/* Establishing and handling network connections.
- Copyright (C) 1995, 1996, 1997, 2001, 2002 Free Software Foundation, Inc.
+ Copyright (C) 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
+ 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011 Free Software
+ Foundation, Inc.
This file is part of GNU Wget.
GNU Wget is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
GNU Wget is distributed in the hope that it will be useful,
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
-along with Wget; if not, write to the Free Software
-Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+along with Wget. If not, see <http://www.gnu.org/licenses/>.
-In addition, as a special exception, the Free Software Foundation
-gives permission to link the code of its release of Wget with the
-OpenSSL project's "OpenSSL" library (or with modified versions of it
-that use the same license as the "OpenSSL" library), and distribute
-the linked executables. You must obey the GNU General Public License
-in all respects for all of the code used other than "OpenSSL". If you
-modify this file, you may extend this exception to your version of the
-file, but you are not obligated to do so. If you do not wish to do
-so, delete this exception statement from your version. */
+Additional permission under GNU GPL version 3 section 7
-#include <config.h>
+If you modify this program, or any covered work, by linking or
+combining it with the OpenSSL project's OpenSSL library (or a
+modified version of that library), containing parts covered by the
+terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
+grants you additional permission to convey the resulting work.
+Corresponding Source for a non-source form of such a combination
+shall include the source code for the parts of OpenSSL used as well
+as that of the covered work. */
+
+#include "wget.h"
#include <stdio.h>
#include <stdlib.h>
-#include <sys/types.h>
-#ifdef HAVE_UNISTD_H
-# include <unistd.h>
-#endif
+#include <unistd.h>
#include <assert.h>
+#include <sys/socket.h>
+#include <sys/select.h>
+
#ifndef WINDOWS
-# include <sys/socket.h>
-# include <netdb.h>
+# ifdef __VMS
+# include "vms_ip.h"
+# else /* def __VMS */
+# include <netdb.h>
+# endif /* def __VMS [else] */
# include <netinet/in.h>
# ifndef __BEOS__
# include <arpa/inet.h>
#endif /* not WINDOWS */
#include <errno.h>
-#ifdef HAVE_STRING_H
-# include <string.h>
-#else
-# include <strings.h>
-#endif /* HAVE_STRING_H */
-#ifdef HAVE_SYS_SELECT_H
-# include <sys/select.h>
-#endif /* HAVE_SYS_SELECT_H */
-
-#include "wget.h"
+#include <string.h>
+#include <sys/time.h>
#include "utils.h"
#include "host.h"
#include "connect.h"
#include "hash.h"
-#ifndef errno
-extern int errno;
+/* Apparently needed for Interix: */
+#ifdef HAVE_STDINT_H
+# include <stdint.h>
#endif
-\f
+/* Define sockaddr_storage where unavailable (presumably on IPv4-only
+ hosts). */
+
+#ifndef ENABLE_IPV6
+# ifndef HAVE_STRUCT_SOCKADDR_STORAGE
+# define sockaddr_storage sockaddr_in
+# endif
+#endif /* ENABLE_IPV6 */
+
/* Fill SA as per the data in IP and PORT. SA shoult point to struct
sockaddr_storage if ENABLE_IPV6 is defined, to struct sockaddr_in
otherwise. */
static void
sockaddr_set_data (struct sockaddr *sa, const ip_address *ip, int port)
{
- switch (ip->type)
+ switch (ip->family)
{
- case IPV4_ADDRESS:
+ case AF_INET:
{
- struct sockaddr_in *sin = (struct sockaddr_in *)sa;
- sin->sin_family = AF_INET;
- sin->sin_port = htons (port);
- sin->sin_addr = ADDRESS_IPV4_IN_ADDR (ip);
- break;
+ struct sockaddr_in *sin = (struct sockaddr_in *)sa;
+ xzero (*sin);
+ sin->sin_family = AF_INET;
+ sin->sin_port = htons (port);
+ sin->sin_addr = ip->data.d4;
+ break;
}
#ifdef ENABLE_IPV6
- case IPV6_ADDRESS:
+ case AF_INET6:
{
- struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
- sin6->sin6_family = AF_INET6;
- sin6->sin6_port = htons (port);
- sin6->sin6_addr = ADDRESS_IPV6_IN6_ADDR (ip);
+ struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
+ xzero (*sin6);
+ sin6->sin6_family = AF_INET6;
+ sin6->sin6_port = htons (port);
+ sin6->sin6_addr = ip->data.d6;
#ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
- sin6->sin6_scope_id = ADDRESS_IPV6_SCOPE (ip);
+ sin6->sin6_scope_id = ip->ipv6_scope;
#endif
- break;
+ break;
}
#endif /* ENABLE_IPV6 */
default:
{
case AF_INET:
{
- struct sockaddr_in *sin = (struct sockaddr_in *)sa;
- if (ip)
- {
- ip->type = IPV4_ADDRESS;
- ADDRESS_IPV4_IN_ADDR (ip) = sin->sin_addr;
- }
- if (port)
- *port = ntohs (sin->sin_port);
- break;
+ struct sockaddr_in *sin = (struct sockaddr_in *)sa;
+ if (ip)
+ {
+ ip->family = AF_INET;
+ ip->data.d4 = sin->sin_addr;
+ }
+ if (port)
+ *port = ntohs (sin->sin_port);
+ break;
}
#ifdef ENABLE_IPV6
case AF_INET6:
{
- struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
- if (ip)
- {
- ip->type = IPV6_ADDRESS;
- ADDRESS_IPV6_IN6_ADDR (ip) = sin6->sin6_addr;
+ struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sa;
+ if (ip)
+ {
+ ip->family = AF_INET6;
+ ip->data.d6 = sin6->sin6_addr;
#ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
- ADDRESS_IPV6_SCOPE (ip) = sin6->sin6_scope_id;
+ ip->ipv6_scope = sin6->sin6_scope_id;
#endif
- }
- if (port)
- *port = ntohs (sin6->sin6_port);
- break;
+ }
+ if (port)
+ *port = ntohs (sin6->sin6_port);
+ break;
}
#endif
default:
#endif
default:
abort ();
- return 0; /* so the compiler shuts up. */
}
}
\f
-static int
+/* Resolve the bind address specified via --bind-address and store it
+ to SA. The resolved value is stored in a static variable and
+ reused after the first invocation of this function.
+
+ Returns true on success, false on failure. */
+
+static bool
resolve_bind_address (struct sockaddr *sa)
{
struct address_list *al;
/* Make sure this is called only once. opt.bind_address doesn't
change during a Wget run. */
- static int called, should_bind;
+ static bool called, should_bind;
static ip_address ip;
if (called)
{
if (should_bind)
- sockaddr_set_data (sa, &ip, 0);
+ sockaddr_set_data (sa, &ip, 0);
return should_bind;
}
- called = 1;
+ called = true;
- al = lookup_host_passive (opt.bind_address);
+ al = lookup_host (opt.bind_address, LH_BIND | LH_SILENT);
if (!al)
{
/* #### We should be able to print the error message here. */
logprintf (LOG_NOTQUIET,
- _("%s: unable to resolve bind address `%s'; disabling bind.\n"),
- exec_name, opt.bind_address);
- should_bind = 0;
- return 0;
+ _("%s: unable to resolve bind address %s; disabling bind.\n"),
+ exec_name, quote (opt.bind_address));
+ should_bind = false;
+ return false;
}
/* Pick the first address in the list and use it as bind address.
address_list_release (al);
sockaddr_set_data (sa, &ip, 0);
- should_bind = 1;
- return 1;
+ should_bind = true;
+ return true;
}
\f
struct cwt_context {
static int
connect_with_timeout (int fd, const struct sockaddr *addr, socklen_t addrlen,
- double timeout)
+ double timeout)
{
struct cwt_context ctx;
ctx.fd = fd;
return ctx.result;
}
\f
-/* Connect to a remote endpoint whose IP address is known. */
+/* Connect via TCP to the specified address and port.
+
+ If PRINT is non-NULL, it is the host name to print that we're
+ connecting to. */
int
connect_to_ip (const ip_address *ip, int port, const char *print)
{
struct sockaddr_storage ss;
struct sockaddr *sa = (struct sockaddr *)&ss;
- int sock = -1;
+ int sock;
/* If PRINT is non-NULL, print the "Connecting to..." line, with
PRINT being the host name we're connecting to. */
if (print)
{
- const char *txt_addr = pretty_print_address (ip);
- if (print && 0 != strcmp (print, txt_addr))
- logprintf (LOG_VERBOSE,
- _("Connecting to %s|%s|:%d... "), print, txt_addr, port);
+ const char *txt_addr = print_address (ip);
+ if (0 != strcmp (print, txt_addr))
+ {
+ char *str = NULL, *name;
+
+ if (opt.enable_iri && (name = idn_decode ((char *) print)) != NULL)
+ {
+ int len = strlen (print) + strlen (name) + 4;
+ str = xmalloc (len);
+ snprintf (str, len, "%s (%s)", name, print);
+ str[len-1] = '\0';
+ xfree (name);
+ }
+
+ logprintf (LOG_VERBOSE, _("Connecting to %s|%s|:%d... "),
+ str ? str : escnonprint_uri (print), txt_addr, port);
+
+ if (str)
+ xfree (str);
+ }
else
- logprintf (LOG_VERBOSE, _("Connecting to %s:%d... "), txt_addr, port);
+ {
+ if (ip->family == AF_INET)
+ logprintf (LOG_VERBOSE, _("Connecting to %s:%d... "), txt_addr, port);
+#ifdef ENABLE_IPV6
+ else if (ip->family == AF_INET6)
+ logprintf (LOG_VERBOSE, _("Connecting to [%s]:%d... "), txt_addr, port);
+#endif
+ }
}
/* Store the sockaddr info to SA. */
if (sock < 0)
goto err;
+#if defined(ENABLE_IPV6) && defined(IPV6_V6ONLY)
+ if (opt.ipv6_only) {
+ int on = 1;
+ /* In case of error, we will go on anyway... */
+ int err = setsockopt (sock, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof (on));
+ IF_DEBUG
+ if (err < 0)
+ DEBUGP (("Failed setting IPV6_V6ONLY: %s", strerror (errno)));
+ }
+#endif
+
/* For very small rate limits, set the buffer size (and hence,
hopefully, the kernel's TCP window size) to the per-second limit.
That way we should never have to sleep for more than 1s between
{
int bufsize = opt.limit_rate;
if (bufsize < 512)
- bufsize = 512; /* avoid pathologically small values */
+ bufsize = 512; /* avoid pathologically small values */
#ifdef SO_RCVBUF
setsockopt (sock, SOL_SOCKET, SO_RCVBUF,
- (void *)&bufsize, (socklen_t)sizeof (bufsize));
+ (void *)&bufsize, (socklen_t)sizeof (bufsize));
#endif
/* When we add limit_rate support for writing, which is useful
- for POST, we should also set SO_SNDBUF here. */
+ for POST, we should also set SO_SNDBUF here. */
}
if (opt.bind_address)
{
/* Bind the client side of the socket to the requested
- address. */
+ address. */
struct sockaddr_storage bind_ss;
struct sockaddr *bind_sa = (struct sockaddr *)&bind_ss;
if (resolve_bind_address (bind_sa))
- {
+ {
if (bind (sock, bind_sa, sockaddr_size (bind_sa)) < 0)
- goto err;
- }
+ goto err;
+ }
}
/* Connect the socket to the remote endpoint. */
if (connect_with_timeout (sock, sa, sockaddr_size (sa),
- opt.connect_timeout) < 0)
+ opt.connect_timeout) < 0)
goto err;
/* Success. */
logprintf. */
int save_errno = errno;
if (sock >= 0)
- xclose (sock);
+ fd_close (sock);
if (print)
- logprintf (LOG_VERBOSE, "failed: %s.\n", strerror (errno));
+ logprintf (LOG_VERBOSE, _("failed: %s.\n"), strerror (errno));
errno = save_errno;
return -1;
}
}
-/* Connect to a remote endpoint specified by host name. */
+/* Connect via TCP to a remote host on the specified port.
+
+ HOST is resolved as an Internet host name. If HOST resolves to
+ more than one IP address, they are tried in the order returned by
+ DNS until connecting to one of them succeeds. */
int
connect_to_host (const char *host, int port)
{
int i, start, end;
- struct address_list *al;
- int sock = -1;
+ int sock;
+
+ struct address_list *al = lookup_host (host, 0);
- again:
- al = lookup_host (host, 0);
+ retry:
if (!al)
- return E_HOST;
+ {
+ logprintf (LOG_NOTQUIET,
+ _("%s: unable to resolve host address %s\n"),
+ exec_name, quote (host));
+ return E_HOST;
+ }
address_list_get_bounds (al, &start, &end);
for (i = start; i < end; i++)
const ip_address *ip = address_list_address_at (al, i);
sock = connect_to_ip (ip, port, host);
if (sock >= 0)
- /* Success. */
- break;
-
- address_list_set_faulty (al, i);
+ {
+ /* Success. */
+ address_list_set_connected (al);
+ address_list_release (al);
+ return sock;
+ }
/* The attempt to connect has failed. Continue with the loop
- and try next address. */
- }
- address_list_release (al);
-
- if (sock >= 0)
- /* Mark a successful connection to one of the addresses. */
- address_list_set_connected (al);
+ and try next address. */
- if (sock < 0 && address_list_connected_p (al))
- {
- /* We are unable to connect to any of HOST's addresses, although
- we were previously able to connect to HOST. That might
- indicate that HOST is under dynamic DNS and the addresses
- we're connecting to have expired. Resolve it again. */
- forget_host_lookup (host);
- goto again;
+ address_list_set_faulty (al, i);
}
- return sock;
-}
+ /* Failed to connect to any of the addresses in AL. */
-int
-test_socket_open (int sock)
-{
-#ifdef HAVE_SELECT
- fd_set check_set;
- struct timeval to;
-
- /* Check if we still have a valid (non-EOF) connection. From Andrew
- * Maholski's code in the Unix Socket FAQ. */
-
- FD_ZERO (&check_set);
- FD_SET (sock, &check_set);
-
- /* Wait one microsecond */
- to.tv_sec = 0;
- to.tv_usec = 1;
-
- /* If we get a timeout, then that means still connected */
- if (select (sock + 1, &check_set, NULL, NULL, &to) == 0)
+ if (address_list_connected_p (al))
{
- /* Connection is valid (not EOF), so continue */
- return 1;
+ /* We connected to AL before, but cannot do so now. That might
+ indicate that our DNS cache entry for HOST has expired. */
+ address_list_release (al);
+ al = lookup_host (host, LH_REFRESH);
+ goto retry;
}
- else
- return 0;
-#else
- /* Without select, it's hard to know for sure. */
- return 1;
-#endif
-}
+ address_list_release (al);
+ return -1;
+}
+\f
/* Create a socket, bind it to local interface BIND_ADDRESS on port
*PORT, set up a listen backlog, and return the resulting socket, or
-1 in case of error.
bind_local (const ip_address *bind_address, int *port)
{
int sock;
- int family = AF_INET;
struct sockaddr_storage ss;
struct sockaddr *sa = (struct sockaddr *)&ss;
void *setopt_ptr = (void *)&setopt_val;
socklen_t setopt_size = sizeof (setopt_val);
-#ifdef ENABLE_IPV6
- if (bind_address->type == IPV6_ADDRESS)
- family = AF_INET6;
-#endif
-
- sock = socket (family, SOCK_STREAM, 0);
+ sock = socket (bind_address->family, SOCK_STREAM, 0);
if (sock < 0)
return -1;
setsockopt (sock, SOL_SOCKET, SO_REUSEADDR, setopt_ptr, setopt_size);
#endif
-#ifdef ENABLE_IPV6
-# ifdef HAVE_IPV6_V6ONLY
- if (family == AF_INET6)
- setsockopt (sock, IPPROTO_IPV6, IPV6_V6ONLY, setopt_ptr, setopt_size);
-# endif
-#endif
-
xzero (ss);
sockaddr_set_data (sa, bind_address, *port);
if (bind (sock, sa, sockaddr_size (sa)) < 0)
{
- xclose (sock);
+ fd_close (sock);
return -1;
}
DEBUGP (("Local socket fd %d bound.\n", sock));
/* If *PORT is 0, find out which port we've bound to. */
if (*port == 0)
{
- socklen_t sa_len = sockaddr_size (sa);
- if (getsockname (sock, sa, &sa_len) < 0)
- {
- /* If we can't find out the socket's local address ("name"),
- something is seriously wrong with the socket, and it's
- unusable for us anyway because we must know the chosen
- port. */
- xclose (sock);
- return -1;
- }
+ socklen_t addrlen = sockaddr_size (sa);
+ if (getsockname (sock, sa, &addrlen) < 0)
+ {
+ /* If we can't find out the socket's local address ("name"),
+ something is seriously wrong with the socket, and it's
+ unusable for us anyway because we must know the chosen
+ port. */
+ fd_close (sock);
+ return -1;
+ }
sockaddr_get_data (sa, NULL, port);
- DEBUGP (("binding to address %s using port %i.\n",
- pretty_print_address (bind_address), *port));
+ DEBUGP (("binding to address %s using port %i.\n",
+ print_address (bind_address), *port));
}
if (listen (sock, 1) < 0)
{
- xclose (sock);
+ fd_close (sock);
return -1;
}
return sock;
struct sockaddr *sa = (struct sockaddr *)&ss;
socklen_t addrlen = sizeof (ss);
-#ifdef HAVE_SELECT
if (opt.connect_timeout)
{
int test = select_fd (local_sock, opt.connect_timeout, WAIT_FOR_READ);
if (test == 0)
- errno = ETIMEDOUT;
+ errno = ETIMEDOUT;
if (test <= 0)
- return -1;
+ return -1;
}
-#endif
sock = accept (local_sock, sa, &addrlen);
DEBUGP (("Accepted client at socket %d.\n", sock));
return sock;
}
/* Get the IP address associated with the connection on FD and store
- it to IP. Return 1 on success, 0 otherwise.
+ it to IP. Return true on success, false otherwise.
If ENDPOINT is ENDPOINT_LOCAL, it returns the address of the local
(client) side of the socket. Else if ENDPOINT is ENDPOINT_PEER, it
returns the address of the remote (peer's) side of the socket. */
-int
+bool
socket_ip_address (int sock, ip_address *ip, int endpoint)
{
struct sockaddr_storage storage;
- struct sockaddr *sockaddr = (struct sockaddr *)&storage;
+ struct sockaddr *sockaddr = (struct sockaddr *) &storage;
socklen_t addrlen = sizeof (storage);
int ret;
+ memset (sockaddr, 0, addrlen);
if (endpoint == ENDPOINT_LOCAL)
ret = getsockname (sock, sockaddr, &addrlen);
else if (endpoint == ENDPOINT_PEER)
else
abort ();
if (ret < 0)
- return 0;
+ return false;
+ memset(ip, 0, sizeof(ip_address));
+ ip->family = sockaddr->sa_family;
switch (sockaddr->sa_family)
{
#ifdef ENABLE_IPV6
case AF_INET6:
{
- struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)&storage;
- ip->type = IPV6_ADDRESS;
- ADDRESS_IPV6_IN6_ADDR (ip) = sa6->sin6_addr;
+ struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)&storage;
+ ip->data.d6 = sa6->sin6_addr;
#ifdef HAVE_SOCKADDR_IN6_SCOPE_ID
- ADDRESS_IPV6_SCOPE (ip) = sa6->sin6_scope_id;
+ ip->ipv6_scope = sa6->sin6_scope_id;
#endif
- DEBUGP (("conaddr is: %s\n", pretty_print_address (ip)));
- return 1;
+ DEBUGP (("conaddr is: %s\n", print_address (ip)));
+ return true;
}
#endif
case AF_INET:
{
- struct sockaddr_in *sa = (struct sockaddr_in *)&storage;
- ip->type = IPV4_ADDRESS;
- ADDRESS_IPV4_IN_ADDR (ip) = sa->sin_addr;
- DEBUGP (("conaddr is: %s\n", pretty_print_address (ip)));
- return 1;
+ struct sockaddr_in *sa = (struct sockaddr_in *)&storage;
+ ip->data.d4 = sa->sin_addr;
+ DEBUGP (("conaddr is: %s\n", print_address (ip)));
+ return true;
}
default:
abort ();
}
-
- return 0;
}
-/* Return non-zero of the errno code passed to the function is a
- result of an attempt to create a socket of unsupported family. */
+/* Get the socket family of connection on FD and store
+ Return family type on success, -1 otherwise.
+
+ If ENDPOINT is ENDPOINT_LOCAL, it returns the sock family of the local
+ (client) side of the socket. Else if ENDPOINT is ENDPOINT_PEER, it
+ returns the sock family of the remote (peer's) side of the socket. */
int
-unsupported_socket_family_error (int err)
+socket_family (int sock, int endpoint)
+{
+ struct sockaddr_storage storage;
+ struct sockaddr *sockaddr = (struct sockaddr *) &storage;
+ socklen_t addrlen = sizeof (storage);
+ int ret;
+
+ memset (sockaddr, 0, addrlen);
+
+ if (endpoint == ENDPOINT_LOCAL)
+ ret = getsockname (sock, sockaddr, &addrlen);
+ else if (endpoint == ENDPOINT_PEER)
+ ret = getpeername (sock, sockaddr, &addrlen);
+ else
+ abort ();
+
+ if (ret < 0)
+ return -1;
+
+ return sockaddr->sa_family;
+}
+
+/* Return true if the error from the connect code can be considered
+ retryable. Wget normally retries after errors, but the exception
+ are the "unsupported protocol" type errors (possible on IPv4/IPv6
+ dual family systems) and "connection refused". */
+
+bool
+retryable_socket_connect_error (int err)
{
/* Have to guard against some of these values not being defined.
- Cannot use switch because some of the values might be equal. */
+ Cannot use a switch statement because some of the values might be
+ equal. */
+ if (false
#ifdef EAFNOSUPPORT
- if (err == EAFNOSUPPORT) return 1;
+ || err == EAFNOSUPPORT
#endif
#ifdef EPFNOSUPPORT
- if (err == EPFNOSUPPORT) return 1;
+ || err == EPFNOSUPPORT
#endif
-#ifdef ESOCKTNOSUPPORT /* no, "sockt" is not a typo! */
- if (err == ESOCKTNOSUPPORT) return 1;
+#ifdef ESOCKTNOSUPPORT /* no, "sockt" is not a typo! */
+ || err == ESOCKTNOSUPPORT
#endif
#ifdef EPROTONOSUPPORT
- if (err == EPROTONOSUPPORT) return 1;
+ || err == EPROTONOSUPPORT
#endif
#ifdef ENOPROTOOPT
- if (err == ENOPROTOOPT) return 1;
+ || err == ENOPROTOOPT
+#endif
+ /* Apparently, older versions of Linux and BSD used EINVAL
+ instead of EAFNOSUPPORT and such. */
+ || err == EINVAL
+ )
+ return false;
+
+ if (!opt.retry_connrefused)
+ if (err == ECONNREFUSED
+#ifdef ENETUNREACH
+ || err == ENETUNREACH /* network is unreachable */
#endif
- /* Apparently, older versions of Linux used EINVAL instead of
- EAFNOSUPPORT. */
- if (err == EINVAL) return 1;
+#ifdef EHOSTUNREACH
+ || err == EHOSTUNREACH /* host is unreachable */
+#endif
+ )
+ return false;
- return 0;
+ return true;
}
-#ifdef HAVE_SELECT
-
-/* Wait for file descriptor FD to be readable or writable or both,
- timing out after MAXTIME seconds. Returns 1 if FD is available, 0
- for timeout and -1 for error. The argument WAIT_FOR can be a
- combination of WAIT_READ and WAIT_WRITE.
+/* Wait for a single descriptor to become available, timing out after
+ MAXTIME seconds. Returns 1 if FD is available, 0 for timeout and
+ -1 for error. The argument WAIT_FOR can be a combination of
+ WAIT_FOR_READ and WAIT_FOR_WRITE.
This is a mere convenience wrapper around the select call, and
- should be taken as such. */
+ should be taken as such (for example, it doesn't implement Wget's
+ 0-timeout-means-no-timeout semantics.) */
int
select_fd (int fd, double maxtime, int wait_for)
wr = &fdset;
tmout.tv_sec = (long) maxtime;
- tmout.tv_usec = 1000000L * (maxtime - (long) maxtime);
+ tmout.tv_usec = 1000000 * (maxtime - (long) maxtime);
do
+ {
result = select (fd + 1, rd, wr, NULL, &tmout);
+#ifdef WINDOWS
+ /* gnulib select() converts blocking sockets to nonblocking in windows.
+ wget uses blocking sockets so we must convert them back to blocking. */
+ set_windows_fd_as_blocking_socket (fd);
+#endif
+ }
while (result < 0 && errno == EINTR);
return result;
}
-#endif /* HAVE_SELECT */
-\f
-/* Basic socket operations, mostly EINTR wrappers. */
+/* Return true iff the connection to the remote site established
+ through SOCK is still open.
+
+ Specifically, this function returns true if SOCK is not ready for
+ reading. This is because, when the connection closes, the socket
+ is ready for reading because EOF is about to be delivered. A side
+ effect of this method is that sockets that have pending data are
+ considered non-open. This is actually a good thing for callers of
+ this function, where such pending data can only be unwanted
+ leftover from a previous request. */
+
+bool
+test_socket_open (int sock)
+{
+ fd_set check_set;
+ struct timeval to;
+ int ret = 0;
+
+ /* Check if we still have a valid (non-EOF) connection. From Andrew
+ * Maholski's code in the Unix Socket FAQ. */
+
+ FD_ZERO (&check_set);
+ FD_SET (sock, &check_set);
+ /* Wait one microsecond */
+ to.tv_sec = 0;
+ to.tv_usec = 1;
+
+ ret = select (sock + 1, &check_set, NULL, NULL, &to);
#ifdef WINDOWS
-# define read(fd, buf, cnt) recv (fd, buf, cnt, 0)
-# define write(fd, buf, cnt) send (fd, buf, cnt, 0)
-# define close(fd) closesocket (fd)
+/* gnulib select() converts blocking sockets to nonblocking in windows.
+wget uses blocking sockets so we must convert them back to blocking
+*/
+ set_windows_fd_as_blocking_socket ( sock );
#endif
-#ifdef __BEOS__
-# define read(fd, buf, cnt) recv (fd, buf, cnt, 0)
-# define write(fd, buf, cnt) send (fd, buf, cnt, 0)
-#endif
+ if ( !ret )
+ /* We got a timeout, it means we're still connected. */
+ return true;
+ else
+ /* Read now would not wait, it means we have either pending data
+ or EOF/error. */
+ return false;
+}
+\f
+/* Basic socket operations, mostly EINTR wrappers. */
static int
sock_read (int fd, char *buf, int bufsize)
static int
sock_write (int fd, char *buf, int bufsize)
{
- int res = 0;
+ int res;
do
res = write (fd, buf, bufsize);
while (res == -1 && errno == EINTR);
static int
sock_poll (int fd, double timeout, int wait_for)
{
-#ifdef HAVE_SELECT
return select_fd (fd, timeout, wait_for);
-#else
- return 1;
-#endif
+}
+
+static int
+sock_peek (int fd, char *buf, int bufsize)
+{
+ int res;
+ do
+ res = recv (fd, buf, bufsize, MSG_PEEK);
+ while (res == -1 && errno == EINTR);
+ return res;
}
static void
that are not mere file descriptors under the hood, such as SSL
sockets.
- That way the user code can call xread(fd, ...) and we'll run read
+ That way the user code can call fd_read(fd, ...) and we'll run read
or SSL_read or whatever is necessary. */
-static struct hash_table *extended_map;
-static int extended_map_modified_tick;
+static struct hash_table *transport_map;
+static unsigned int transport_map_modified_tick;
-struct extended_info {
- xreader_t reader;
- xwriter_t writer;
- xpoller_t poller;
- xcloser_t closer;
+struct transport_info {
+ struct transport_implementation *imp;
void *ctx;
};
-/* Register the handlers for operations on FD. This is meant
- primarily for transport layers like SSL that piggyback on sockets,
- but with their own readers, writers, etc. */
+/* Register the transport layer operations that will be used when
+ reading, writing, and polling FD.
+
+ This should be used for transport layers like SSL that piggyback on
+ sockets. FD should otherwise be a real socket, on which you can
+ call getpeername, etc. */
void
-register_extended (int fd, xreader_t reader, xwriter_t writer,
- xpoller_t poller, xcloser_t closer, void *ctx)
+fd_register_transport (int fd, struct transport_implementation *imp, void *ctx)
{
- struct extended_info *info;
+ struct transport_info *info;
/* The file descriptor must be non-negative to be registered.
- Negative values are ignored by xclose(), and -1 cannot be used as
+ Negative values are ignored by fd_close(), and -1 cannot be used as
hash key. */
assert (fd >= 0);
- info = xnew (struct extended_info);
- info->reader = reader;
- info->writer = writer;
- info->poller = poller;
- info->closer = closer;
+ info = xnew (struct transport_info);
+ info->imp = imp;
info->ctx = ctx;
- if (!extended_map)
- extended_map = hash_table_new (0, NULL, NULL);
- hash_table_put (extended_map, (void *) fd, info);
- ++extended_map_modified_tick;
+ if (!transport_map)
+ transport_map = hash_table_new (0, NULL, NULL);
+ hash_table_put (transport_map, (void *)(intptr_t) fd, info);
+ ++transport_map_modified_tick;
+}
+
+/* Return context of the transport registered with
+ fd_register_transport. This assumes fd_register_transport was
+ previously called on FD. */
+
+void *
+fd_transport_context (int fd)
+{
+ struct transport_info *info = hash_table_get (transport_map, (void *)(intptr_t) fd);
+ return info->ctx;
}
-/* When xread/xwrite are called multiple times in a loop, they should
+/* When fd_read/fd_write are called multiple times in a loop, they should
remember the INFO pointer instead of fetching it every time. It is
not enough to compare FD to LAST_FD because FD might have been
closed and reopened. modified_tick ensures that changes to
- extended_map will not be unnoticed.
+ transport_map will not be unnoticed.
This is a macro because we want the static storage variables to be
per-function. */
-#define LAZY_RETRIEVE_INFO(info) do { \
- static struct extended_info *last_info; \
- static int last_fd = -1, last_tick; \
- if (!extended_map) \
- info = NULL; \
- else if (last_fd == fd && last_tick == extended_map_modified_tick) \
- info = last_info; \
- else \
- { \
- info = hash_table_get (extended_map, (void *) fd); \
- last_fd = fd; \
- last_info = info; \
- last_tick = extended_map_modified_tick; \
- } \
+#define LAZY_RETRIEVE_INFO(info) do { \
+ static struct transport_info *last_info; \
+ static int last_fd = -1; \
+ static unsigned int last_tick; \
+ if (!transport_map) \
+ info = NULL; \
+ else if (last_fd == fd && last_tick == transport_map_modified_tick) \
+ info = last_info; \
+ else \
+ { \
+ info = hash_table_get (transport_map, (void *)(intptr_t) fd); \
+ last_fd = fd; \
+ last_info = info; \
+ last_tick = transport_map_modified_tick; \
+ } \
} while (0)
-/* Read no more than BUFSIZE bytes of data from FD, storing them to
- BUF. If TIMEOUT is non-zero, the operation aborts if no data is
- received after that many seconds. If TIMEOUT is -1, the value of
- opt.timeout is used for TIMEOUT. */
-
-int
-xread (int fd, char *buf, int bufsize, double timeout)
+static bool
+poll_internal (int fd, struct transport_info *info, int wf, double timeout)
{
- struct extended_info *info;
- LAZY_RETRIEVE_INFO (info);
if (timeout == -1)
timeout = opt.read_timeout;
if (timeout)
{
int test;
- if (info && info->poller)
- test = info->poller (fd, timeout, WAIT_FOR_READ, info->ctx);
+ if (info && info->imp->poller)
+ test = info->imp->poller (fd, timeout, wf, info->ctx);
else
- test = sock_poll (fd, timeout, WAIT_FOR_READ);
+ test = sock_poll (fd, timeout, wf);
if (test == 0)
- errno = ETIMEDOUT;
+ errno = ETIMEDOUT;
if (test <= 0)
- return -1;
+ return false;
}
- if (info && info->reader)
- return info->reader (fd, buf, bufsize, info->ctx);
+ return true;
+}
+
+/* Read no more than BUFSIZE bytes of data from FD, storing them to
+ BUF. If TIMEOUT is non-zero, the operation aborts if no data is
+ received after that many seconds. If TIMEOUT is -1, the value of
+ opt.timeout is used for TIMEOUT. */
+
+int
+fd_read (int fd, char *buf, int bufsize, double timeout)
+{
+ struct transport_info *info;
+ LAZY_RETRIEVE_INFO (info);
+ if (!poll_internal (fd, info, WAIT_FOR_READ, timeout))
+ return -1;
+ if (info && info->imp->reader)
+ return info->imp->reader (fd, buf, bufsize, info->ctx);
else
return sock_read (fd, buf, bufsize);
}
+/* Like fd_read, except it provides a "preview" of the data that will
+ be read by subsequent calls to fd_read. Specifically, it copies no
+ more than BUFSIZE bytes of the currently available data to BUF and
+ returns the number of bytes copied. Return values and timeout
+ semantics are the same as those of fd_read.
+
+ CAVEAT: Do not assume that the first subsequent call to fd_read
+ will retrieve the same amount of data. Reading can return more or
+ less data, depending on the TCP implementation and other
+ circumstances. However, barring an error, it can be expected that
+ all the peeked data will eventually be read by fd_read. */
+
+int
+fd_peek (int fd, char *buf, int bufsize, double timeout)
+{
+ struct transport_info *info;
+ LAZY_RETRIEVE_INFO (info);
+ if (!poll_internal (fd, info, WAIT_FOR_READ, timeout))
+ return -1;
+ if (info && info->imp->peeker)
+ return info->imp->peeker (fd, buf, bufsize, info->ctx);
+ else
+ return sock_peek (fd, buf, bufsize);
+}
+
/* Write the entire contents of BUF to FD. If TIMEOUT is non-zero,
the operation aborts if no data is received after that many
seconds. If TIMEOUT is -1, the value of opt.timeout is used for
TIMEOUT. */
int
-xwrite (int fd, char *buf, int bufsize, double timeout)
+fd_write (int fd, char *buf, int bufsize, double timeout)
{
int res;
- struct extended_info *info;
+ struct transport_info *info;
LAZY_RETRIEVE_INFO (info);
- if (timeout == -1)
- timeout = opt.read_timeout;
/* `write' may write less than LEN bytes, thus the loop keeps trying
it until all was written, or an error occurred. */
res = 0;
while (bufsize > 0)
{
- if (timeout)
- {
- int test;
- if (info && info->poller)
- test = info->poller (fd, timeout, WAIT_FOR_WRITE, info->ctx);
- else
- test = sock_poll (fd, timeout, WAIT_FOR_WRITE);
- if (test == 0)
- errno = ETIMEDOUT;
- if (test <= 0)
- return -1;
- }
- if (info && info->writer)
- res = info->writer (fd, buf, bufsize, info->ctx);
+ if (!poll_internal (fd, info, WAIT_FOR_WRITE, timeout))
+ return -1;
+ if (info && info->imp->writer)
+ res = info->imp->writer (fd, buf, bufsize, info->ctx);
else
- res = sock_write (fd, buf, bufsize);
+ res = sock_write (fd, buf, bufsize);
if (res <= 0)
- break;
+ break;
buf += res;
bufsize -= res;
}
return res;
}
+/* Report the most recent error(s) on FD. This should only be called
+ after fd_* functions, such as fd_read and fd_write, and only if
+ they return a negative result. For errors coming from other calls
+ such as setsockopt or fopen, strerror should continue to be
+ used.
+
+ If the transport doesn't support error messages or doesn't supply
+ one, strerror(errno) is returned. The returned error message
+ should not be used after fd_close has been called. */
+
+const char *
+fd_errstr (int fd)
+{
+ /* Don't bother with LAZY_RETRIEVE_INFO, as this will only be called
+ in case of error, never in a tight loop. */
+ struct transport_info *info = NULL;
+ if (transport_map)
+ info = hash_table_get (transport_map, (void *)(intptr_t) fd);
+
+ if (info && info->imp->errstr)
+ {
+ const char *err = info->imp->errstr (fd, info->ctx);
+ if (err)
+ return err;
+ /* else, fall through and print the system error. */
+ }
+ return strerror (errno);
+}
+
/* Close the file descriptor FD. */
void
-xclose (int fd)
+fd_close (int fd)
{
- struct extended_info *info;
+ struct transport_info *info;
if (fd < 0)
return;
- /* We don't need to be extra-fast here, so save some code by
- avoiding LAZY_RETRIEVE_INFO. */
+ /* Don't use LAZY_RETRIEVE_INFO because fd_close() is only called once
+ per socket, so that particular optimization wouldn't work. */
info = NULL;
- if (extended_map)
- info = hash_table_get (extended_map, (void *) fd);
+ if (transport_map)
+ info = hash_table_get (transport_map, (void *)(intptr_t) fd);
- if (info && info->closer)
- info->closer (fd, info->ctx);
+ if (info && info->imp->closer)
+ info->imp->closer (fd, info->ctx);
else
sock_close (fd);
if (info)
{
- hash_table_remove (extended_map, (void *) fd);
+ hash_table_remove (transport_map, (void *)(intptr_t) fd);
xfree (info);
- ++extended_map_modified_tick;
+ ++transport_map_modified_tick;
}
}
projects / wget / blobdiff