/*
- * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
+ * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 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 <stdio.h>
#include <stdlib.h>
#include <string.h>
+#include <sys/ioctl.h>
#include <sys/resource.h>
#include <sys/socket.h>
#include <sys/stat.h>
#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);
if (fcntl(fd, F_SETFL, flags | O_NONBLOCK) != -1) {
return 0;
} else {
- VLOG_ERR("fcntl(F_SETFL) failed: %s", strerror(errno));
+ VLOG_ERR("fcntl(F_SETFL) failed: %s", ovs_strerror(errno));
return errno;
}
} else {
- VLOG_ERR("fcntl(F_GETFL) failed: %s", strerror(errno));
+ VLOG_ERR("fcntl(F_GETFL) failed: %s", ovs_strerror(errno));
return errno;
}
}
int
get_max_fds(void)
{
- static int max_fds = -1;
- if (max_fds < 0) {
+ 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;
VLOG_WARN("failed to obtain fd limit, defaulting to 1024");
max_fds = 1024;
}
+ ovsthread_once_done(&once);
}
+
return max_fds;
}
switch (getaddrinfo(host_name, NULL, &hints, &result)) {
case 0:
- *addr = ((struct sockaddr_in *) result->ai_addr)->sin_addr;
+ *addr = ALIGNED_CAST(struct sockaddr_in *,
+ result->ai_addr)->sin_addr;
freeaddrinfo(result);
return 0;
}
return 0;
} else if (retval < 0) {
- VLOG_ERR_RL(&rl, "poll: %s", strerror(errno));
+ VLOG_ERR_RL(&rl, "poll: %s", ovs_strerror(errno));
return errno;
} else {
return EAGAIN;
}
}
-/* Stores in '*un' a sockaddr_un that refers to file 'name'. Stores in
- * '*un_len' the size of the sockaddr_un. */
-static void
-make_sockaddr_un__(const char *name, struct sockaddr_un *un, socklen_t *un_len)
+/* 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)
{
- 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);
+ 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])
+{
+ 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;
}
/* 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. On success,
- * '*dirfdp' is either -1 or a nonnegative file descriptor that the caller
- * should close after using '*un' to bind or connect. On failure, '*dirfdp' is
- * -1. */
+ * 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)
+ int *dirfdp, char linkname[MAX_UN_LEN + 1])
{
- enum { MAX_UN_LEN = sizeof un->sun_path - 1 };
+ char short_name[MAX_UN_LEN + 1];
*dirfdp = -1;
+ linkname[0] = '\0';
if (strlen(name) > MAX_UN_LEN) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
-
- if (LINUX_DATAPATH) {
- /* 'name' is too long to fit in a sockaddr_un, but we have a
- * workaround for that on Linux: shorten it by opening a file
- * descriptor for the directory part of the name and indirecting
- * through /proc/self/fd/<dirfd>/<basename>. */
- char *dir, *base;
- char *short_name;
- int dirfd;
-
- dir = dir_name(name);
- base = base_name(name);
-
- dirfd = open(dir, O_DIRECTORY | O_RDONLY);
- if (dirfd < 0) {
- free(base);
- free(dir);
- return errno;
- }
-
- short_name = xasprintf("/proc/self/fd/%d/%s", dirfd, base);
- free(dir);
- free(base);
-
- if (strlen(short_name) <= MAX_UN_LEN) {
- make_sockaddr_un__(short_name, un, un_len);
- free(short_name);
- *dirfdp = dirfd;
- return 0;
- }
- free(short_name);
- close(dirfd);
+ /* '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);
VLOG_WARN_RL(&rl, "Unix socket name %s is longer than maximum "
- "%d bytes (even shortened)", name, MAX_UN_LEN);
- } else {
- /* 'name' is too long and we have no workaround. */
- VLOG_WARN_RL(&rl, "Unix socket name %s is longer than maximum "
- "%d bytes", name, MAX_UN_LEN);
+ "%"PRIuSIZE" bytes", name, MAX_UN_LEN);
+ return error;
}
- return ENAMETOOLONG;
- } else {
- make_sockaddr_un__(name, un, un_len);
- return 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;
+}
+
+/* Clean up after make_sockaddr_un(). */
+static void
+free_sockaddr_un(int dirfd, const char *linkname)
+{
+ if (dirfd >= 0) {
+ close(dirfd);
+ }
+ if (linkname[0]) {
+ fatal_signal_unlink_file_now(linkname);
}
}
}
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, strerror(errno));
+ 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);
+ error = make_sockaddr_un(bind_path, &un, &un_len, &dirfd, linkname);
if (!error) {
error = bind_unix_socket(fd, (struct sockaddr *) &un, un_len);
}
- if (dirfd >= 0) {
- close(dirfd);
- }
+ free_sockaddr_un(dirfd, linkname);
+
if (error) {
goto error;
}
}
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);
+ error = make_sockaddr_un(connect_path, &un, &un_len, &dirfd, linkname);
if (!error
&& connect(fd, (struct sockaddr*) &un, un_len)
&& errno != EINPROGRESS) {
error = errno;
}
- if (dirfd >= 0) {
- close(dirfd);
- }
+ free_sockaddr_un(dirfd, linkname);
+
if (error) {
goto error;
}
: htonl(0)); /* ??? */
}
-/* Parses 'target', which should be a string in the format "<host>[:<port>]".
- * <host> is required. If 'default_port' is nonzero then <port> is optional
- * and defaults to 'default_port'.
+/* This is like strsep() except:
*
- * On success, returns true and stores the parsed remote address into '*sinp'.
- * On failure, logs an error, stores zeros into '*sinp', and returns false. */
-bool
-inet_parse_active(const char *target_, uint16_t default_port,
- struct sockaddr_in *sinp)
+ * - The separator string is ":".
+ *
+ * - Square brackets [] quote ":" separators and are removed from the
+ * tokens. */
+static char *
+parse_bracketed_token(char **pp)
{
- char *target = xstrdup(target_);
- char *save_ptr = NULL;
- const char *host_name;
- const char *port_string;
- bool ok = false;
-
- /* Defaults. */
- sinp->sin_family = AF_INET;
- sinp->sin_port = htons(default_port);
-
- /* Tokenize. */
- host_name = strtok_r(target, ":", &save_ptr);
- port_string = strtok_r(NULL, ":", &save_ptr);
- if (!host_name) {
- VLOG_ERR("%s: bad peer name format", target_);
- goto exit;
+ 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;
}
+}
- /* Look up IP, port. */
- if (lookup_ip(host_name, &sinp->sin_addr)) {
- goto exit;
+static bool
+parse_sockaddr_components(struct sockaddr_storage *ss,
+ const char *host_s,
+ const char *port_s, uint16_t default_port,
+ const char *s)
+{
+ struct sockaddr_in *sin = ALIGNED_CAST(struct sockaddr_in *, 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);
+ }
+ } else {
+ port = default_port;
}
- if (port_string && atoi(port_string)) {
- sinp->sin_port = htons(atoi(port_string));
- } else if (!default_port) {
- VLOG_ERR("%s: port number must be specified", target_);
- goto exit;
+
+ memset(ss, 0, sizeof *ss);
+ if (strchr(host_s, ':')) {
+ struct sockaddr_in6 *sin6
+ = ALIGNED_CAST(struct sockaddr_in6 *, ss);
+
+ 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);
+ goto exit;
+ }
+ } 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;
+ }
}
- ok = true;
+ return true;
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'.
+ *
+ * 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)
+{
+ char *target = xstrdup(target_);
+ const char *port;
+ const char *host;
+ char *p;
+ bool ok;
+
+ p = target;
+ host = parse_bracketed_token(&p);
+ port = parse_bracketed_token(&p);
+ if (!host) {
+ VLOG_ERR("%s: host must be specified", target_);
+ ok = false;
+ } else if (!port && !default_port) {
+ VLOG_ERR("%s: port must be specified", target_);
+ ok = false;
+ } else {
+ ok = parse_sockaddr_components(ss, host, port, default_port, target_);
+ }
if (!ok) {
- memset(sinp, 0, sizeof *sinp);
+ memset(ss, 0, sizeof *ss);
}
free(target);
return ok;
}
-/* Opens a non-blocking IPv4 socket of the specified 'style' and connects to
- * 'target', which should be a string in the format "<host>[:<port>]". <host>
- * is required. If 'default_port' is nonzero then <port> is optional and
- * defaults to 'default_port'.
+
+/* 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'.
*
* 'style' should be SOCK_STREAM (for TCP) or SOCK_DGRAM (for UDP).
*
* into '*fdp'. On failure, returns a positive errno value other than EAGAIN
* and stores -1 into '*fdp'.
*
- * If 'sinp' is non-null, then on success the target address is stored into
- * '*sinp'.
+ * If 'ss' is non-null, then on success stores the target address into '*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. */
int
inet_open_active(int style, const char *target, uint16_t default_port,
- struct sockaddr_in *sinp, int *fdp, uint8_t dscp)
+ struct sockaddr_storage *ssp, int *fdp, uint8_t dscp)
{
- struct sockaddr_in sin;
+ struct sockaddr_storage ss;
int fd = -1;
int error;
/* Parse. */
- if (!inet_parse_active(target, default_port, &sin)) {
+ if (!inet_parse_active(target, default_port, &ss)) {
error = EAFNOSUPPORT;
goto exit;
}
/* Create non-blocking socket. */
- fd = socket(AF_INET, style, 0);
+ fd = socket(ss.ss_family, style, 0);
if (fd < 0) {
- VLOG_ERR("%s: socket: %s", target, strerror(errno));
+ VLOG_ERR("%s: socket: %s", target, ovs_strerror(errno));
error = errno;
goto exit;
}
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
+ /* 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);
if (error) {
- VLOG_ERR("%s: socket: %s", target, strerror(error));
+ VLOG_ERR("%s: socket: %s", target, ovs_strerror(error));
goto exit;
}
/* Connect. */
- error = connect(fd, (struct sockaddr *) &sin, sizeof sin) == 0 ? 0 : errno;
+ error = connect(fd, (struct sockaddr *) &ss, ss_length(&ss)) == 0
+ ? 0
+ : errno;
if (error == EINPROGRESS) {
error = EAGAIN;
}
exit:
- if (!error || error == EAGAIN) {
- if (sinp) {
- *sinp = sin;
+ if (error && error != EAGAIN) {
+ if (ssp) {
+ memset(ssp, 0, sizeof *ssp);
+ }
+ if (fd >= 0) {
+ close(fd);
+ fd = -1;
+ }
+ } else {
+ if (ssp) {
+ *ssp = ss;
}
- } else if (fd >= 0) {
- close(fd);
- fd = -1;
}
*fdp = fd;
return error;
}
-/* Parses 'target', which should be a string in the format "[<port>][:<ip>]":
+/* Parses 'target', which should be a string in the format "[<port>][:<host>]":
*
* - If 'default_port' is -1, then <port> is required. Otherwise, if
* <port> is omitted, then 'default_port' is used instead.
* - If <port> (or 'default_port', if used) is 0, then no port is bound
* and the TCP/IP stack will select a port.
*
- * - If <ip> is omitted then the IP address is wildcarded.
+ * - <host> is optional. If supplied, it may be an IPv4 address or an
+ * IPv6 address enclosed in square brackets. If omitted, the IP address
+ * is wildcarded.
*
- * If successful, stores the address into '*sinp' and returns true; otherwise
- * zeros '*sinp' and returns false. */
+ * If successful, stores the address into '*ss' and returns true; otherwise
+ * zeros '*ss' and returns false. */
bool
inet_parse_passive(const char *target_, int default_port,
- struct sockaddr_in *sinp)
+ struct sockaddr_storage *ss)
{
char *target = xstrdup(target_);
- char *string_ptr = target;
- const char *host_name;
- const char *port_string;
- bool ok = false;
- int port;
-
- /* Address defaults. */
- memset(sinp, 0, sizeof *sinp);
- sinp->sin_family = AF_INET;
- sinp->sin_addr.s_addr = htonl(INADDR_ANY);
- sinp->sin_port = htons(default_port);
-
- /* Parse optional port number. */
- port_string = strsep(&string_ptr, ":");
- if (port_string && str_to_int(port_string, 10, &port)) {
- sinp->sin_port = htons(port);
- } else if (default_port < 0) {
- VLOG_ERR("%s: port number must be specified", target_);
- goto exit;
- }
-
- /* Parse optional bind IP. */
- host_name = strsep(&string_ptr, ":");
- if (host_name && host_name[0] && lookup_ip(host_name, &sinp->sin_addr)) {
- goto exit;
+ const char *port;
+ const char *host;
+ char *p;
+ bool ok;
+
+ p = target;
+ port = parse_bracketed_token(&p);
+ host = parse_bracketed_token(&p);
+ 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 = true;
-
-exit:
if (!ok) {
- memset(sinp, 0, sizeof *sinp);
+ memset(ss, 0, sizeof *ss);
}
free(target);
return ok;
}
-/* Opens a non-blocking IPv4 socket of the specified 'style', binds to
+/* Opens a non-blocking IPv4 or IPv6 socket of the specified 'style', binds to
* 'target', and listens for incoming connections. Parses 'target' in the same
* way was inet_parse_passive().
*
* On success, returns a non-negative file descriptor. On failure, returns a
* negative errno value.
*
- * If 'sinp' is non-null, then on success the bound address is stored into
- * '*sinp'.
+ * If 'ss' is non-null, then on success stores the bound address into '*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. */
int
inet_open_passive(int style, const char *target, int default_port,
- struct sockaddr_in *sinp, uint8_t dscp)
+ struct sockaddr_storage *ssp, uint8_t dscp)
{
bool kernel_chooses_port;
- struct sockaddr_in sin;
+ struct sockaddr_storage ss;
int fd = 0, error;
unsigned int yes = 1;
- if (!inet_parse_passive(target, default_port, &sin)) {
+ if (!inet_parse_passive(target, default_port, &ss)) {
return -EAFNOSUPPORT;
}
+ kernel_chooses_port = ss_get_port(&ss) == 0;
/* Create non-blocking socket, set SO_REUSEADDR. */
- fd = socket(AF_INET, style, 0);
+ fd = socket(ss.ss_family, style, 0);
if (fd < 0) {
error = errno;
- VLOG_ERR("%s: socket: %s", target, strerror(error));
+ VLOG_ERR("%s: socket: %s", target, ovs_strerror(error));
return -error;
}
error = set_nonblocking(fd);
if (style == SOCK_STREAM
&& setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof yes) < 0) {
error = errno;
- VLOG_ERR("%s: setsockopt(SO_REUSEADDR): %s", target, strerror(error));
+ VLOG_ERR("%s: setsockopt(SO_REUSEADDR): %s",
+ target, ovs_strerror(error));
goto error;
}
/* Bind. */
- if (bind(fd, (struct sockaddr *) &sin, sizeof sin) < 0) {
+ if (bind(fd, (struct sockaddr *) &ss, ss_length(&ss)) < 0) {
error = errno;
- VLOG_ERR("%s: bind: %s", target, strerror(error));
+ VLOG_ERR("%s: bind: %s", target, ovs_strerror(error));
goto error;
}
* connect(), the handshake SYN frames will be sent with a TOS of 0. */
error = set_dscp(fd, dscp);
if (error) {
- VLOG_ERR("%s: socket: %s", target, strerror(error));
+ VLOG_ERR("%s: socket: %s", target, ovs_strerror(error));
goto error;
}
/* Listen. */
if (style == SOCK_STREAM && listen(fd, 10) < 0) {
error = errno;
- VLOG_ERR("%s: listen: %s", target, strerror(error));
+ VLOG_ERR("%s: listen: %s", target, ovs_strerror(error));
goto error;
}
- kernel_chooses_port = sin.sin_port == htons(0);
- if (sinp || kernel_chooses_port) {
- socklen_t sin_len = sizeof sin;
- if (getsockname(fd, (struct sockaddr *) &sin, &sin_len) < 0) {
+ if (ssp || kernel_chooses_port) {
+ socklen_t ss_len = sizeof ss;
+ if (getsockname(fd, (struct sockaddr *) &ss, &ss_len) < 0) {
error = errno;
- VLOG_ERR("%s: getsockname: %s", target, strerror(error));
+ VLOG_ERR("%s: getsockname: %s", target, ovs_strerror(error));
goto error;
}
- if (sin.sin_family != AF_INET || sin_len != sizeof sin) {
- error = EAFNOSUPPORT;
- VLOG_ERR("%s: getsockname: invalid socket name", target);
- goto error;
- }
- if (sinp) {
- *sinp = sin;
- }
if (kernel_chooses_port) {
VLOG_INFO("%s: listening on port %"PRIu16,
- target, ntohs(sin.sin_port));
+ target, ss_get_port(&ss));
+ }
+ if (ssp) {
+ *ssp = ss;
}
}
return fd;
error:
+ if (ssp) {
+ memset(ssp, 0, sizeof *ssp);
+ }
close(fd);
return -error;
}
int
get_null_fd(void)
{
- static int null_fd = -1;
- if (null_fd < 0) {
+ 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", strerror(error));
- return -error;
+ VLOG_ERR("could not open /dev/null: %s", ovs_strerror(error));
+ null_fd = -error;
}
+ ovsthread_once_done(&once);
}
+
return null_fd;
}
* really an error. */
} else {
error = errno;
- VLOG_ERR("%s: fsync failed (%s)", dir, strerror(error));
+ VLOG_ERR("%s: fsync failed (%s)", dir, ovs_strerror(error));
}
}
close(fd);
} else {
error = errno;
- VLOG_ERR("%s: open failed (%s)", dir, strerror(error));
+ VLOG_ERR("%s: open failed (%s)", dir, ovs_strerror(error));
}
free(dir);
xpipe(int fds[2])
{
if (pipe(fds)) {
- VLOG_FATAL("failed to create pipe (%s)", strerror(errno));
+ VLOG_FATAL("failed to create pipe (%s)", ovs_strerror(errno));
}
}
xset_nonblocking(fds[1]);
}
-void
-xsocketpair(int domain, int type, int protocol, int fds[2])
-{
- if (socketpair(domain, type, protocol, fds)) {
- VLOG_FATAL("failed to create socketpair (%s)", strerror(errno));
- }
-}
-
static int
getsockopt_int(int fd, int level, int option, const char *optname, int *valuep)
{
len = sizeof value;
if (getsockopt(fd, level, option, &value, &len)) {
error = errno;
- VLOG_ERR_RL(&rl, "getsockopt(%s): %s", optname, strerror(error));
+ VLOG_ERR_RL(&rl, "getsockopt(%s): %s", optname, ovs_strerror(error));
} else if (len != sizeof value) {
error = EINVAL;
- VLOG_ERR_RL(&rl, "getsockopt(%s): value is %u bytes (expected %zu)",
+ VLOG_ERR_RL(&rl, "getsockopt(%s): value is %u bytes (expected %"PRIuSIZE")",
optname, (unsigned int) len, sizeof value);
} else {
error = 0;
socklen_t len = sizeof ss;
if (!getaddr(fd, (struct sockaddr *) &ss, &len)) {
- if (ss.ss_family == AF_INET) {
- struct sockaddr_in sin;
+ if (ss.ss_family == AF_INET || ss.ss_family == AF_INET6) {
+ char addrbuf[SS_NTOP_BUFSIZE];
- memcpy(&sin, &ss, sizeof sin);
- ds_put_format(string, IP_FMT":%"PRIu16,
- IP_ARGS(sin.sin_addr.s_addr), ntohs(sin.sin_port));
+ ds_put_format(string, "%s:%"PRIu16,
+ ss_format_address(&ss, addrbuf, sizeof addrbuf),
+ ss_get_port(&ss));
} else if (ss.ss_family == AF_UNIX) {
struct sockaddr_un sun;
const char *null;
ds_init(&string);
if (fstat(fd, &s)) {
- ds_put_format(&string, "fstat failed (%s)", strerror(errno));
+ ds_put_format(&string, "fstat failed (%s)", ovs_strerror(errno));
} else if (S_ISSOCK(s.st_mode)) {
describe_sockaddr(&string, fd, getsockname);
ds_put_cstr(&string, "<->");
return ds_steal_cstr(&string);
}
-/* Returns the total of the 'iov_len' members of the 'n_iovs' in 'iovs'.
- * The caller must ensure that the total does not exceed SIZE_MAX. */
-size_t
-iovec_len(const struct iovec iovs[], size_t n_iovs)
-{
- size_t len = 0;
- size_t i;
-
- for (i = 0; i < n_iovs; i++) {
- len += iovs[i].iov_len;
- }
- return len;
-}
-
-/* Returns true if all of the 'n_iovs' iovecs in 'iovs' have length zero. */
-bool
-iovec_is_empty(const struct iovec iovs[], size_t n_iovs)
+/* 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)
{
- size_t i;
-
- for (i = 0; i < n_iovs; i++) {
- if (iovs[i].iov_len) {
- return false;
+ 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) {
+ sock = -errno;
+ VLOG_ERR("failed to create inet socket: %s", ovs_strerror(errno));
}
+ ovsthread_once_done(&once);
}
- return true;
-}
-/* Sends the 'n_iovs' iovecs of data in 'iovs' and the 'n_fds' file descriptors
- * in 'fds' on Unix domain socket 'sock'. Returns the number of bytes
- * successfully sent or -1 if an error occurred. On error, sets errno
- * appropriately. */
-int
-send_iovec_and_fds(int sock,
- const struct iovec *iovs, size_t n_iovs,
- const int fds[], size_t n_fds)
-{
- ovs_assert(sock >= 0);
- if (n_fds > 0) {
- union {
- struct cmsghdr cm;
- char control[CMSG_SPACE(SOUTIL_MAX_FDS * sizeof *fds)];
- } cmsg;
- struct msghdr msg;
-
- ovs_assert(!iovec_is_empty(iovs, n_iovs));
- ovs_assert(n_fds <= SOUTIL_MAX_FDS);
-
- memset(&cmsg, 0, sizeof cmsg);
- cmsg.cm.cmsg_len = CMSG_LEN(n_fds * sizeof *fds);
- cmsg.cm.cmsg_level = SOL_SOCKET;
- cmsg.cm.cmsg_type = SCM_RIGHTS;
- memcpy(CMSG_DATA(&cmsg.cm), fds, n_fds * sizeof *fds);
-
- msg.msg_name = NULL;
- msg.msg_namelen = 0;
- msg.msg_iov = CONST_CAST(struct iovec *, iovs);
- msg.msg_iovlen = n_iovs;
- msg.msg_control = &cmsg.cm;
- msg.msg_controllen = CMSG_SPACE(n_fds * sizeof *fds);
- msg.msg_flags = 0;
-
- return sendmsg(sock, &msg, 0);
- } else {
- return writev(sock, iovs, n_iovs);
- }
+ return (sock < 0 ? -sock
+ : ioctl(sock, command, arg) == -1 ? errno
+ : 0);
}
-/* Sends the 'n_iovs' iovecs of data in 'iovs' and the 'n_fds' file descriptors
- * in 'fds' on Unix domain socket 'sock'. If 'skip_bytes' is nonzero, then the
- * first 'skip_bytes' of data in the iovecs are not sent, and none of the file
- * descriptors are sent. The function continues to retry sending until an
- * error (other than EINTR) occurs or all the data and fds are sent.
- *
- * Returns 0 if all the data and fds were successfully sent, otherwise a
- * positive errno value. Regardless of success, stores the number of bytes
- * sent (always at least 'skip_bytes') in '*bytes_sent'. (If at least one byte
- * is sent, then all the fds have been sent.)
- *
- * 'skip_bytes' must be less than or equal to iovec_len(iovs, n_iovs). */
int
-send_iovec_and_fds_fully(int sock,
- const struct iovec iovs[], size_t n_iovs,
- const int fds[], size_t n_fds,
- size_t skip_bytes, size_t *bytes_sent)
+af_inet_ifreq_ioctl(const char *name, struct ifreq *ifr, unsigned long int cmd,
+ const char *cmd_name)
{
- *bytes_sent = 0;
- while (n_iovs > 0) {
- int retval;
-
- if (skip_bytes) {
- retval = skip_bytes;
- skip_bytes = 0;
- } else if (!*bytes_sent) {
- retval = send_iovec_and_fds(sock, iovs, n_iovs, fds, n_fds);
- } else {
- retval = writev(sock, iovs, n_iovs);
- }
+ int error;
- if (retval > 0) {
- *bytes_sent += retval;
- while (retval > 0) {
- const uint8_t *base = iovs->iov_base;
- size_t len = iovs->iov_len;
-
- if (retval < len) {
- size_t sent;
- int error;
-
- error = write_fully(sock, base + retval, len - retval,
- &sent);
- *bytes_sent += sent;
- retval += sent;
- if (error) {
- return error;
- }
- }
- retval -= len;
- iovs++;
- n_iovs--;
- }
- } else if (retval == 0) {
- if (iovec_is_empty(iovs, n_iovs)) {
- break;
- }
- VLOG_WARN("send returned 0");
- return EPROTO;
- } else if (errno != EINTR) {
- return errno;
- }
+ 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 0;
+ return error;
}
+\f
+/* sockaddr_storage helpers. */
-/* Sends the 'n_iovs' iovecs of data in 'iovs' and the 'n_fds' file descriptors
- * in 'fds' on Unix domain socket 'sock'. The function continues to retry
- * sending until an error (other than EAGAIN or EINTR) occurs or all the data
- * and fds are sent. Upon EAGAIN, the function blocks until the socket is
- * ready for more data.
- *
- * Returns 0 if all the data and fds were successfully sent, otherwise a
- * positive errno value. */
-int
-send_iovec_and_fds_fully_block(int sock,
- const struct iovec iovs[], size_t n_iovs,
- const int fds[], size_t n_fds)
+/* Returns the IPv4 or IPv6 port in 'ss'. */
+uint16_t
+ss_get_port(const struct sockaddr_storage *ss)
{
- size_t sent = 0;
-
- for (;;) {
- int error;
-
- error = send_iovec_and_fds_fully(sock, iovs, n_iovs,
- fds, n_fds, sent, &sent);
- if (error != EAGAIN) {
- return error;
- }
- poll_fd_wait(sock, POLLOUT);
- poll_block();
+ 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();
}
}
-/* Attempts to receive from Unix domain socket 'sock' up to 'size' bytes of
- * data into 'data' and up to SOUTIL_MAX_FDS file descriptors into 'fds'.
- *
- * - Upon success, returns the number of bytes of data copied into 'data'
- * and stores the number of received file descriptors into '*n_fdsp'.
+/* 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.
*
- * - On failure, returns a negative errno value and stores 0 in
- * '*n_fdsp'.
- *
- * - On EOF, returns 0 and stores 0 in '*n_fdsp'. */
-int
-recv_data_and_fds(int sock,
- void *data, size_t size,
- int fds[SOUTIL_MAX_FDS], size_t *n_fdsp)
+ * Returns 'buf'. */
+char *
+ss_format_address(const struct sockaddr_storage *ss,
+ char *buf, size_t bufsize)
{
- union {
- struct cmsghdr cm;
- char control[CMSG_SPACE(SOUTIL_MAX_FDS * sizeof *fds)];
- } cmsg;
- struct msghdr msg;
- int retval;
- struct cmsghdr *p;
- size_t i;
-
- *n_fdsp = 0;
-
- do {
- struct iovec iov;
-
- iov.iov_base = data;
- iov.iov_len = size;
-
- msg.msg_name = NULL;
- msg.msg_namelen = 0;
- msg.msg_iov = &iov;
- msg.msg_iovlen = 1;
- msg.msg_control = &cmsg.cm;
- msg.msg_controllen = sizeof cmsg.control;
- msg.msg_flags = 0;
-
- retval = recvmsg(sock, &msg, 0);
- } while (retval < 0 && errno == EINTR);
- if (retval <= 0) {
- return retval < 0 ? -errno : 0;
+ 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'), "]");
+ } else {
+ OVS_NOT_REACHED();
}
- for (p = CMSG_FIRSTHDR(&msg); p; p = CMSG_NXTHDR(&msg, p)) {
- if (p->cmsg_level != SOL_SOCKET || p->cmsg_type != SCM_RIGHTS) {
- VLOG_ERR("unexpected control message %d:%d",
- p->cmsg_level, p->cmsg_type);
- goto error;
- } else if (*n_fdsp) {
- VLOG_ERR("multiple SCM_RIGHTS received");
- goto error;
- } else {
- size_t n_fds = (p->cmsg_len - CMSG_LEN(0)) / sizeof *fds;
- const int *fds_data = (const int *) CMSG_DATA(p);
-
- ovs_assert(n_fds > 0);
- if (n_fds > SOUTIL_MAX_FDS) {
- VLOG_ERR("%zu fds received but only %d supported",
- n_fds, SOUTIL_MAX_FDS);
- for (i = 0; i < n_fds; i++) {
- close(fds_data[i]);
- }
- goto error;
- }
+ return buf;
+}
- *n_fdsp = n_fds;
- memcpy(fds, fds_data, n_fds * sizeof *fds);
- }
- }
+size_t
+ss_length(const struct sockaddr_storage *ss)
+{
+ switch (ss->ss_family) {
+ case AF_INET:
+ return sizeof(struct sockaddr_in);
- return retval;
+ case AF_INET6:
+ return sizeof(struct sockaddr_in6);
-error:
- for (i = 0; i < *n_fdsp; i++) {
- close(fds[i]);
+ default:
+ OVS_NOT_REACHED();
}
- *n_fdsp = 0;
- return EPROTO;
}