X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fsocket-util.c;h=2dff9f5924331495790e43f89e244f053fa8ecbf;hb=a39c3ff9f586d3416976d3c3bd095fa81d73efa0;hp=7c40ab8330d6d298a59a8308986d4cf4e0baaf70;hpb=a7a27d07190ee51904c0daeac57b88f881104227;p=sliver-openvswitch.git

diff --git a/lib/socket-util.c b/lib/socket-util.c
index 7c40ab833..2dff9f592 100644
--- a/lib/socket-util.c
+++ b/lib/socket-util.c
@@ -1,5 +1,5 @@
 /*
- * 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.
@@ -17,7 +17,6 @@
 #include <config.h>
 #include "socket-util.h"
 #include <arpa/inet.h>
-#include <assert.h>
 #include <errno.h>
 #include <fcntl.h>
 #include <net/if.h>
@@ -39,7 +38,7 @@
 #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
@@ -51,11 +50,9 @@ 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)" 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
@@ -92,7 +89,7 @@ xset_nonblocking(int fd)
     }
 }
 
-static int
+int
 set_dscp(int fd, uint8_t dscp)
 {
     int val;
@@ -181,48 +178,67 @@ lookup_ipv6(const char *host_name, struct in6_addr *addr)
  * 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;
 
@@ -232,9 +248,17 @@ check_connection_completion(int fd)
         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));
         return errno;
     } else {
@@ -266,7 +290,7 @@ drain_rcvbuf(int fd)
          *
          * 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) {
@@ -335,7 +359,7 @@ make_sockaddr_un(const char *name, struct sockaddr_un *un, socklen_t *un_len,
     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
@@ -416,13 +440,8 @@ make_unix_socket(int style, bool nonblock,
      * 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;
         }
     }
@@ -614,6 +633,7 @@ exit:
         }
     } else if (fd >= 0) {
         close(fd);
+        fd = -1;
     }
     *fdp = fd;
     return error;
@@ -695,6 +715,7 @@ int
 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;
@@ -744,9 +765,10 @@ inet_open_passive(int style, const char *target, int default_port,
         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));
             goto error;
@@ -756,7 +778,13 @@ inet_open_passive(int style, const char *target, int default_port,
             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;
@@ -893,6 +921,14 @@ xpipe(int fds[2])
     }
 }
 
+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])
 {
@@ -938,7 +974,7 @@ describe_sockaddr(struct ds *string, int fd,
 
             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;
@@ -980,7 +1016,7 @@ describe_sockaddr(struct ds *string, int fd,
             }
         }
 #endif
-#if AF_PACKET && __linux__
+#if AF_PACKET && LINUX_DATAPATH
         else if (ss.ss_family == AF_PACKET) {
             struct sockaddr_ll sll;
 
@@ -1010,7 +1046,7 @@ describe_sockaddr(struct ds *string, int fd,
 }
 
 
-#ifdef __linux__
+#ifdef LINUX_DATAPATH
 static void
 put_fd_filename(struct ds *string, int fd)
 {
@@ -1054,7 +1090,7 @@ describe_fd(int fd)
                               : 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
     }
@@ -1098,7 +1134,7 @@ send_iovec_and_fds(int sock,
                    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;
@@ -1106,8 +1142,8 @@ send_iovec_and_fds(int sock,
         } 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);
@@ -1117,7 +1153,7 @@ send_iovec_and_fds(int sock,
 
         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);
@@ -1282,7 +1318,7 @@ recv_data_and_fds(int sock,
             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);