blocking
- * in connect(), if connect_path != NULL. (In turn, that's a corner case:
- * it will only happen if style is SOCK_STREAM or SOCK_SEQPACKET, and only
- * if a backlog of un-accepted connections has built up in the kernel.) */
- if (nonblock) {
- error = set_nonblocking(fd);
- if (error) {
- goto error;
- }
- }
-
- if (bind_path) {
- char linkname[MAX_UN_LEN + 1];
- struct sockaddr_un un;
- socklen_t un_len;
- int dirfd;
-
- if (unlink(bind_path) && errno != ENOENT) {
- VLOG_WARN("unlinking \"%s\": %s\n",
- bind_path, ovs_strerror(errno));
- }
- fatal_signal_add_file_to_unlink(bind_path);
-
- error = make_sockaddr_un(bind_path, &un, &un_len, &dirfd, linkname);
- if (!error) {
- error = bind_unix_socket(fd, (struct sockaddr *) &un, un_len);
- }
- free_sockaddr_un(dirfd, linkname);
-
- if (error) {
- goto error;
+ char *colon = NULL;
+ bool in_brackets = false;
+ int n_colons = 0;
+ for (char *p = s; *p; p++) {
+ if (*p == '[') {
+ in_brackets = true;
+ } else if (*p == ']') {
+ in_brackets = false;
+ } else if (*p == ':' && !in_brackets) {
+ n_colons++;
+ colon = p;
}
}
- if (connect_path) {
- char linkname[MAX_UN_LEN + 1];
- struct sockaddr_un un;
- socklen_t un_len;
- int dirfd;
-
- error = make_sockaddr_un(connect_path, &un, &un_len, &dirfd, linkname);
- if (!error
- && connect(fd, (struct sockaddr*) &un, un_len)
- && errno != EINPROGRESS) {
- error = errno;
- }
- free_sockaddr_un(dirfd, linkname);
+ *hostp = *portp = NULL;
+ if (n_colons > 1) {
+ *hostp = s;
+ } else {
+ char **tokens[2];
+ tokens[host_index] = hostp;
+ tokens[!host_index] = portp;
- if (error) {
- goto error;
+ if (colon) {
+ *colon = '\0';
+ *tokens[1] = unbracket(colon + 1);
}
+ *tokens[0] = unbracket(s);
}
-
- return fd;
-
-error:
- if (error == EAGAIN) {
- error = EPROTO;
- }
- if (bind_path) {
- fatal_signal_unlink_file_now(bind_path);
- }
- close(fd);
- return -error;
}
-int
-get_unix_name_len(socklen_t sun_len)
+/* Parses 's', a string in the form "<host>[:<port>]", into its (required) host
+ * and (optional) port components, and stores pointers to them in '*hostp' and
+ * '*portp' respectively. Always sets '*hostp' nonnull, although possibly to
+ * an empty string. Can set '*portp' to the null string.
+ *
+ * Supports both IPv4 and IPv6. IPv6 addresses may be quoted with square
+ * brackets. Resolves ambiguous cases that might represent an IPv6 address or
+ * an IPv6 address and a port as representing just a host, e.g. "::1:2:3:4:80"
+ * is a host but "[::1:2:3:4]:80" is a host and a port.
+ *
+ * Modifies 's' and points '*hostp' and '*portp' (if nonnull) into it.
+ */
+void
+inet_parse_host_port_tokens(char *s, char **hostp, char **portp)
{
- return (sun_len >= offsetof(struct sockaddr_un, sun_path)
- ? sun_len - offsetof(struct sockaddr_un, sun_path)
- : 0);
+ inet_parse_tokens__(s, 0, hostp, portp);
}
-ovs_be32
-guess_netmask(ovs_be32 ip_)
+/* Parses 's', a string in the form "<port>[:<host>]", into its port and host
+ * components, and stores pointers to them in '*portp' and '*hostp'
+ * respectively. Either '*portp' and '*hostp' (but not both) can end up null.
+ *
+ * Supports both IPv4 and IPv6. IPv6 addresses may be quoted with square
+ * brackets. Resolves ambiguous cases that might represent an IPv6 address or
+ * an IPv6 address and a port as representing just a host, e.g. "::1:2:3:4:80"
+ * is a host but "[::1:2:3:4]:80" is a host and a port.
+ *
+ * Modifies 's' and points '*hostp' and '*portp' (if nonnull) into it.
+ */
+void
+inet_parse_port_host_tokens(char *s, char **portp, char **hostp)
{
- uint32_t ip = ntohl(ip_);
- return ((ip >> 31) == 0 ? htonl(0xff000000) /* Class A */
- : (ip >> 30) == 2 ? htonl(0xffff0000) /* Class B */
- : (ip >> 29) == 6 ? htonl(0xffffff00) /* Class C */
- : htonl(0)); /* ??? */
+ inet_parse_tokens__(s, 1, hostp, portp);
}
-/* This is like strsep() except:
- *
- * - The separator string is ":".
- *
- * - Square brackets [] quote ":" separators and are removed from the
- * tokens. */
-static char *
-parse_bracketed_token(char **pp)
+static bool
+parse_sockaddr_components_dns(struct sockaddr_storage *ss OVS_UNUSED,
+ char *host_s,
+ const char *port_s OVS_UNUSED,
+ uint16_t default_port OVS_UNUSED,
+ const char *s OVS_UNUSED)
{
- char *p = *pp;
-
- if (p == NULL) {
- return NULL;
- } else if (*p == '\0') {
- *pp = NULL;
- return p;
- } else if (*p == '[') {
- char *start = p + 1;
- char *end = start + strcspn(start, "]");
- *pp = (*end == '\0' ? NULL
- : end[1] == ':' ? end + 2
- : end + 1);
- *end = '\0';
- return start;
- } else {
- char *start = p;
- char *end = start + strcspn(start, ":");
- *pp = *end == '\0' ? NULL : end + 1;
- *end = '\0';
- return start;
+ char *tmp_host_s;
+
+ dns_resolve(host_s, &tmp_host_s);
+ if (tmp_host_s != NULL) {
+ parse_sockaddr_components(ss, tmp_host_s, port_s,
+ default_port, s, false);
+ free(tmp_host_s);
+ return true;
}
+ return false;
}
static bool
parse_sockaddr_components(struct sockaddr_storage *ss,
- const char *host_s,
+ char *host_s,
const char *port_s, uint16_t default_port,
- const char *s)
+ const char *s,
+ bool resolve_host)
{
- struct sockaddr_in *sin = ALIGNED_CAST(struct sockaddr_in *, ss);
+ struct sockaddr_in *sin = sin_cast(sa_cast(ss));
int port;
if (port_s && port_s[0]) {
if (!str_to_int(port_s, 10, &port) || port < 0 || port > 65535) {
VLOG_ERR("%s: bad port number \"%s\"", s, port_s);
+ goto exit;
}
} else {
port = default_port;
}
memset(ss, 0, sizeof *ss);
- if (strchr(host_s, ':')) {
- struct sockaddr_in6 *sin6
- = ALIGNED_CAST(struct sockaddr_in6 *, ss);
+ if (host_s && strchr(host_s, ':')) {
+ struct sockaddr_in6 *sin6 = sin6_cast(sa_cast(ss));
+ char *addr = strsep(&host_s, "%");
sin6->sin6_family = AF_INET6;
sin6->sin6_port = htons(port);
- if (!inet_pton(AF_INET6, host_s, sin6->sin6_addr.s6_addr)) {
- VLOG_ERR("%s: bad IPv6 address \"%s\"", s, host_s);
+ if (!addr || !*addr || !ipv6_parse(addr, &sin6->sin6_addr)) {
goto exit;
}
+
+#ifdef HAVE_STRUCT_SOCKADDR_IN6_SIN6_SCOPE_ID
+ char *scope = strsep(&host_s, "%");
+ if (scope && *scope) {
+ if (!scope[strspn(scope, "0123456789")]) {
+ sin6->sin6_scope_id = atoi(scope);
+ } else {
+ sin6->sin6_scope_id = if_nametoindex(scope);
+ if (!sin6->sin6_scope_id) {
+ VLOG_ERR("%s: bad IPv6 scope \"%s\" (%s)",
+ s, scope, ovs_strerror(errno));
+ goto exit;
+ }
+ }
+ }
+#endif
} else {
sin->sin_family = AF_INET;
sin->sin_port = htons(port);
- if (!inet_pton(AF_INET, host_s, &sin->sin_addr.s_addr)) {
- VLOG_ERR("%s: bad IPv4 address \"%s\"", s, host_s);
- goto exit;
+ if (host_s && !ip_parse(host_s, &sin->sin_addr.s_addr)) {
+ goto resolve;
}
}
return true;
+resolve:
+ if (resolve_host && parse_sockaddr_components_dns(ss, host_s, port_s,
+ default_port, s)) {
+ return true;
+ } else if (!resolve_host) {
+ VLOG_ERR("%s: bad IP address \"%s\"", s, host_s);
+ }
exit:
memset(ss, 0, sizeof *ss);
return false;
/* Parses 'target', which should be a string in the format "<host>[:<port>]".
* <host>, which is required, may be an IPv4 address or an IPv6 address
- * enclosed in square brackets. If 'default_port' is nonzero then <port> is
- * optional and defaults to 'default_port'.
+ * enclosed in square brackets. If 'default_port' is nonnegative then <port>
+ * is optional and defaults to 'default_port' (use 0 to make the kernel choose
+ * an available port, although this isn't usually appropriate for active
+ * connections). If 'default_port' is negative, then <port> is required.
+ * It resolves the host if 'resolve_host' is true.
*
* On success, returns true and stores the parsed remote address into '*ss'.
* On failure, logs an error, stores zeros into '*ss', and returns false. */
bool
-inet_parse_active(const char *target_, uint16_t default_port,
- struct sockaddr_storage *ss)
+inet_parse_active(const char *target_, int default_port,
+ struct sockaddr_storage *ss, bool resolve_host)
{
char *target = xstrdup(target_);
- const char *port;
- const char *host;
- char *p;
+ char *port, *host;
bool ok;
- p = target;
- host = parse_bracketed_token(&p);
- port = parse_bracketed_token(&p);
+ inet_parse_host_port_tokens(target, &host, &port);
if (!host) {
VLOG_ERR("%s: host must be specified", target_);
ok = false;
- } else if (!port && !default_port) {
+ } else if (!port && default_port < 0) {
VLOG_ERR("%s: port must be specified", target_);
ok = false;
} else {
- ok = parse_sockaddr_components(ss, host, port, default_port, target_);
+ ok = parse_sockaddr_components(ss, host, port, default_port,
+ target_, resolve_host);
}
if (!ok) {
memset(ss, 0, sizeof *ss);
/* Opens a non-blocking IPv4 or IPv6 socket of the specified 'style' and
* connects to 'target', which should be a string in the format
* "<host>[:<port>]". <host>, which is required, may be an IPv4 address or an
- * IPv6 address enclosed in square brackets. If 'default_port' is nonzero then
- * <port> is optional and defaults to 'default_port'.
+ * IPv6 address enclosed in square brackets. If 'default_port' is nonnegative
+ * then <port> is optional and defaults to 'default_port'.
*
* 'style' should be SOCK_STREAM (for TCP) or SOCK_DGRAM (for UDP).
*
* should be in the range [0, 63] and will automatically be shifted to the
* appropriately place in the IP tos field. */
int
-inet_open_active(int style, const char *target, uint16_t default_port,
+inet_open_active(int style, const char *target, int default_port,
struct sockaddr_storage *ssp, int *fdp, uint8_t dscp)
{
struct sockaddr_storage ss;
int error;
/* Parse. */
- if (!inet_parse_active(target, default_port, &ss)) {
+ if (!inet_parse_active(target, default_port, &ss, true)) {
error = EAFNOSUPPORT;
goto exit;
}
/* The dscp bits must be configured before connect() to ensure that the
* TOS field is set during the connection establishment. If set after
* connect(), the handshake SYN frames will be sent with a TOS of 0. */
- error = set_dscp(fd, dscp);
+ error = set_dscp(fd, ss.ss_family, dscp);
if (error) {
- VLOG_ERR("%s: socket: %s", target, sock_strerror(error));
+ VLOG_ERR("%s: set_dscp: %s", target, sock_strerror(error));
goto exit;
}
struct sockaddr_storage *ss)
{
char *target = xstrdup(target_);
- const char *port;
- const char *host;
- char *p;
+ char *port, *host;
bool ok;
- p = target;
- port = parse_bracketed_token(&p);
- host = parse_bracketed_token(&p);
+ inet_parse_port_host_tokens(target, &port, &host);
if (!port && default_port < 0) {
VLOG_ERR("%s: port must be specified", target_);
ok = false;
} else {
- ok = parse_sockaddr_components(ss, host ? host : "0.0.0.0",
- port, default_port, target_);
+ ok = parse_sockaddr_components(ss, host, port, default_port,
+ target_, true);
}
if (!ok) {
memset(ss, 0, sizeof *ss);
*
* 'dscp' becomes the DSCP bits in the IP headers for the new connection. It
* should be in the range [0, 63] and will automatically be shifted to the
- * appropriately place in the IP tos field. */
+ * appropriately place in the IP tos field.
+ *
+ * If 'kernel_print_port' is true and the port is dynamically assigned by
+ * the kernel, print the chosen port. */
int
inet_open_passive(int style, const char *target, int default_port,
- struct sockaddr_storage *ssp, uint8_t dscp)
+ struct sockaddr_storage *ssp, uint8_t dscp,
+ bool kernel_print_port)
{
bool kernel_chooses_port;
struct sockaddr_storage ss;
/* Bind. */
if (bind(fd, (struct sockaddr *) &ss, ss_length(&ss)) < 0) {
error = sock_errno();
- VLOG_ERR("%s: bind: %s", target, sock_strerror(error));
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
+ VLOG_ERR_RL(&rl, "%s: bind: %s", target, sock_strerror(error));
goto error;
}
/* The dscp bits must be configured before connect() to ensure that the TOS
* field is set during the connection establishment. If set after
* connect(), the handshake SYN frames will be sent with a TOS of 0. */
- error = set_dscp(fd, dscp);
+ error = set_dscp(fd, ss.ss_family, dscp);
if (error) {
- VLOG_ERR("%s: socket: %s", target, sock_strerror(error));
+ VLOG_ERR("%s: set_dscp: %s", target, sock_strerror(error));
goto error;
}
VLOG_ERR("%s: getsockname: %s", target, sock_strerror(error));
goto error;
}
- if (kernel_chooses_port) {
+ if (kernel_chooses_port && kernel_print_port) {
VLOG_INFO("%s: listening on port %"PRIu16,
target, ss_get_port(&ss));
}
return -error;
}
-/* Returns a readable and writable fd for /dev/null, if successful, otherwise
- * a negative errno value. The caller must not close the returned fd (because
- * the same fd will be handed out to subsequent callers). */
-int
-get_null_fd(void)
+/* Parses 'target', which may be an IPv4 address or an IPv6 address
+ * enclosed in square brackets.
+ *
+ * On success, returns true and stores the parsed remote address into '*ss'.
+ * On failure, logs an error, stores zeros into '*ss', and returns false. */
+bool
+inet_parse_address(const char *target_, struct sockaddr_storage *ss)
{
- static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
- static int null_fd;
-
- if (ovsthread_once_start(&once)) {
- null_fd = open("/dev/null", O_RDWR);
- if (null_fd < 0) {
- int error = errno;
- VLOG_ERR("could not open /dev/null: %s", ovs_strerror(error));
- null_fd = -error;
- }
- ovsthread_once_done(&once);
+ char *target = xstrdup(target_);
+ char *host = unbracket(target);
+ bool ok = parse_sockaddr_components(ss, host, NULL, 0, target_, false);
+ if (!ok) {
+ memset(ss, 0, sizeof *ss);
}
-
- return null_fd;
+ free(target);
+ return ok;
}
int
}
}
-void
-xpipe(int fds[2])
-{
- if (pipe(fds)) {
- VLOG_FATAL("failed to create pipe (%s)", ovs_strerror(errno));
- }
-}
-
-void
-xpipe_nonblocking(int fds[2])
-{
- xpipe(fds);
- xset_nonblocking(fds[0]);
- xset_nonblocking(fds[1]);
-}
-
static int
getsockopt_int(int fd, int level, int option, const char *optname, int *valuep)
{
if (!getaddr(fd, (struct sockaddr *) &ss, &len)) {
if (ss.ss_family == AF_INET || ss.ss_family == AF_INET6) {
- char addrbuf[SS_NTOP_BUFSIZE];
-
- ds_put_format(string, "%s:%"PRIu16,
- ss_format_address(&ss, addrbuf, sizeof addrbuf),
- ss_get_port(&ss));
+ ss_format_address(&ss, string);
+ ds_put_format(string, ":%"PRIu16, ss_get_port(&ss));
+#ifndef _WIN32
} else if (ss.ss_family == AF_UNIX) {
struct sockaddr_un sun;
const char *null;
null = memchr(sun.sun_path, '\0', maxlen);
ds_put_buffer(string, sun.sun_path,
null ? null - sun.sun_path : maxlen);
+#endif
}
#ifdef HAVE_NETLINK
else if (ss.ss_family == AF_NETLINK) {
}
}
#endif
-#if AF_PACKET && LINUX_DATAPATH
+#if __linux__
else if (ss.ss_family == AF_PACKET) {
struct sockaddr_ll sll;
}
-#ifdef LINUX_DATAPATH
+#ifdef __linux__
static void
put_fd_filename(struct ds *string, int fd)
{
: S_ISFIFO(s.st_mode) ? "FIFO"
: S_ISLNK(s.st_mode) ? "symbolic link"
: "unknown"));
-#ifdef LINUX_DATAPATH
+#ifdef __linux__
put_fd_filename(&string, fd);
#endif
}
#endif /* _WIN32 */
return ds_steal_cstr(&string);
}
+\f
+/* sockaddr helpers. */
-/* Calls ioctl() on an AF_INET sock, passing the specified 'command' and
- * 'arg'. Returns 0 if successful, otherwise a positive errno value. */
-int
-af_inet_ioctl(unsigned long int command, const void *arg)
+static struct sockaddr_in *
+sin_cast(const struct sockaddr *sa)
{
- static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
- static int sock;
-
- if (ovsthread_once_start(&once)) {
- sock = socket(AF_INET, SOCK_DGRAM, 0);
- if (sock < 0) {
- int error = sock_errno();
- VLOG_ERR("failed to create inet socket: %s", sock_strerror(error));
- sock = -error;
- }
- ovsthread_once_done(&once);
- }
+ return ALIGNED_CAST(struct sockaddr_in *, sa);
+}
- return (sock < 0 ? -sock
- : ioctl(sock, command, arg) == -1 ? errno
- : 0);
+static struct sockaddr_in6 *
+sin6_cast(const struct sockaddr *sa)
+{
+ return ALIGNED_CAST(struct sockaddr_in6 *, sa);
}
-int
-af_inet_ifreq_ioctl(const char *name, struct ifreq *ifr, unsigned long int cmd,
- const char *cmd_name)
+/* Returns true if 'sa' represents an IPv4 or IPv6 address, false otherwise. */
+bool
+sa_is_ip(const struct sockaddr *sa)
{
- int error;
+ return sa->sa_family == AF_INET || sa->sa_family == AF_INET6;
+}
- ovs_strzcpy(ifr->ifr_name, name, sizeof ifr->ifr_name);
- error = af_inet_ioctl(cmd, ifr);
- if (error) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
- VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s", name, cmd_name,
- ovs_strerror(error));
- }
- return error;
+/* Returns the IPv4 or IPv6 address in 'sa'. Returns IPv4 addresses as
+ * v6-mapped. */
+struct in6_addr
+sa_get_address(const struct sockaddr *sa)
+{
+ ovs_assert(sa_is_ip(sa));
+ return (sa->sa_family == AF_INET
+ ? in6_addr_mapped_ipv4(sin_cast(sa)->sin_addr.s_addr)
+ : sin6_cast(sa)->sin6_addr);
}
-\f
-/* sockaddr_storage helpers. */
-/* Returns the IPv4 or IPv6 port in 'ss'. */
+/* Returns the IPv4 or IPv6 port in 'sa'. */
uint16_t
-ss_get_port(const struct sockaddr_storage *ss)
+sa_get_port(const struct sockaddr *sa)
{
- if (ss->ss_family == AF_INET) {
- const struct sockaddr_in *sin
- = ALIGNED_CAST(const struct sockaddr_in *, ss);
- return ntohs(sin->sin_port);
- } else if (ss->ss_family == AF_INET6) {
- const struct sockaddr_in6 *sin6
- = ALIGNED_CAST(const struct sockaddr_in6 *, ss);
- return ntohs(sin6->sin6_port);
- } else {
- OVS_NOT_REACHED();
+ ovs_assert(sa_is_ip(sa));
+ return ntohs(sa->sa_family == AF_INET
+ ? sin_cast(sa)->sin_port
+ : sin6_cast(sa)->sin6_port);
+}
+
+/* Returns true if 'name' is safe to include inside a network address field.
+ * We want to avoid names that include confusing punctuation, etc. */
+static bool OVS_UNUSED
+is_safe_name(const char *name)
+{
+ if (!name[0] || isdigit((unsigned char) name[0])) {
+ return false;
+ }
+ for (const char *p = name; *p; p++) {
+ if (!isalnum((unsigned char) *p) && *p != '-' && *p != '_') {
+ return false;
+ }
}
+ return true;
}
-/* Formats the IPv4 or IPv6 address in 'ss' into the 'bufsize' bytes in 'buf'.
- * If 'ss' is an IPv6 address, puts square brackets around the address.
- * 'bufsize' should be at least SS_NTOP_BUFSIZE.
- *
- * Returns 'buf'. */
-char *
-ss_format_address(const struct sockaddr_storage *ss,
- char *buf, size_t bufsize)
+static void
+sa_format_address__(const struct sockaddr *sa,
+ const char *lbrack, const char *rbrack,
+ struct ds *s)
{
- ovs_assert(bufsize >= SS_NTOP_BUFSIZE);
- if (ss->ss_family == AF_INET) {
- const struct sockaddr_in *sin
- = ALIGNED_CAST(const struct sockaddr_in *, ss);
-
- snprintf(buf, bufsize, IP_FMT, IP_ARGS(sin->sin_addr.s_addr));
- } else if (ss->ss_family == AF_INET6) {
- const struct sockaddr_in6 *sin6
- = ALIGNED_CAST(const struct sockaddr_in6 *, ss);
-
- buf[0] = '[';
- inet_ntop(AF_INET6, sin6->sin6_addr.s6_addr, buf + 1, bufsize - 1);
- strcpy(strchr(buf, '\0'), "]");
+ ovs_assert(sa_is_ip(sa));
+ if (sa->sa_family == AF_INET) {
+ ds_put_format(s, IP_FMT, IP_ARGS(sin_cast(sa)->sin_addr.s_addr));
} else {
- OVS_NOT_REACHED();
+ const struct sockaddr_in6 *sin6 = sin6_cast(sa);
+
+ ds_put_cstr(s, lbrack);
+ ds_reserve(s, s->length + INET6_ADDRSTRLEN);
+ char *tail = &s->string[s->length];
+ inet_ntop(AF_INET6, sin6->sin6_addr.s6_addr, tail, INET6_ADDRSTRLEN);
+ s->length += strlen(tail);
+
+#ifdef HAVE_STRUCT_SOCKADDR_IN6_SIN6_SCOPE_ID
+ uint32_t scope = sin6->sin6_scope_id;
+ if (scope) {
+ char namebuf[IF_NAMESIZE];
+ char *name = if_indextoname(scope, namebuf);
+ ds_put_char(s, '%');
+ if (name && is_safe_name(name)) {
+ ds_put_cstr(s, name);
+ } else {
+ ds_put_format(s, "%"PRIu32, scope);
+ }
+ }
+#endif
+
+ ds_put_cstr(s, rbrack);
}
+}
- return buf;
+/* Formats the IPv4 or IPv6 address in 'sa' into 's'. If 'sa' is an IPv6
+ * address, puts square brackets around the address. */
+void
+sa_format_address(const struct sockaddr *sa, struct ds *s)
+{
+ sa_format_address__(sa, "[", "]", s);
+}
+
+/* Formats the IPv4 or IPv6 address in 'sa' into 's'. Does not add square
+ * brackets around IPv6 addresses. */
+void
+sa_format_address_nobracks(const struct sockaddr *sa, struct ds *s)
+{
+ sa_format_address__(sa, "", "", s);
}
size_t
-ss_length(const struct sockaddr_storage *ss)
+sa_length(const struct sockaddr *sa)
{
- switch (ss->ss_family) {
+ switch (sa->sa_family) {
case AF_INET:
return sizeof(struct sockaddr_in);
OVS_NOT_REACHED();
}
}
+\f
+/* sockaddr_storage helpers. */
+
+static const struct sockaddr *
+sa_cast(const struct sockaddr_storage *ss)
+{
+ return ALIGNED_CAST(const struct sockaddr *, ss);
+}
+bool
+ss_is_ip(const struct sockaddr_storage *ss)
+{
+ return sa_is_ip(sa_cast(ss));
+}
+
+uint16_t
+ss_get_port(const struct sockaddr_storage *ss)
+{
+ return sa_get_port(sa_cast(ss));
+}
+
+struct in6_addr
+ss_get_address(const struct sockaddr_storage *ss)
+{
+ return sa_get_address(sa_cast(ss));
+}
+
+void
+ss_format_address(const struct sockaddr_storage *ss, struct ds *s)
+{
+ sa_format_address(sa_cast(ss), s);
+}
+
+void
+ss_format_address_nobracks(const struct sockaddr_storage *ss, struct ds *s)
+{
+ sa_format_address_nobracks(sa_cast(ss), s);
+}
+
+size_t
+ss_length(const struct sockaddr_storage *ss)
+{
+ return sa_length(sa_cast(ss));
+}
+\f
/* For Windows socket calls, 'errno' is not set. One has to call
* WSAGetLastError() to get the error number and then pass it to
* this function to get the correct error string.
return ovs_strerror(error);
#endif
}
+\f
+#ifdef __linux__
+static int
+emulate_sendmmsg(int fd, struct mmsghdr *msgs, unsigned int n,
+ unsigned int flags)
+{
+ for (unsigned int i = 0; i < n; i++) {
+ ssize_t retval = sendmsg(fd, &msgs[i].msg_hdr, flags);
+ if (retval < 0) {
+ return i ? i : retval;
+ }
+ msgs[i].msg_len = retval;
+ }
+ return n;
+}
+
+#ifndef HAVE_SENDMMSG
+int
+sendmmsg(int fd, struct mmsghdr *msgs, unsigned int n, unsigned int flags)
+{
+ return emulate_sendmmsg(fd, msgs, n, flags);
+}
+#else
+/* sendmmsg was redefined in lib/socket-util.c, should undef sendmmsg here
+ * to avoid recursion */
+#undef sendmmsg
+int
+wrap_sendmmsg(int fd, struct mmsghdr *msgs, unsigned int n, unsigned int flags)
+{
+ static bool sendmmsg_broken = false;
+ if (!sendmmsg_broken) {
+ int save_errno = errno;
+ int retval = sendmmsg(fd, msgs, n, flags);
+ if (retval >= 0 || errno != ENOSYS) {
+ return retval;
+ }
+ sendmmsg_broken = true;
+ errno = save_errno;
+ }
+ return emulate_sendmmsg(fd, msgs, n, flags);
+}
+#endif
+
+static int
+emulate_recvmmsg(int fd, struct mmsghdr *msgs, unsigned int n,
+ int flags, struct timespec *timeout OVS_UNUSED)
+{
+ ovs_assert(!timeout); /* XXX not emulated */
+
+ bool waitforone = flags & MSG_WAITFORONE;
+ flags &= ~MSG_WAITFORONE;
+
+ for (unsigned int i = 0; i < n; i++) {
+ ssize_t retval = recvmsg(fd, &msgs[i].msg_hdr, flags);
+ if (retval < 0) {
+ return i ? i : retval;
+ }
+ msgs[i].msg_len = retval;
+
+ if (waitforone) {
+ flags |= MSG_DONTWAIT;
+ }
+ }
+ return n;
+}
+
+#ifndef HAVE_SENDMMSG
+int
+recvmmsg(int fd, struct mmsghdr *msgs, unsigned int n,
+ int flags, struct timespec *timeout)
+{
+ return emulate_recvmmsg(fd, msgs, n, flags, timeout);
+}
+#else
+/* recvmmsg was redefined in lib/socket-util.c, should undef recvmmsg here
+ * to avoid recursion */
+#undef recvmmsg
+int
+wrap_recvmmsg(int fd, struct mmsghdr *msgs, unsigned int n,
+ int flags, struct timespec *timeout)
+{
+ ovs_assert(!timeout); /* XXX not emulated */
+
+ static bool recvmmsg_broken = false;
+ if (!recvmmsg_broken) {
+ int save_errno = errno;
+ int retval = recvmmsg(fd, msgs, n, flags, timeout);
+ if (retval >= 0 || errno != ENOSYS) {
+ return retval;
+ }
+ recvmmsg_broken = true;
+ errno = save_errno;
+ }
+ return emulate_recvmmsg(fd, msgs, n, flags, timeout);
+}
+#endif
+#endif /* __linux__ */