[Qemu-devel] [PATCH v6 0/5] Do not set SO_REUSEADDR on Windows

[Qemu-devel] [PATCH v6 0/5] Do not set SO_REUSEADDR on Windows

[Sebastian Ottlik]

This patchset disables most uses of SO_REUSEADDR on Windows and replaces it with
calls to the new function socket_set_fast_reuse. On Windows systems the default
behaviour is equivalent to SO_REUSEADDR on other operating systems. SO_REUSEADDR
can still be set but results in undesired behaviour in most cases. It may even
lead to situations were system behaviour is unspecified. More information on
this can be found at:
http://msdn.microsoft.com/en-us/library/windows/desktop/ms740621.aspx

I originally encountered this issue when accidentally launching two QEMU
instances with identical GDB ports at the same time. In which case QEMU won't
fail as one might expect.

Note that patch #4 fails checkpatch.pl. This is intentional (see v3 changes).

v6 Changes:
- dropped error output and the silent parameter in favor of an assertion

  Actually I wanted to remove the return value from the function too, as the
  assertion pretty much states that the function will not fail and thus always
  return 0. However this would make the code a little ugly to prevent unused
  variable warnings if NDEBUG is set (see patch 1) and also would require some
  ugly changes to slirp/socket.c (see patch 4). Thus I decided to keep it.

- Rebased to current master (a684f3cf9b9b9c3cb82be87aafc463de8974610c)

v5 Changes:
- Fixed inverted silent flag on all call sites of socket_set_fast_reuse

v4 Changes:
- Added the silent flag to socket_set_fast_reuse controlling error reporting
  One location where SO_REUSEADDR was set would report errors if setting the
  option failed. Keeping the reporting code there would be somewhat unclean, so
  I moved it to socket_set_fast_reuse. A side effect of this was that the error
  reporting was added for all locations that now use socket_set_fast_reuse. Here
  a new flag is added to control error reporting, which means this patchset
  won't change QEMU behaviour (except for not setting SO_REUSEADDR on Windows).

- Fixed a commit message typo

- Rebased to current master (2d1fe1873a984d1c2c89ffa3d12949cafc718551).

v3 Changes:
- Fixed coding style issues.
  According to checkpatch.pl patch #4 still introduces style errors as tabs are
  used instead of space for some indentation. I keept the tabs to stay
  consistent with the sourrounding code, as tabs seem to be used consitently in
  parts (all?) of the slirp code.

- Changed patch #3 to keep SO_REUSEADDR for multicast sockets on windows and
  added an explainatory comment.

- Rebased to current master (94c2b6aff43cdfcfdfb552773a6b6b973a72ef0b).

v2 Changes:

- Introduce a function with os specific implementation instead of using #ifdef
  I named it socket_set_fast_reuse instead of the suggested qemu_set_reuseaddr
  so the name better reflects what the function actually does.

gdbstub.c              |    6 ++----
include/qemu/sockets.h |    1 +
net/socket.c           |   19 ++++++++++---------
slirp/misc.c           |    3 +--
slirp/socket.c         |    4 +---
slirp/tcp_subr.c       |    6 ++----
slirp/udp.c            |    4 ++--
util/oslib-posix.c     |   12 ++++++++++++
util/oslib-win32.c     |   10 ++++++++++
util/qemu-sockets.c    |    6 +++---
10 files changed, 44 insertions(+), 27 deletions(-)

util: add socket_set_fast_reuse function which will
gdbstub: call socket_set_fast_reuse instead of
net: call socket_set_fast_reuse instead of setting
slirp: call socket_set_fast_reuse instead of setting
util: call socket_set_fast_reuse instead of setting

[Qemu-devel] [PATCH v6 1/5] util: add socket_set_fast_reuse function which will replace setting SO_REUSEADDR

[Sebastian Ottlik]

If a socket is closed it remains in TIME_WAIT state for some time. On operating
systems using BSD sockets the endpoint of the socket may not be reused while in
this state unless SO_REUSEADDR was set on the socket. On windows on the other
hand the default behaviour is to allow reuse (i.e. identical to SO_REUSEADDR on
other operating systems) and setting SO_REUSEADDR on a socket allows it to be
bound to a endpoint even if the endpoint is already used by another socket
independently of the other sockets state. This can even result in undefined
behaviour.

Many sockets used by QEMU should not block the use of their endpoint after being
closed while they are still in TIME_WAIT state. Currently QEMU sets SO_REUSEADDR
for such sockets, which can lead to problems on Windows. This patch introduces
the function socket_set_fast_reuse that should be used instead of setting
SO_REUSEADDR when fast socket reuse is desired and behaves correctly on all
operating systems.

As a failure of this function can only be caused by bad QEMU internal errors, an
assertion handles these situations. The return value is still passed on, to
minimize changes in client code and prevent unused variable warnings if NDEBUG
is defined.

Signed-off-by: Sebastian Ottlik <ottlik@fzi.de>
---
 include/qemu/sockets.h |    1 +
 util/oslib-posix.c     |   12 ++++++++++++
 util/oslib-win32.c     |   10 ++++++++++
 3 files changed, 23 insertions(+)

diff --git a/include/qemu/sockets.h b/include/qemu/sockets.h
index c5174d7..45588d7 100644
--- a/include/qemu/sockets.h
+++ b/include/qemu/sockets.h
@@ -39,6 +39,7 @@ int socket_set_cork(int fd, int v);
 int socket_set_nodelay(int fd);
 void qemu_set_block(int fd);
 void qemu_set_nonblock(int fd);
+int socket_set_fast_reuse(int fd);
 int send_all(int fd, const void *buf, int len1);
 int recv_all(int fd, void *buf, int len1, bool single_read);
 
diff --git a/util/oslib-posix.c b/util/oslib-posix.c
index 253bc3d..e00a44c 100644
--- a/util/oslib-posix.c
+++ b/util/oslib-posix.c
@@ -157,6 +157,18 @@ void qemu_set_nonblock(int fd)
     fcntl(fd, F_SETFL, f | O_NONBLOCK);
 }
 
+int socket_set_fast_reuse(int fd)
+{
+    int val = 1, ret;
+
+    ret = setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
+                     (const char *)&val, sizeof(val));
+
+    assert(ret == 0);
+
+    return ret;
+}
+
 void qemu_set_cloexec(int fd)
 {
     int f;
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index 983b7a2..776ccfa 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -124,6 +124,16 @@ void qemu_set_nonblock(int fd)
     qemu_fd_register(fd);
 }
 
+int socket_set_fast_reuse(int fd)
+{
+    /* Enabling the reuse of an endpoint that was used by a socket still in
+     * TIME_WAIT state is usually performed by setting SO_REUSEADDR. On Windows
+     * fast reuse is the default and SO_REUSEADDR does strange things. So we
+     * don't have to do anything here. More info can be found at:
+     * http://msdn.microsoft.com/en-us/library/windows/desktop/ms740621.aspx */
+    return 0;
+}
+
 int inet_aton(const char *cp, struct in_addr *ia)
 {
     uint32_t addr = inet_addr(cp);
-- 
1.7.9.5

[Qemu-devel] [PATCH v6 2/5] gdbstub: call socket_set_fast_reuse instead of setting SO_REUSEADDR

[Sebastian Ottlik]

SO_REUSEADDR should be avoided on Windows but is desired on other operating
systems. So instead of setting it we call socket_set_fast_reuse that will result
in the appropriate behaviour on all operating systems.

Signed-off-by: Sebastian Ottlik <ottlik@fzi.de>
---
 gdbstub.c |    6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/gdbstub.c b/gdbstub.c
index 2b7f22b..0e5a3f5 100644
--- a/gdbstub.c
+++ b/gdbstub.c
@@ -1553,7 +1553,7 @@ static void gdb_accept(void)
 static int gdbserver_open(int port)
 {
     struct sockaddr_in sockaddr;
-    int fd, val, ret;
+    int fd, ret;
 
     fd = socket(PF_INET, SOCK_STREAM, 0);
     if (fd < 0) {
@@ -1564,9 +1564,7 @@ static int gdbserver_open(int port)
     fcntl(fd, F_SETFD, FD_CLOEXEC);
 #endif
 
-    /* allow fast reuse */
-    val = 1;
-    qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
+    socket_set_fast_reuse(fd);
 
     sockaddr.sin_family = AF_INET;
     sockaddr.sin_port = htons(port);
-- 
1.7.9.5

[Qemu-devel] [PATCH v6 3/5] net: call socket_set_fast_reuse instead of setting SO_REUSEADDR

[Sebastian Ottlik]

SO_REUSEADDR should be avoided on Windows but is desired on other operating
systems. So instead of setting it we call socket_set_fast_reuse that will result
in the appropriate behaviour on all operating systems.

An exception to this rule are multicast sockets where it is sensible to have
multiple sockets listen on the same ip and port and we should set SO_REUSEADDR
on windows.

Signed-off-by: Sebastian Ottlik <ottlik@fzi.de>
---
 net/socket.c |   19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/net/socket.c b/net/socket.c
index e61309d..fb21e20 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -262,6 +262,11 @@ static int net_socket_mcast_create(struct sockaddr_in *mcastaddr, struct in_addr
         return -1;
     }
 
+    /* Allow multiple sockets to bind the same multicast ip and port by setting
+     * SO_REUSEADDR. This is the only situation where SO_REUSEADDR should be set
+     * on windows. Use socket_set_fast_reuse otherwise as it sets SO_REUSEADDR
+     * only on posix systems.
+     */
     val = 1;
     ret = qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
     if (ret < 0) {
@@ -510,7 +515,7 @@ static int net_socket_listen_init(NetClientState *peer,
     NetClientState *nc;
     NetSocketState *s;
     struct sockaddr_in saddr;
-    int fd, val, ret;
+    int fd, ret;
 
     if (parse_host_port(&saddr, host_str) < 0)
         return -1;
@@ -522,9 +527,7 @@ static int net_socket_listen_init(NetClientState *peer,
     }
     qemu_set_nonblock(fd);
 
-    /* allow fast reuse */
-    val = 1;
-    qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
+    socket_set_fast_reuse(fd);
 
     ret = bind(fd, (struct sockaddr *)&saddr, sizeof(saddr));
     if (ret < 0) {
@@ -645,7 +648,7 @@ static int net_socket_udp_init(NetClientState *peer,
                                  const char *lhost)
 {
     NetSocketState *s;
-    int fd, val, ret;
+    int fd, ret;
     struct sockaddr_in laddr, raddr;
 
     if (parse_host_port(&laddr, lhost) < 0) {
@@ -661,11 +664,9 @@ static int net_socket_udp_init(NetClientState *peer,
         perror("socket(PF_INET, SOCK_DGRAM)");
         return -1;
     }
-    val = 1;
-    ret = qemu_setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
-                          &val, sizeof(val));
+
+    ret = socket_set_fast_reuse(fd);
     if (ret < 0) {
-        perror("setsockopt(SOL_SOCKET, SO_REUSEADDR)");
         closesocket(fd);
         return -1;
     }
-- 
1.7.9.5

[Qemu-devel] [PATCH v6 4/5] slirp: call socket_set_fast_reuse instead of setting SO_REUSEADDR

[Sebastian Ottlik]

SO_REUSEADDR should be avoided on Windows but is desired on other operating
systems. So instead of setting it we call socket_set_fast_reuse that will result
in the appropriate behaviour on all operating systems.

Signed-off-by: Sebastian Ottlik <ottlik@fzi.de>
---
 slirp/misc.c     |    3 +--
 slirp/socket.c   |    4 +---
 slirp/tcp_subr.c |    6 ++----
 slirp/udp.c      |    4 ++--
 4 files changed, 6 insertions(+), 11 deletions(-)

diff --git a/slirp/misc.c b/slirp/misc.c
index c0d4899..6c1636f 100644
--- a/slirp/misc.c
+++ b/slirp/misc.c
@@ -212,8 +212,7 @@ fork_exec(struct socket *so, const char *ex, int do_pty)
                     so->s = accept(s, (struct sockaddr *)&addr, &addrlen);
                 } while (so->s < 0 && errno == EINTR);
                 closesocket(s);
-                opt = 1;
-                qemu_setsockopt(so->s, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof(int));
+                socket_set_fast_reuse(so->s);
                 opt = 1;
                 qemu_setsockopt(so->s, SOL_SOCKET, SO_OOBINLINE, &opt, sizeof(int));
 		qemu_set_nonblock(so->s);
diff --git a/slirp/socket.c b/slirp/socket.c
index 25d60e7..37ac5cf 100644
--- a/slirp/socket.c
+++ b/slirp/socket.c
@@ -627,9 +627,7 @@ tcp_listen(Slirp *slirp, uint32_t haddr, u_int hport, uint32_t laddr,
 	addr.sin_port = hport;
 
 	if (((s = qemu_socket(AF_INET,SOCK_STREAM,0)) < 0) ||
-#ifndef _WIN32
-	    (qemu_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof(int)) < 0) ||
-#endif
+	    (socket_set_fast_reuse(s) < 0) ||
 	    (bind(s,(struct sockaddr *)&addr, sizeof(addr)) < 0) ||
 	    (listen(s,1) < 0)) {
 		int tmperrno = errno; /* Don't clobber the real reason we failed */
diff --git a/slirp/tcp_subr.c b/slirp/tcp_subr.c
index 043f28f..7571c5a 100644
--- a/slirp/tcp_subr.c
+++ b/slirp/tcp_subr.c
@@ -337,8 +337,7 @@ int tcp_fconnect(struct socket *so)
     struct sockaddr_in addr;
 
     qemu_set_nonblock(s);
-    opt = 1;
-    qemu_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof(opt));
+    socket_set_fast_reuse(s);
     opt = 1;
     qemu_setsockopt(s, SOL_SOCKET, SO_OOBINLINE, &opt, sizeof(opt));
 
@@ -426,8 +425,7 @@ void tcp_connect(struct socket *inso)
         return;
     }
     qemu_set_nonblock(s);
-    opt = 1;
-    qemu_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof(int));
+    socket_set_fast_reuse(s);
     opt = 1;
     qemu_setsockopt(s, SOL_SOCKET, SO_OOBINLINE, &opt, sizeof(int));
     socket_set_nodelay(s);
diff --git a/slirp/udp.c b/slirp/udp.c
index b105f87..8cc6cb6 100644
--- a/slirp/udp.c
+++ b/slirp/udp.c
@@ -354,7 +354,7 @@ udp_listen(Slirp *slirp, uint32_t haddr, u_int hport, uint32_t laddr,
 {
 	struct sockaddr_in addr;
 	struct socket *so;
-	socklen_t addrlen = sizeof(struct sockaddr_in), opt = 1;
+	socklen_t addrlen = sizeof(struct sockaddr_in);
 
 	so = socreate(slirp);
 	if (!so) {
@@ -372,7 +372,7 @@ udp_listen(Slirp *slirp, uint32_t haddr, u_int hport, uint32_t laddr,
 		udp_detach(so);
 		return NULL;
 	}
-	qemu_setsockopt(so->s, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof(int));
+	socket_set_fast_reuse(so->s);
 
 	getsockname(so->s,(struct sockaddr *)&addr,&addrlen);
 	so->so_fport = addr.sin_port;
-- 
1.7.9.5

[Qemu-devel] [PATCH v6 5/5] util: call socket_set_fast_reuse instead of setting SO_REUSEADDR

[Sebastian Ottlik]

SO_REUSEADDR should be avoided on Windows but is desired on other operating
systems. So instead of setting it we call socket_set_fast_reuse that will result
in the appropriate behaviour on all operating systems.

Signed-off-by: Sebastian Ottlik <ottlik@fzi.de>
---
 util/qemu-sockets.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/util/qemu-sockets.c b/util/qemu-sockets.c
index 095716e..6b97dc1 100644
--- a/util/qemu-sockets.c
+++ b/util/qemu-sockets.c
@@ -155,7 +155,7 @@ int inet_listen_opts(QemuOpts *opts, int port_offset, Error **errp)
             continue;
         }
 
-        qemu_setsockopt(slisten, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on));
+        socket_set_fast_reuse(slisten);
 #ifdef IPV6_V6ONLY
         if (e->ai_family == PF_INET6) {
             /* listen on both ipv4 and ipv6 */
@@ -274,7 +274,7 @@ static int inet_connect_addr(struct addrinfo *addr, bool *in_progress,
         error_set_errno(errp, errno, QERR_SOCKET_CREATE_FAILED);
         return -1;
     }
-    qemu_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on));
+    socket_set_fast_reuse(sock);
     if (connect_state != NULL) {
         qemu_set_nonblock(sock);
     }
@@ -455,7 +455,7 @@ int inet_dgram_opts(QemuOpts *opts, Error **errp)
         error_set_errno(errp, errno, QERR_SOCKET_CREATE_FAILED);
         goto err;
     }
-    qemu_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on));
+    socket_set_fast_reuse(sock);
 
     /* bind socket */
     if (bind(sock, local->ai_addr, local->ai_addrlen) < 0) {
-- 
1.7.9.5

Re: [Qemu-devel] [PATCH v6 0/5] Do not set SO_REUSEADDR on Windows

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 1584 bytes --]

On 10/02/2013 04:23 AM, Sebastian Ottlik wrote:
> This patchset disables most uses of SO_REUSEADDR on Windows and replaces it with
> calls to the new function socket_set_fast_reuse. On Windows systems the default
> behaviour is equivalent to SO_REUSEADDR on other operating systems. SO_REUSEADDR
> can still be set but results in undesired behaviour in most cases. It may even
> lead to situations were system behaviour is unspecified. More information on
> this can be found at:
> http://msdn.microsoft.com/en-us/library/windows/desktop/ms740621.aspx
> 
> I originally encountered this issue when accidentally launching two QEMU
> instances with identical GDB ports at the same time. In which case QEMU won't
> fail as one might expect.
> 
> Note that patch #4 fails checkpatch.pl. This is intentional (see v3 changes).
> 
> v6 Changes:
> - dropped error output and the silent parameter in favor of an assertion
> 
>   Actually I wanted to remove the return value from the function too, as the
>   assertion pretty much states that the function will not fail and thus always
>   return 0. However this would make the code a little ugly to prevent unused
>   variable warnings if NDEBUG is set (see patch 1) and also would require some
>   ugly changes to slirp/socket.c (see patch 4). Thus I decided to keep it.
> 
> - Rebased to current master (a684f3cf9b9b9c3cb82be87aafc463de8974610c)
> 

Series: Reviewed-by: Eric Blake <eblake@redhat.com>

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 621 bytes --]

Re: [Qemu-devel] [PATCH v6 0/5] Do not set SO_REUSEADDR on Windows

[Stefan Weil]

Am 02.10.2013 14:28, schrieb Eric Blake:
> On 10/02/2013 04:23 AM, Sebastian Ottlik wrote:
>> This patchset disables most uses of SO_REUSEADDR on Windows and
replaces it with
>> calls to the new function socket_set_fast_reuse. On Windows systems
the default
>> behaviour is equivalent to SO_REUSEADDR on other operating systems.
SO_REUSEADDR
>> can still be set but results in undesired behaviour in most cases. It
may even
>> lead to situations were system behaviour is unspecified. More
information on
>> this can be found at:
>> http://msdn.microsoft.com/en-us/library/windows/desktop/ms740621.aspx
>>
>> I originally encountered this issue when accidentally launching two QEMU
>> instances with identical GDB ports at the same time. In which case
QEMU won't
>> fail as one might expect.
>>
>> Note that patch #4 fails checkpatch.pl. This is intentional (see v3
changes).
>>
>> v6 Changes:
>> - dropped error output and the silent parameter in favor of an assertion
>>
>>   Actually I wanted to remove the return value from the function too,
as the
>>   assertion pretty much states that the function will not fail and
thus always
>>   return 0. However this would make the code a little ugly to prevent
unused
>>   variable warnings if NDEBUG is set (see patch 1) and also would
require some
>>   ugly changes to slirp/socket.c (see patch 4). Thus I decided to
keep it.
>>
>> - Rebased to current master (a684f3cf9b9b9c3cb82be87aafc463de8974610c)
>>
>
> Series: Reviewed-by: Eric Blake <eblake@redhat.com>

Thanks for the patches and for the review.

I applied all to my mingw patch queue and have sent a pull request.

Stefan