X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=lib%2Fsocket-util.c;h=6554e972192b47a6310123345312d705f701e9f1;hb=b86545a659a7146006e229d035f874106303f261;hp=07687ba383b9ec1de11aac0dcb683f4b337bfa57;hpb=d98fa5032eaf49b073058dbf390283c9a71001e1;p=openvswitch

diff --git a/lib/socket-util.c b/lib/socket-util.c
index 07687ba3..6554e972 100644
--- a/lib/socket-util.c
+++ b/lib/socket-util.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks.
+ * Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira Networks.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -59,6 +59,9 @@ VLOG_DEFINE_THIS_MODULE(socket_util);
 #define O_DIRECTORY 0
 #endif
 
+static int getsockopt_int(int fd, int level, int option, const char *optname,
+                          int *valuep);
+
 /* Sets 'fd' to non-blocking mode.  Returns 0 if successful, otherwise a
  * positive errno value. */
 int
@@ -145,17 +148,46 @@ lookup_ipv6(const char *host_name, struct in6_addr *addr)
     return 0;
 }
 
+/* Translates 'host_name', which must be a host name or a string representation
+ * of an IP address, into a numeric IP address in '*addr'.  Returns 0 if
+ * 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
+ * OVS must set up a flow, but it can't because it's waiting for a DNS reply.
+ * The synchronous lookup also delays other activty.  (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;
+
+    if (inet_aton(host_name, addr)) {
+        return 0;
+    }
+
+    h = gethostbyname(host_name);
+    if (h) {
+        *addr = *(struct in_addr *) h->h_addr;
+        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);
+}
+
 /* Returns the error condition associated with socket 'fd' and resets the
  * socket's error status. */
 int
 get_socket_error(int fd)
 {
     int error;
-    socklen_t len = sizeof(error);
-    if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &error, &len) < 0) {
-        static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10);
+
+    if (getsockopt_int(fd, SOL_SOCKET, SO_ERROR, "SO_ERROR", &error)) {
         error = errno;
-        VLOG_ERR_RL(&rl, "getsockopt(SO_ERROR): %s", strerror(error));
     }
     return error;
 }
@@ -190,15 +222,13 @@ check_connection_completion(int fd)
 int
 drain_rcvbuf(int fd)
 {
-    socklen_t rcvbuf_len;
-    size_t rcvbuf;
+    int rcvbuf;
 
-    rcvbuf_len = sizeof rcvbuf;
-    if (getsockopt(fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf, &rcvbuf_len) < 0) {
-        static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10);
-        VLOG_ERR_RL(&rl, "getsockopt(SO_RCVBUF) failed: %s", strerror(errno));
-        return errno;
+    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
@@ -219,6 +249,18 @@ drain_rcvbuf(int fd)
     return 0;
 }
 
+/* Returns the size of socket 'sock''s receive buffer (SO_RCVBUF), or a
+ * negative errno value if an error occurs. */
+int
+get_socket_rcvbuf(int sock)
+{
+    int rcvbuf;
+    int error;
+
+    error = getsockopt_int(sock, SOL_SOCKET, SO_RCVBUF, "SO_RCVBUF", &rcvbuf);
+    return error ? -error : rcvbuf;
+}
+
 /* Reads and discards up to 'n' datagrams from 'fd', stopping as soon as no
  * more data can be immediately read.  ('fd' should therefore be in
  * non-blocking mode.)*/
@@ -416,7 +458,7 @@ error:
         error = EPROTO;
     }
     if (bind_path) {
-        fatal_signal_remove_file_to_unlink(bind_path);
+        fatal_signal_unlink_file_now(bind_path);
     }
     close(fd);
     return -error;
@@ -502,10 +544,13 @@ exit:
  * and stores -1 into '*fdp'.
  *
  * If 'sinp' is non-null, then on success the target address is stored into
- * '*sinp'. */
+ * '*sinp'.
+ *
+ * 'dscp'  If not DSCP_INVALID, its value becomes the DSCP bits in the IP
+ * headers for the new connection. */
 int
 inet_open_active(int style, const char *target, uint16_t default_port,
-                 struct sockaddr_in *sinp, int *fdp)
+                 struct sockaddr_in *sinp, int *fdp, uint8_t dscp)
 {
     struct sockaddr_in sin;
     int fd = -1;
@@ -529,6 +574,17 @@ inet_open_active(int style, const char *target, uint16_t default_port,
         goto exit_close;
     }
 
+    /* The socket options set here ensure that the TOS bits are set during
+     * the connection establishment.  If set after connect(), the handshake
+     * SYN frames will be sent with a TOS of 0. */
+    if (dscp != DSCP_INVALID) {
+        if (setsockopt(fd, IPPROTO_IP, IP_TOS, &dscp, sizeof dscp)) {
+            VLOG_ERR("%s: socket: %s", target, strerror(errno));
+            error = errno;
+            goto exit;
+        }
+    }
+
     /* Connect. */
     error = connect(fd, (struct sockaddr *) &sin, sizeof sin) == 0 ? 0 : errno;
     if (error == EINPROGRESS) {
@@ -567,7 +623,7 @@ exit:
  * If successful, stores the address into '*sinp' and returns true; otherwise
  * zeros '*sinp' and returns false. */
 bool
-inet_parse_passive(const char *target_, uint16_t default_port,
+inet_parse_passive(const char *target_, int default_port,
                    struct sockaddr_in *sinp)
 {
     char *target = xstrdup(target_);
@@ -621,17 +677,20 @@ exit:
  * negative errno value.
  *
  * If 'sinp' is non-null, then on success the bound address is stored into
- * '*sinp'. */
+ * '*sinp'.
+ *
+ * 'dscp'  If not DSCP_INVALID, its value becomes the DSCP bits in the IP
+ * headers for the new connection. */
 int
 inet_open_passive(int style, const char *target, int default_port,
-                  struct sockaddr_in *sinp)
+                  struct sockaddr_in *sinp, uint8_t dscp)
 {
     struct sockaddr_in sin;
     int fd = 0, error;
     unsigned int yes = 1;
 
     if (!inet_parse_passive(target, default_port, &sin)) {
-        return EAFNOSUPPORT;
+        return -EAFNOSUPPORT;
     }
 
     /* Create non-blocking socket, set SO_REUSEADDR. */
@@ -639,7 +698,7 @@ inet_open_passive(int style, const char *target, int default_port,
     if (fd < 0) {
         error = errno;
         VLOG_ERR("%s: socket: %s", target, strerror(error));
-        return error;
+        return -error;
     }
     error = set_nonblocking(fd);
     if (error) {
@@ -659,8 +718,19 @@ inet_open_passive(int style, const char *target, int default_port,
         goto error;
     }
 
+    /* The socket options set here ensure that the TOS bits are set during
+     * the connection establishment.  If set after connect(), the handshake
+     * SYN frames will be sent with a TOS of 0. */
+    if (dscp != DSCP_INVALID) {
+        if (setsockopt(fd, IPPROTO_IP, IP_TOS, &dscp, sizeof dscp)) {
+            VLOG_ERR("%s: socket: %s", target, strerror(errno));
+            error = errno;
+            goto error;
+        }
+    }
+
     /* Listen. */
-    if (listen(fd, 10) < 0) {
+    if (style == SOCK_STREAM && listen(fd, 10) < 0) {
         error = errno;
         VLOG_ERR("%s: listen: %s", target, strerror(error));
         goto error;
@@ -674,6 +744,7 @@ inet_open_passive(int style, const char *target, int default_port,
             goto error;
         }
         if (sin.sin_family != AF_INET || sin_len != sizeof sin) {
+            error = EAFNOSUPPORT;
             VLOG_ERR("%s: getsockname: invalid socket name", target);
             goto error;
         }
@@ -684,7 +755,7 @@ inet_open_passive(int style, const char *target, int default_port,
 
 error:
     close(fd);
-    return error;
+    return -error;
 }
 
 /* Returns a readable and writable fd for /dev/null, if successful, otherwise
@@ -815,13 +886,27 @@ xpipe(int fds[2])
 }
 
 static int
-getsockopt_int(int fd, int level, int optname, int *valuep)
+getsockopt_int(int fd, int level, int option, const char *optname, int *valuep)
 {
-    socklen_t len = sizeof *valuep;
+    static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10);
+    socklen_t len;
+    int value;
+    int error;
 
-    return (getsockopt(fd, level, optname, valuep, &len) ? errno
-            : len == sizeof *valuep ? 0
-            : EINVAL);
+    len = sizeof value;
+    if (getsockopt(fd, level, option, &value, &len)) {
+        error = errno;
+        VLOG_ERR_RL(&rl, "getsockopt(%s): %s", optname, strerror(error));
+    } else if (len != sizeof value) {
+        error = EINVAL;
+        VLOG_ERR_RL(&rl, "getsockopt(%s): value is %u bytes (expected %zu)",
+                    optname, (unsigned int) len, sizeof value);
+    } else {
+        error = 0;
+    }
+
+    *valuep = error ? 0 : value;
+    return error;
 }
 
 static void
@@ -859,7 +944,8 @@ describe_sockaddr(struct ds *string, int fd,
 #define SO_PROTOCOL 38
 #endif
 
-            if (!getsockopt_int(fd, SOL_SOCKET, SO_PROTOCOL, &protocol)) {
+            if (!getsockopt_int(fd, SOL_SOCKET, SO_PROTOCOL, "SO_PROTOCOL",
+                                &protocol)) {
                 switch (protocol) {
                 case NETLINK_ROUTE:
                     ds_put_cstr(string, "NETLINK_ROUTE");