/*
- * Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc.
+ * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 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 <arpa/inet.h>
-#include <assert.h>
#include <errno.h>
#include <fcntl.h>
#include <net/if.h>
#include "poll-loop.h"
#include "util.h"
#include "vlog.h"
-#if AF_PACKET && __linux__
+#if AF_PACKET && LINUX_DATAPATH
#include <linux/if_packet.h>
#endif
#ifdef HAVE_NETLINK
/* #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)" instead of "#ifdef __linux__". */
-#ifdef __linux__
-#define LINUX 1
-#else
-#define LINUX 0
+ * writing "if (LINUX_DATAPATH)" instead of "#ifdef __linux__". */
+#ifndef LINUX_DATAPATH
+#define LINUX_DATAPATH 0
#endif
#ifndef O_DIRECTORY
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;
}
}
* successful, otherwise a positive errno value.
*
* Most Open vSwitch code should not use this because it causes deadlocks:
- * gethostbyname() sends out a DNS request but that starts a new flow for which
+ * getaddrinfo() sends out a DNS request but that starts a new flow for which
* OVS must set up a flow, but it can't because it's waiting for a DNS reply.
* The synchronous lookup also delays other activity. (Of course we can solve
* this but it doesn't seem worthwhile quite yet.) */
int
lookup_hostname(const char *host_name, struct in_addr *addr)
{
- struct hostent *h;
+ struct addrinfo *result;
+ struct addrinfo hints;
if (inet_aton(host_name, addr)) {
return 0;
}
- h = gethostbyname(host_name);
- if (h) {
- *addr = *(struct in_addr *) h->h_addr;
+ memset(&hints, 0, sizeof hints);
+ hints.ai_family = AF_INET;
+
+ switch (getaddrinfo(host_name, NULL, &hints, &result)) {
+ case 0:
+ *addr = ((struct sockaddr_in *) result->ai_addr)->sin_addr;
+ freeaddrinfo(result);
return 0;
- }
- return (h_errno == HOST_NOT_FOUND ? ENOENT
- : h_errno == TRY_AGAIN ? EAGAIN
- : h_errno == NO_RECOVERY ? EIO
- : h_errno == NO_ADDRESS ? ENXIO
- : EINVAL);
-}
+#ifdef EAI_ADDRFAMILY
+ case EAI_ADDRFAMILY:
+#endif
+ case EAI_NONAME:
+ case EAI_SERVICE:
+ return ENOENT;
-/* Returns the error condition associated with socket 'fd' and resets the
- * socket's error status. */
-int
-get_socket_error(int fd)
-{
- int error;
+ case EAI_AGAIN:
+ return EAGAIN;
- if (getsockopt_int(fd, SOL_SOCKET, SO_ERROR, "SO_ERROR", &error)) {
- error = errno;
+ case EAI_BADFLAGS:
+ case EAI_FAMILY:
+ case EAI_SOCKTYPE:
+ return EINVAL;
+
+ case EAI_FAIL:
+ return EIO;
+
+ case EAI_MEMORY:
+ return ENOMEM;
+
+#ifdef EAI_NODATA
+ case EAI_NODATA:
+ return ENXIO;
+#endif
+
+ case EAI_SYSTEM:
+ return errno;
+
+ default:
+ return EPROTO;
}
- return error;
}
int
check_connection_completion(int fd)
{
+ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10);
struct pollfd pfd;
int retval;
retval = poll(&pfd, 1, 0);
} while (retval < 0 && errno == EINTR);
if (retval == 1) {
- return get_socket_error(fd);
+ if (pfd.revents & POLLERR) {
+ ssize_t n = send(fd, "", 1, MSG_DONTWAIT);
+ if (n < 0) {
+ return errno;
+ } else {
+ VLOG_ERR_RL(&rl, "poll return POLLERR but send succeeded");
+ return EPROTO;
+ }
+ }
+ return 0;
} else if (retval < 0) {
- static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10);
- VLOG_ERR_RL(&rl, "poll: %s", strerror(errno));
+ VLOG_ERR_RL(&rl, "poll: %s", ovs_strerror(errno));
return errno;
} else {
return EAGAIN;
*
* On other Unix-like OSes, MSG_TRUNC has no effect in the flags
* argument. */
- char buffer[LINUX ? 1 : 2048];
+ 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) {
if (strlen(name) > MAX_UN_LEN) {
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
- if (LINUX) {
+ 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
* 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) {
- int flags = fcntl(fd, F_GETFL, 0);
- if (flags == -1) {
- error = errno;
- goto error;
- }
- if (fcntl(fd, F_SETFL, flags | O_NONBLOCK) == -1) {
- error = errno;
+ error = set_nonblocking(fd);
+ if (error) {
goto error;
}
}
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);
/* Create non-blocking socket. */
fd = socket(AF_INET, 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;
}
* 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;
}
}
} else if (fd >= 0) {
close(fd);
+ fd = -1;
}
*fdp = fd;
return error;
inet_open_passive(int style, const char *target, int default_port,
struct sockaddr_in *sinp, uint8_t dscp)
{
+ bool kernel_chooses_port;
struct sockaddr_in sin;
int fd = 0, error;
unsigned int yes = 1;
fd = socket(AF_INET, 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) {
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;
}
- if (sinp) {
+ 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 (getsockname(fd, (struct sockaddr *) &sin, &sin_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) {
VLOG_ERR("%s: getsockname: invalid socket name", target);
goto error;
}
- *sinp = sin;
+ if (sinp) {
+ *sinp = sin;
+ }
+ if (kernel_chooses_port) {
+ VLOG_INFO("%s: listening on port %"PRIu16,
+ target, ntohs(sin.sin_port));
+ }
}
return fd;
null_fd = open("/dev/null", O_RDWR);
if (null_fd < 0) {
int error = errno;
- VLOG_ERR("could not open /dev/null: %s", strerror(error));
+ VLOG_ERR("could not open /dev/null: %s", ovs_strerror(error));
return -error;
}
}
* 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));
}
}
+void
+xpipe_nonblocking(int fds[2])
+{
+ xpipe(fds);
+ xset_nonblocking(fds[0]);
+ 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));
+ VLOG_FATAL("failed to create socketpair (%s)", ovs_strerror(errno));
}
}
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)",
memcpy(&sin, &ss, sizeof sin);
ds_put_format(string, IP_FMT":%"PRIu16,
- IP_ARGS(&sin.sin_addr.s_addr), ntohs(sin.sin_port));
+ IP_ARGS(sin.sin_addr.s_addr), ntohs(sin.sin_port));
} else if (ss.ss_family == AF_UNIX) {
struct sockaddr_un sun;
const char *null;
}
}
#endif
-#if AF_PACKET && __linux__
+#if AF_PACKET && LINUX_DATAPATH
else if (ss.ss_family == AF_PACKET) {
struct sockaddr_ll sll;
}
-#ifdef __linux__
+#ifdef LINUX_DATAPATH
static void
put_fd_filename(struct ds *string, int fd)
{
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, "<->");
: S_ISFIFO(s.st_mode) ? "FIFO"
: S_ISLNK(s.st_mode) ? "symbolic link"
: "unknown"));
-#ifdef __linux__
+#ifdef LINUX_DATAPATH
put_fd_filename(&string, fd);
#endif
}
const struct iovec *iovs, size_t n_iovs,
const int fds[], size_t n_fds)
{
- assert(sock >= 0);
+ ovs_assert(sock >= 0);
if (n_fds > 0) {
union {
struct cmsghdr cm;
} cmsg;
struct msghdr msg;
- assert(!iovec_is_empty(iovs, n_iovs));
- assert(n_fds <= SOUTIL_MAX_FDS);
+ 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);
msg.msg_name = NULL;
msg.msg_namelen = 0;
- msg.msg_iov = (struct iovec *) iovs;
+ 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);
size_t n_fds = (p->cmsg_len - CMSG_LEN(0)) / sizeof *fds;
const int *fds_data = (const int *) CMSG_DATA(p);
- assert(n_fds > 0);
+ 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);