[PATCH v3 00/16] win32: do not mix SOCKET and fd space

[PATCH v3 00/16] win32: do not mix SOCKET and fd space

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Hi,

A win32 SOCKET handle is often cast to an int file descriptor, as this is what
other OS use for sockets. When necessary, QEMU eventually queries whether it's a
socket with the help of fd_is_socket(). However, there is no guarantee of
conflict between the fd and SOCKET space. Such conflict would have surprising
consequences. We can fix this by using FDs only.

After fixing a few missed closesocket(), this patch series makes the win32
socket API wrappers take FDs. It finally get rid of closesocket() usage by using
a close() wrapper instead. (note that fdopen/fclose would not be enough either
to close the underlying socket appropriately)

v3:
- fix closesocket() to prevent CloseHandle() from close()
- fix direct closesocket() usage (#undef closesocket before)
- add a test for &error_warn
- add r-b tags

v2:
- add clean up patch "util: drop qemu_fork()"
- add a "&error_warn", to help with basic error reporting
- fix errno handling after _get_osfhandle()
- introduce qemu_socket_(un)select() helpers
- add patch "aio_set_fd_handler() only supports SOCKET"
- add meson slirp.wrap RFC
- various misc cleanups
- add r-b tags

Marc-André Lureau (16):
  util: drop qemu_fork()
  tests: use closesocket()
  io: use closesocket()
  tests: add test-error-report
  error: add global &error_warn destination
  win32/socket: introduce qemu_socket_select() helper
  win32/socket: introduce qemu_socket_unselect() helper
  aio: make aio_set_fd_poll() static to aio-posix.c
  aio/win32: aio_set_fd_handler() only supports SOCKET
  RFC: build-sys: add slirp.wrap
  main-loop: remove qemu_fd_register(), win32/slirp/socket specific
  slirp: unregister the win32 SOCKET
  slirp: open-code qemu_socket_(un)select()
  win32: avoid mixing SOCKET and file descriptor space
  os-posix: remove useless ioctlsocket() define
  win32: replace closesocket() with close() wrapper

 include/block/aio.h            |   8 --
 include/qapi/error.h           |   6 +
 include/qemu/main-loop.h       |   2 -
 include/qemu/osdep.h           |  14 --
 include/sysemu/os-posix.h      |   3 -
 include/sysemu/os-win32.h      |  15 ++-
 backends/tpm/tpm_emulator.c    |   6 +-
 crypto/afalg.c                 |   6 +-
 hw/hyperv/syndbg.c             |   4 +-
 io/channel-socket.c            |   8 +-
 io/channel-watch.c             |  10 +-
 net/dgram.c                    |  14 +-
 net/slirp.c                    |  16 ++-
 net/socket.c                   |  22 +--
 tests/qtest/libqtest.c         |   8 +-
 tests/qtest/microbit-test.c    |   2 +-
 tests/qtest/netdev-socket.c    |  10 +-
 tests/unit/socket-helpers.c    |   2 +-
 tests/unit/test-error-report.c | 139 +++++++++++++++++++
 util/aio-posix.c               |   6 +-
 util/aio-win32.c               |  23 ++--
 util/error.c                   |  10 +-
 util/main-loop.c               |  11 --
 util/oslib-posix.c             |  70 ----------
 util/oslib-win32.c             | 240 ++++++++++++++++++++++++++++-----
 util/qemu-sockets.c            |  22 +--
 .gitignore                     |   2 +
 subprojects/slirp.wrap         |   6 +
 tests/unit/meson.build         |   1 +
 29 files changed, 461 insertions(+), 225 deletions(-)
 create mode 100644 tests/unit/test-error-report.c
 create mode 100644 subprojects/slirp.wrap

-- 
2.39.2

[PATCH v3 01/16] util: drop qemu_fork()

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Fortunately, qemu_fork() is no longer used since commit
a95570e3e4d6 ("io/command: use glib GSpawn, instead of open-coding
fork/exec"). (GSpawn uses posix_spawn() whenever possible instead)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
---
 include/qemu/osdep.h | 14 ---------
 util/oslib-posix.c   | 70 --------------------------------------------
 util/oslib-win32.c   |  9 ------
 3 files changed, 93 deletions(-)

diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h
index 88c9facbf2..f68b5d8708 100644
--- a/include/qemu/osdep.h
+++ b/include/qemu/osdep.h
@@ -665,20 +665,6 @@ void qemu_prealloc_mem(int fd, char *area, size_t sz, int max_threads,
  */
 char *qemu_get_pid_name(pid_t pid);
 
-/**
- * qemu_fork:
- *
- * A version of fork that avoids signal handler race
- * conditions that can lead to child process getting
- * signals that are otherwise only expected by the
- * parent. It also resets all signal handlers to the
- * default settings.
- *
- * Returns 0 to child process, pid number to parent
- * or -1 on failure.
- */
-pid_t qemu_fork(Error **errp);
-
 /* Using intptr_t ensures that qemu_*_page_mask is sign-extended even
  * when intptr_t is 32-bit and we are aligning a long long.
  */
diff --git a/util/oslib-posix.c b/util/oslib-posix.c
index 77d882e681..760390b31e 100644
--- a/util/oslib-posix.c
+++ b/util/oslib-posix.c
@@ -583,76 +583,6 @@ char *qemu_get_pid_name(pid_t pid)
 }
 
 
-pid_t qemu_fork(Error **errp)
-{
-    sigset_t oldmask, newmask;
-    struct sigaction sig_action;
-    int saved_errno;
-    pid_t pid;
-
-    /*
-     * Need to block signals now, so that child process can safely
-     * kill off caller's signal handlers without a race.
-     */
-    sigfillset(&newmask);
-    if (pthread_sigmask(SIG_SETMASK, &newmask, &oldmask) != 0) {
-        error_setg_errno(errp, errno,
-                         "cannot block signals");
-        return -1;
-    }
-
-    pid = fork();
-    saved_errno = errno;
-
-    if (pid < 0) {
-        /* attempt to restore signal mask, but ignore failure, to
-         * avoid obscuring the fork failure */
-        (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
-        error_setg_errno(errp, saved_errno,
-                         "cannot fork child process");
-        errno = saved_errno;
-        return -1;
-    } else if (pid) {
-        /* parent process */
-
-        /* Restore our original signal mask now that the child is
-         * safely running. Only documented failures are EFAULT (not
-         * possible, since we are using just-grabbed mask) or EINVAL
-         * (not possible, since we are using correct arguments).  */
-        (void)pthread_sigmask(SIG_SETMASK, &oldmask, NULL);
-    } else {
-        /* child process */
-        size_t i;
-
-        /* Clear out all signal handlers from parent so nothing
-         * unexpected can happen in our child once we unblock
-         * signals */
-        sig_action.sa_handler = SIG_DFL;
-        sig_action.sa_flags = 0;
-        sigemptyset(&sig_action.sa_mask);
-
-        for (i = 1; i < NSIG; i++) {
-            /* Only possible errors are EFAULT or EINVAL The former
-             * won't happen, the latter we expect, so no need to check
-             * return value */
-            (void)sigaction(i, &sig_action, NULL);
-        }
-
-        /* Unmask all signals in child, since we've no idea what the
-         * caller's done with their signal mask and don't want to
-         * propagate that to children */
-        sigemptyset(&newmask);
-        if (pthread_sigmask(SIG_SETMASK, &newmask, NULL) != 0) {
-            Error *local_err = NULL;
-            error_setg_errno(&local_err, errno,
-                             "cannot unblock signals");
-            error_report_err(local_err);
-            _exit(1);
-        }
-    }
-    return pid;
-}
-
 void *qemu_alloc_stack(size_t *sz)
 {
     void *ptr, *guardpage;
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index 07ade41800..528c9ee156 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -283,15 +283,6 @@ char *qemu_get_pid_name(pid_t pid)
 }
 
 
-pid_t qemu_fork(Error **errp)
-{
-    errno = ENOSYS;
-    error_setg_errno(errp, errno,
-                     "cannot fork child process");
-    return -1;
-}
-
-
 #undef connect
 int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
                       socklen_t addrlen)
-- 
2.39.2

[PATCH v3 02/16] tests: use closesocket()

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Because they are actually sockets...

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
---
 tests/unit/socket-helpers.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/tests/unit/socket-helpers.c b/tests/unit/socket-helpers.c
index eecadf3a3c..914b3aa0cf 100644
--- a/tests/unit/socket-helpers.c
+++ b/tests/unit/socket-helpers.c
@@ -117,13 +117,13 @@ static int socket_can_bind_connect(const char *hostname, int family)
 
  cleanup:
     if (afd != -1) {
-        close(afd);
+        closesocket(afd);
     }
     if (cfd != -1) {
-        close(cfd);
+        closesocket(cfd);
     }
     if (lfd != -1) {
-        close(lfd);
+        closesocket(lfd);
     }
     if (res) {
         freeaddrinfo(res);
-- 
2.39.2

[PATCH v3 03/16] io: use closesocket()

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Because they are actually sockets...

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
---
 io/channel-socket.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/io/channel-socket.c b/io/channel-socket.c
index 7aca84f61a..2040297d2b 100644
--- a/io/channel-socket.c
+++ b/io/channel-socket.c
@@ -159,7 +159,7 @@ int qio_channel_socket_connect_sync(QIOChannelSocket *ioc,
 
     trace_qio_channel_socket_connect_complete(ioc, fd);
     if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
-        close(fd);
+        closesocket(fd);
         return -1;
     }
 
@@ -233,7 +233,7 @@ int qio_channel_socket_listen_sync(QIOChannelSocket *ioc,
 
     trace_qio_channel_socket_listen_complete(ioc, fd);
     if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
-        close(fd);
+        closesocket(fd);
         return -1;
     }
     qio_channel_set_feature(QIO_CHANNEL(ioc), QIO_CHANNEL_FEATURE_LISTEN);
@@ -310,7 +310,7 @@ int qio_channel_socket_dgram_sync(QIOChannelSocket *ioc,
 
     trace_qio_channel_socket_dgram_complete(ioc, fd);
     if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
-        close(fd);
+        closesocket(fd);
         return -1;
     }
 
-- 
2.39.2

[PATCH v3 04/16] tests: add test-error-report

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 tests/unit/test-error-report.c | 121 +++++++++++++++++++++++++++++++++
 tests/unit/meson.build         |   1 +
 2 files changed, 122 insertions(+)
 create mode 100644 tests/unit/test-error-report.c

diff --git a/tests/unit/test-error-report.c b/tests/unit/test-error-report.c
new file mode 100644
index 0000000000..b09650687b
--- /dev/null
+++ b/tests/unit/test-error-report.c
@@ -0,0 +1,121 @@
+/*
+ * Error reporting test
+ *
+ * Copyright (C) 2022 Red Hat Inc.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "glib-compat.h"
+#include <locale.h>
+
+#include "qemu/error-report.h"
+
+static void
+test_error_report_simple(void)
+{
+    if (g_test_subprocess()) {
+        error_report("%s", "test error");
+        warn_report("%s", "test warn");
+        info_report("%s", "test info");
+        return;
+    }
+
+    g_test_trap_subprocess(NULL, 0, 0);
+    g_test_trap_assert_passed();
+    g_test_trap_assert_stderr("\
+test-error-report: test error*\
+test-error-report: warning: test warn*\
+test-error-report: info: test info*\
+");
+}
+
+static void
+test_error_report_loc(void)
+{
+    if (g_test_subprocess()) {
+        loc_set_file("some-file.c", 7717);
+        error_report("%s", "test error1");
+        loc_set_none();
+        error_report("%s", "test error2");
+        return;
+    }
+
+    g_test_trap_subprocess(NULL, 0, 0);
+    g_test_trap_assert_passed();
+    g_test_trap_assert_stderr("\
+test-error-report:some-file.c:7717: test error1*\
+test-error-report: test error2*\
+");
+}
+
+static void
+test_error_report_glog(void)
+{
+    if (g_test_subprocess()) {
+        g_message("gmessage");
+        return;
+    }
+
+    g_test_trap_subprocess(NULL, 0, 0);
+    g_test_trap_assert_passed();
+    g_test_trap_assert_stderr("test-error-report: info: gmessage*");
+}
+
+static void
+test_error_report_once(void)
+{
+    int i;
+
+    if (g_test_subprocess()) {
+        for (i = 0; i < 3; i++) {
+            warn_report_once("warn");
+            error_report_once("err");
+        }
+        return;
+    }
+
+    g_test_trap_subprocess(NULL, 0, 0);
+    g_test_trap_assert_passed();
+    g_test_trap_assert_stderr("\
+test-error-report: warning: warn*\
+test-error-report: err*\
+");
+}
+
+static void
+test_error_report_timestamp(void)
+{
+    if (g_test_subprocess()) {
+        message_with_timestamp = true;
+        warn_report("warn");
+        error_report("err");
+        return;
+    }
+
+    g_test_trap_subprocess(NULL, 0, 0);
+    g_test_trap_assert_passed();
+    g_test_trap_assert_stderr("\
+*-*-*:*:* test-error-report: warning: warn*\
+*-*-*:*:* test-error-report: err*\
+");
+}
+
+int
+main(int argc, char *argv[])
+{
+    setlocale(LC_ALL, "");
+
+    g_test_init(&argc, &argv, NULL);
+    error_init("test-error-report");
+
+    g_test_add_func("/error-report/simple", test_error_report_simple);
+    g_test_add_func("/error-report/loc", test_error_report_loc);
+    g_test_add_func("/error-report/glog", test_error_report_glog);
+    g_test_add_func("/error-report/once", test_error_report_once);
+    g_test_add_func("/error-report/timestamp", test_error_report_timestamp);
+
+    return g_test_run();
+}
diff --git a/tests/unit/meson.build b/tests/unit/meson.build
index ffa444f432..015e8131df 100644
--- a/tests/unit/meson.build
+++ b/tests/unit/meson.build
@@ -11,6 +11,7 @@ tests = {
   'check-qobject': [],
   'check-qjson': [],
   'check-qlit': [],
+  'test-error-report': [],
   'test-qobject-output-visitor': [testqapi],
   'test-clone-visitor': [testqapi],
   'test-qobject-input-visitor': [testqapi],
-- 
2.39.2

[PATCH v3 05/16] error: add global &error_warn destination

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

This can help debugging issues or develop, when error handling is
introduced.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/qapi/error.h           |  6 ++++++
 tests/unit/test-error-report.c | 18 ++++++++++++++++++
 util/error.c                   | 10 +++++++---
 3 files changed, 31 insertions(+), 3 deletions(-)

diff --git a/include/qapi/error.h b/include/qapi/error.h
index d798faeec3..f21a231bb1 100644
--- a/include/qapi/error.h
+++ b/include/qapi/error.h
@@ -519,6 +519,12 @@ static inline void error_propagator_cleanup(ErrorPropagator *prop)
 
 G_DEFINE_AUTO_CLEANUP_CLEAR_FUNC(ErrorPropagator, error_propagator_cleanup);
 
+/*
+ * Special error destination to warn on error.
+ * See error_setg() and error_propagate() for details.
+ */
+extern Error *error_warn;
+
 /*
  * Special error destination to abort on error.
  * See error_setg() and error_propagate() for details.
diff --git a/tests/unit/test-error-report.c b/tests/unit/test-error-report.c
index b09650687b..54319c86c9 100644
--- a/tests/unit/test-error-report.c
+++ b/tests/unit/test-error-report.c
@@ -12,6 +12,7 @@
 #include <locale.h>
 
 #include "qemu/error-report.h"
+#include "qapi/error.h"
 
 static void
 test_error_report_simple(void)
@@ -103,6 +104,22 @@ test_error_report_timestamp(void)
 ");
 }
 
+static void
+test_error_warn(void)
+{
+    if (g_test_subprocess()) {
+        error_setg(&error_warn, "Testing &error_warn");
+        return;
+    }
+
+    g_test_trap_subprocess(NULL, 0, 0);
+    g_test_trap_assert_passed();
+    g_test_trap_assert_stderr("\
+test-error-report: warning: Testing &error_warn*\
+");
+}
+
+
 int
 main(int argc, char *argv[])
 {
@@ -116,6 +133,7 @@ main(int argc, char *argv[])
     g_test_add_func("/error-report/glog", test_error_report_glog);
     g_test_add_func("/error-report/once", test_error_report_once);
     g_test_add_func("/error-report/timestamp", test_error_report_timestamp);
+    g_test_add_func("/error-report/warn", test_error_warn);
 
     return g_test_run();
 }
diff --git a/util/error.c b/util/error.c
index 1e7af665b8..5537245da6 100644
--- a/util/error.c
+++ b/util/error.c
@@ -27,8 +27,9 @@ struct Error
 
 Error *error_abort;
 Error *error_fatal;
+Error *error_warn;
 
-static void error_handle_fatal(Error **errp, Error *err)
+static void error_handle(Error **errp, Error *err)
 {
     if (errp == &error_abort) {
         fprintf(stderr, "Unexpected error in %s() at %s:%d:\n",
@@ -43,6 +44,9 @@ static void error_handle_fatal(Error **errp, Error *err)
         error_report_err(err);
         exit(1);
     }
+    if (errp == &error_warn) {
+        warn_report_err(err);
+    }
 }
 
 G_GNUC_PRINTF(6, 0)
@@ -71,7 +75,7 @@ static void error_setv(Error **errp,
     err->line = line;
     err->func = func;
 
-    error_handle_fatal(errp, err);
+    error_handle(errp, err);
     *errp = err;
 
     errno = saved_errno;
@@ -284,7 +288,7 @@ void error_propagate(Error **dst_errp, Error *local_err)
     if (!local_err) {
         return;
     }
-    error_handle_fatal(dst_errp, local_err);
+    error_handle(dst_errp, local_err);
     if (dst_errp && !*dst_errp) {
         *dst_errp = local_err;
     } else {
-- 
2.39.2

[PATCH v3 06/16] win32/socket: introduce qemu_socket_select() helper

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

This is a wrapper for WSAEventSelect, with Error handling. By default,
it will produce a warning, so callers don't have to be modified
now, and yet we can spot potential mis-use.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/sysemu/os-win32.h |  5 +++++
 io/channel-socket.c       |  4 ++--
 io/channel-watch.c        |  6 +++---
 util/aio-win32.c          |  2 +-
 util/main-loop.c          |  6 +++---
 util/oslib-win32.c        | 17 ++++++++++++++++-
 6 files changed, 30 insertions(+), 10 deletions(-)

diff --git a/include/sysemu/os-win32.h b/include/sysemu/os-win32.h
index 5b38c7bd04..0afb79cc2e 100644
--- a/include/sysemu/os-win32.h
+++ b/include/sysemu/os-win32.h
@@ -29,6 +29,7 @@
 #include <winsock2.h>
 #include <windows.h>
 #include <ws2tcpip.h>
+#include "qemu/typedefs.h"
 
 #ifdef HAVE_AFUNIX_H
 #include <afunix.h>
@@ -144,6 +145,10 @@ static inline void qemu_funlockfile(FILE *f)
 #endif
 }
 
+/* Helper for WSAEventSelect, to report errors */
+bool qemu_socket_select(SOCKET s, WSAEVENT hEventObject,
+                        long lNetworkEvents, Error **errp);
+
 /* We wrap all the sockets functions so that we can
  * set errno based on WSAGetLastError()
  */
diff --git a/io/channel-socket.c b/io/channel-socket.c
index 2040297d2b..0bc29c4808 100644
--- a/io/channel-socket.c
+++ b/io/channel-socket.c
@@ -442,7 +442,7 @@ static void qio_channel_socket_finalize(Object *obj)
             }
         }
 #ifdef WIN32
-        WSAEventSelect(ioc->fd, NULL, 0);
+        qemu_socket_select(ioc->fd, NULL, 0, NULL);
 #endif
         closesocket(ioc->fd);
         ioc->fd = -1;
@@ -846,7 +846,7 @@ qio_channel_socket_close(QIOChannel *ioc,
 
     if (sioc->fd != -1) {
 #ifdef WIN32
-        WSAEventSelect(sioc->fd, NULL, 0);
+        qemu_socket_select(sioc->fd, NULL, 0, NULL);
 #endif
         if (qio_channel_has_feature(ioc, QIO_CHANNEL_FEATURE_LISTEN)) {
             socket_listen_cleanup(sioc->fd, errp);
diff --git a/io/channel-watch.c b/io/channel-watch.c
index ad7c568a84..6ac41009fa 100644
--- a/io/channel-watch.c
+++ b/io/channel-watch.c
@@ -281,9 +281,9 @@ GSource *qio_channel_create_socket_watch(QIOChannel *ioc,
     GSource *source;
     QIOChannelSocketSource *ssource;
 
-    WSAEventSelect(socket, ioc->event,
-                   FD_READ | FD_ACCEPT | FD_CLOSE |
-                   FD_CONNECT | FD_WRITE | FD_OOB);
+    qemu_socket_select(socket, ioc->event,
+                       FD_READ | FD_ACCEPT | FD_CLOSE |
+                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
 
     source = g_source_new(&qio_channel_socket_source_funcs,
                           sizeof(QIOChannelSocketSource));
diff --git a/util/aio-win32.c b/util/aio-win32.c
index 80cfe012ad..be5136e486 100644
--- a/util/aio-win32.c
+++ b/util/aio-win32.c
@@ -115,7 +115,7 @@ void aio_set_fd_handler(AioContext *ctx,
 
         QLIST_INSERT_HEAD_RCU(&ctx->aio_handlers, node, node);
         event = event_notifier_get_handle(&ctx->notifier);
-        WSAEventSelect(node->pfd.fd, event, bitmask);
+        qemu_socket_select(node->pfd.fd, event, bitmask, NULL);
     }
     if (old_node) {
         aio_remove_fd_handler(ctx, old_node);
diff --git a/util/main-loop.c b/util/main-loop.c
index 3c0f525192..16e837fb12 100644
--- a/util/main-loop.c
+++ b/util/main-loop.c
@@ -416,9 +416,9 @@ void qemu_del_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
 
 void qemu_fd_register(int fd)
 {
-    WSAEventSelect(fd, event_notifier_get_handle(&qemu_aio_context->notifier),
-                   FD_READ | FD_ACCEPT | FD_CLOSE |
-                   FD_CONNECT | FD_WRITE | FD_OOB);
+    qemu_socket_select(fd, event_notifier_get_handle(&qemu_aio_context->notifier),
+                       FD_READ | FD_ACCEPT | FD_CLOSE |
+                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
 }
 
 static int pollfds_fill(GArray *pollfds, fd_set *rfds, fd_set *wfds,
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index 528c9ee156..df752fc762 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -180,7 +180,7 @@ static int socket_error(void)
 void qemu_socket_set_block(int fd)
 {
     unsigned long opt = 0;
-    WSAEventSelect(fd, NULL, 0);
+    qemu_socket_select(fd, NULL, 0, NULL);
     ioctlsocket(fd, FIONBIO, &opt);
 }
 
@@ -283,6 +283,21 @@ char *qemu_get_pid_name(pid_t pid)
 }
 
 
+bool qemu_socket_select(SOCKET s, WSAEVENT hEventObject,
+                        long lNetworkEvents, Error **errp)
+{
+    if (errp == NULL) {
+        errp = &error_warn;
+    }
+
+    if (WSAEventSelect(s, hEventObject, lNetworkEvents) != 0) {
+        error_setg_win32(errp, WSAGetLastError(), "failed to WSAEventSelect()");
+        return false;
+    }
+
+    return true;
+}
+
 #undef connect
 int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
                       socklen_t addrlen)
-- 
2.39.2

[PATCH v3 07/16] win32/socket: introduce qemu_socket_unselect() helper

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

A more explicit version of qemu_socket_select() with no events.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/sysemu/os-win32.h | 2 ++
 io/channel-socket.c       | 4 ++--
 util/oslib-win32.c        | 7 ++++++-
 3 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/include/sysemu/os-win32.h b/include/sysemu/os-win32.h
index 0afb79cc2e..ae0c9a3659 100644
--- a/include/sysemu/os-win32.h
+++ b/include/sysemu/os-win32.h
@@ -149,6 +149,8 @@ static inline void qemu_funlockfile(FILE *f)
 bool qemu_socket_select(SOCKET s, WSAEVENT hEventObject,
                         long lNetworkEvents, Error **errp);
 
+bool qemu_socket_unselect(SOCKET s, Error **errp);
+
 /* We wrap all the sockets functions so that we can
  * set errno based on WSAGetLastError()
  */
diff --git a/io/channel-socket.c b/io/channel-socket.c
index 0bc29c4808..03757c7a7e 100644
--- a/io/channel-socket.c
+++ b/io/channel-socket.c
@@ -442,7 +442,7 @@ static void qio_channel_socket_finalize(Object *obj)
             }
         }
 #ifdef WIN32
-        qemu_socket_select(ioc->fd, NULL, 0, NULL);
+        qemu_socket_unselect(ioc->fd, NULL);
 #endif
         closesocket(ioc->fd);
         ioc->fd = -1;
@@ -846,7 +846,7 @@ qio_channel_socket_close(QIOChannel *ioc,
 
     if (sioc->fd != -1) {
 #ifdef WIN32
-        qemu_socket_select(sioc->fd, NULL, 0, NULL);
+        qemu_socket_unselect(sioc->fd, NULL);
 #endif
         if (qio_channel_has_feature(ioc, QIO_CHANNEL_FEATURE_LISTEN)) {
             socket_listen_cleanup(sioc->fd, errp);
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index df752fc762..dbd32acc98 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -180,7 +180,7 @@ static int socket_error(void)
 void qemu_socket_set_block(int fd)
 {
     unsigned long opt = 0;
-    qemu_socket_select(fd, NULL, 0, NULL);
+    qemu_socket_unselect(fd, NULL);
     ioctlsocket(fd, FIONBIO, &opt);
 }
 
@@ -298,6 +298,11 @@ bool qemu_socket_select(SOCKET s, WSAEVENT hEventObject,
     return true;
 }
 
+bool qemu_socket_unselect(SOCKET s, Error **errp)
+{
+    return qemu_socket_select(s, NULL, 0, errp);
+}
+
 #undef connect
 int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
                       socklen_t addrlen)
-- 
2.39.2

[PATCH v3 08/16] aio: make aio_set_fd_poll() static to aio-posix.c

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/block/aio.h | 8 --------
 util/aio-posix.c    | 6 +++---
 util/aio-win32.c    | 7 -------
 3 files changed, 3 insertions(+), 18 deletions(-)

diff --git a/include/block/aio.h b/include/block/aio.h
index 8fba6a3584..543717f294 100644
--- a/include/block/aio.h
+++ b/include/block/aio.h
@@ -482,14 +482,6 @@ void aio_set_fd_handler(AioContext *ctx,
                         IOHandler *io_poll_ready,
                         void *opaque);
 
-/* Set polling begin/end callbacks for a file descriptor that has already been
- * registered with aio_set_fd_handler.  Do nothing if the file descriptor is
- * not registered.
- */
-void aio_set_fd_poll(AioContext *ctx, int fd,
-                     IOHandler *io_poll_begin,
-                     IOHandler *io_poll_end);
-
 /* Register an event notifier and associated callbacks.  Behaves very similarly
  * to event_notifier_set_handler.  Unlike event_notifier_set_handler, these callbacks
  * will be invoked when using aio_poll().
diff --git a/util/aio-posix.c b/util/aio-posix.c
index 6cc6256d53..a8be940f76 100644
--- a/util/aio-posix.c
+++ b/util/aio-posix.c
@@ -180,9 +180,9 @@ void aio_set_fd_handler(AioContext *ctx,
     }
 }
 
-void aio_set_fd_poll(AioContext *ctx, int fd,
-                     IOHandler *io_poll_begin,
-                     IOHandler *io_poll_end)
+static void aio_set_fd_poll(AioContext *ctx, int fd,
+                            IOHandler *io_poll_begin,
+                            IOHandler *io_poll_end)
 {
     AioHandler *node = find_aio_handler(ctx, fd);
 
diff --git a/util/aio-win32.c b/util/aio-win32.c
index be5136e486..74d63fa21e 100644
--- a/util/aio-win32.c
+++ b/util/aio-win32.c
@@ -125,13 +125,6 @@ void aio_set_fd_handler(AioContext *ctx,
     aio_notify(ctx);
 }
 
-void aio_set_fd_poll(AioContext *ctx, int fd,
-                     IOHandler *io_poll_begin,
-                     IOHandler *io_poll_end)
-{
-    /* Not implemented */
-}
-
 void aio_set_event_notifier(AioContext *ctx,
                             EventNotifier *e,
                             bool is_external,
-- 
2.39.2

[PATCH v3 09/16] aio/win32: aio_set_fd_handler() only supports SOCKET

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Let's check if the argument is actually a SOCKET, else report an error
and return.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 util/aio-win32.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/util/aio-win32.c b/util/aio-win32.c
index 74d63fa21e..08e8f5615d 100644
--- a/util/aio-win32.c
+++ b/util/aio-win32.c
@@ -22,6 +22,7 @@
 #include "qemu/sockets.h"
 #include "qapi/error.h"
 #include "qemu/rcu_queue.h"
+#include "qemu/error-report.h"
 
 struct AioHandler {
     EventNotifier *e;
@@ -70,10 +71,14 @@ void aio_set_fd_handler(AioContext *ctx,
                         IOHandler *io_poll_ready,
                         void *opaque)
 {
-    /* fd is a SOCKET in our case */
     AioHandler *old_node;
     AioHandler *node = NULL;
 
+    if (!fd_is_socket(fd)) {
+        error_report("fd=%d is not a socket, AIO implementation is missing", fd);
+        return;
+    }
+
     qemu_lockcnt_lock(&ctx->list_lock);
     QLIST_FOREACH(old_node, &ctx->aio_handlers, node) {
         if (old_node->pfd.fd == fd && !old_node->deleted) {
-- 
2.39.2

[PATCH v3 10/16] RFC: build-sys: add slirp.wrap

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

This allows to build with --enable-slirp / -D slirp=enabled, even when
libslirp is not installed on the system. Meson will pull it from git in
that case.

RFC because this is very convenient, for a developper targetting
different environments, but might not be considered appropriate, as it
is "a kind of" git submodule (without git submodule integration issues
though, afaik, experience should tell).

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 .gitignore             | 2 ++
 subprojects/slirp.wrap | 6 ++++++
 2 files changed, 8 insertions(+)
 create mode 100644 subprojects/slirp.wrap

diff --git a/.gitignore b/.gitignore
index 61fa39967b..1ea59f4819 100644
--- a/.gitignore
+++ b/.gitignore
@@ -20,3 +20,5 @@ GTAGS
 *.swp
 *.patch
 *.gcov
+
+/subprojects/slirp
diff --git a/subprojects/slirp.wrap b/subprojects/slirp.wrap
new file mode 100644
index 0000000000..87cdd8dcd8
--- /dev/null
+++ b/subprojects/slirp.wrap
@@ -0,0 +1,6 @@
+[wrap-git]
+url = https://gitlab.freedesktop.org/slirp/libslirp
+revision = 15c52d69
+
+[provide]
+slirp = libslirp_dep
-- 
2.39.2

[PATCH v3 11/16] main-loop: remove qemu_fd_register(), win32/slirp/socket specific

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Open-code the socket registration where it's needed, to avoid
artificially used or unclear generic interface.

Furthermore, the following patches are going to make socket handling use
FD-only inside QEMU, but we need to handle win32 SOCKET from libslirp.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/qemu/main-loop.h |  2 --
 net/slirp.c              |  8 +++++++-
 util/main-loop.c         | 11 -----------
 3 files changed, 7 insertions(+), 14 deletions(-)

diff --git a/include/qemu/main-loop.h b/include/qemu/main-loop.h
index c25f390696..b3e54e00bc 100644
--- a/include/qemu/main-loop.h
+++ b/include/qemu/main-loop.h
@@ -387,8 +387,6 @@ void qemu_cond_timedwait_iothread(QemuCond *cond, int ms);
 
 /* internal interfaces */
 
-void qemu_fd_register(int fd);
-
 #define qemu_bh_new(cb, opaque) \
     qemu_bh_new_full((cb), (opaque), (stringify(cb)))
 QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name);
diff --git a/net/slirp.c b/net/slirp.c
index 2ee3f1a0d7..0730a935ba 100644
--- a/net/slirp.c
+++ b/net/slirp.c
@@ -248,7 +248,13 @@ static void net_slirp_timer_mod(void *timer, int64_t expire_timer,
 
 static void net_slirp_register_poll_fd(int fd, void *opaque)
 {
-    qemu_fd_register(fd);
+#ifdef WIN32
+    AioContext *ctxt = qemu_get_aio_context();
+
+    qemu_socket_select(fd, event_notifier_get_handle(&ctxt->notifier),
+                       FD_READ | FD_ACCEPT | FD_CLOSE |
+                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
+#endif
 }
 
 static void net_slirp_unregister_poll_fd(int fd, void *opaque)
diff --git a/util/main-loop.c b/util/main-loop.c
index 16e837fb12..e180c85145 100644
--- a/util/main-loop.c
+++ b/util/main-loop.c
@@ -252,10 +252,6 @@ static int max_priority;
 static int glib_pollfds_idx;
 static int glib_n_poll_fds;
 
-void qemu_fd_register(int fd)
-{
-}
-
 static void glib_pollfds_fill(int64_t *cur_timeout)
 {
     GMainContext *context = g_main_context_default();
@@ -414,13 +410,6 @@ void qemu_del_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
     }
 }
 
-void qemu_fd_register(int fd)
-{
-    qemu_socket_select(fd, event_notifier_get_handle(&qemu_aio_context->notifier),
-                       FD_READ | FD_ACCEPT | FD_CLOSE |
-                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
-}
-
 static int pollfds_fill(GArray *pollfds, fd_set *rfds, fd_set *wfds,
                         fd_set *xfds)
 {
-- 
2.39.2

[PATCH v3 12/16] slirp: unregister the win32 SOCKET

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Presumably, this is what should happen when the SOCKET is to be removed.
(it probably worked until now because closesocket() does it implicitly,
but we never now how the slirp library could use the SOCKET later)

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 net/slirp.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/slirp.c b/net/slirp.c
index 0730a935ba..a7c35778a6 100644
--- a/net/slirp.c
+++ b/net/slirp.c
@@ -259,7 +259,9 @@ static void net_slirp_register_poll_fd(int fd, void *opaque)
 
 static void net_slirp_unregister_poll_fd(int fd, void *opaque)
 {
-    /* no qemu_fd_unregister */
+#ifdef WIN32
+    qemu_socket_unselect(fd, NULL);
+#endif
 }
 
 static void net_slirp_notify(void *opaque)
-- 
2.39.2

[PATCH v3 13/16] slirp: open-code qemu_socket_(un)select()

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

We are about to make the QEMU socket API use file-descriptor space only,
but libslirp gives us SOCKET as fd, still.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 net/slirp.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/net/slirp.c b/net/slirp.c
index a7c35778a6..c33b3e02e7 100644
--- a/net/slirp.c
+++ b/net/slirp.c
@@ -251,16 +251,20 @@ static void net_slirp_register_poll_fd(int fd, void *opaque)
 #ifdef WIN32
     AioContext *ctxt = qemu_get_aio_context();
 
-    qemu_socket_select(fd, event_notifier_get_handle(&ctxt->notifier),
+    if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
                        FD_READ | FD_ACCEPT | FD_CLOSE |
-                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
+                       FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
+        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
+    }
 #endif
 }
 
 static void net_slirp_unregister_poll_fd(int fd, void *opaque)
 {
 #ifdef WIN32
-    qemu_socket_unselect(fd, NULL);
+    if (WSAEventSelect(fd, NULL, 0) != 0) {
+        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
+    }
 #endif
 }
 
-- 
2.39.2

[PATCH v3 14/16] win32: avoid mixing SOCKET and file descriptor space

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Until now, a win32 SOCKET handle is often cast to an int file
descriptor, as this is what other OS use for sockets. When necessary,
QEMU eventually queries whether it's a socket with the help of
fd_is_socket(). However, there is no guarantee of conflict between the
fd and SOCKET space. Such conflict would have surprising consequences,
we shouldn't mix them.

Also, it is often forgotten that SOCKET must be closed with
closesocket(), and not close().

Instead, let's make the win32 socket wrapper functions return and take a
file descriptor, and let util/ wrappers do the fd/SOCKET conversion as
necessary. A bit of adaptation is necessary in io/ as well.

Unfortunately, we can't drop closesocket() usage, despite
_open_osfhandle() documentation claiming transfer of ownership, testing
shows bad behaviour if you forget to call closesocket().

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/sysemu/os-win32.h |   4 +-
 io/channel-watch.c        |   6 +-
 util/aio-win32.c          |   9 +-
 util/oslib-win32.c        | 219 ++++++++++++++++++++++++++++++++------
 4 files changed, 197 insertions(+), 41 deletions(-)

diff --git a/include/sysemu/os-win32.h b/include/sysemu/os-win32.h
index ae0c9a3659..d440f4e03e 100644
--- a/include/sysemu/os-win32.h
+++ b/include/sysemu/os-win32.h
@@ -146,10 +146,10 @@ static inline void qemu_funlockfile(FILE *f)
 }
 
 /* Helper for WSAEventSelect, to report errors */
-bool qemu_socket_select(SOCKET s, WSAEVENT hEventObject,
+bool qemu_socket_select(int sockfd, WSAEVENT hEventObject,
                         long lNetworkEvents, Error **errp);
 
-bool qemu_socket_unselect(SOCKET s, Error **errp);
+bool qemu_socket_unselect(int sockfd, Error **errp);
 
 /* We wrap all the sockets functions so that we can
  * set errno based on WSAGetLastError()
diff --git a/io/channel-watch.c b/io/channel-watch.c
index 6ac41009fa..64b486e378 100644
--- a/io/channel-watch.c
+++ b/io/channel-watch.c
@@ -275,13 +275,13 @@ GSource *qio_channel_create_fd_watch(QIOChannel *ioc,
 
 #ifdef CONFIG_WIN32
 GSource *qio_channel_create_socket_watch(QIOChannel *ioc,
-                                         int socket,
+                                         int sockfd,
                                          GIOCondition condition)
 {
     GSource *source;
     QIOChannelSocketSource *ssource;
 
-    qemu_socket_select(socket, ioc->event,
+    qemu_socket_select(sockfd, ioc->event,
                        FD_READ | FD_ACCEPT | FD_CLOSE |
                        FD_CONNECT | FD_WRITE | FD_OOB, NULL);
 
@@ -293,7 +293,7 @@ GSource *qio_channel_create_socket_watch(QIOChannel *ioc,
     object_ref(OBJECT(ioc));
 
     ssource->condition = condition;
-    ssource->socket = socket;
+    ssource->socket = _get_osfhandle(sockfd);
     ssource->revents = 0;
 
     ssource->fd.fd = (gintptr)ioc->event;
diff --git a/util/aio-win32.c b/util/aio-win32.c
index 08e8f5615d..6bded009a4 100644
--- a/util/aio-win32.c
+++ b/util/aio-win32.c
@@ -73,15 +73,18 @@ void aio_set_fd_handler(AioContext *ctx,
 {
     AioHandler *old_node;
     AioHandler *node = NULL;
+    SOCKET s;
 
     if (!fd_is_socket(fd)) {
         error_report("fd=%d is not a socket, AIO implementation is missing", fd);
         return;
     }
 
+    s = _get_osfhandle(fd);
+
     qemu_lockcnt_lock(&ctx->list_lock);
     QLIST_FOREACH(old_node, &ctx->aio_handlers, node) {
-        if (old_node->pfd.fd == fd && !old_node->deleted) {
+        if (old_node->pfd.fd == s && !old_node->deleted) {
             break;
         }
     }
@@ -92,7 +95,7 @@ void aio_set_fd_handler(AioContext *ctx,
 
         /* Alloc and insert if it's not already there */
         node = g_new0(AioHandler, 1);
-        node->pfd.fd = fd;
+        node->pfd.fd = s;
 
         node->pfd.events = 0;
         if (node->io_read) {
@@ -120,7 +123,7 @@ void aio_set_fd_handler(AioContext *ctx,
 
         QLIST_INSERT_HEAD_RCU(&ctx->aio_handlers, node, node);
         event = event_notifier_get_handle(&ctx->notifier);
-        qemu_socket_select(node->pfd.fd, event, bitmask, NULL);
+        qemu_socket_select(fd, event, bitmask, NULL);
     }
     if (old_node) {
         aio_remove_fd_handler(ctx, old_node);
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index dbd32acc98..7836fb0be3 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -283,13 +283,20 @@ char *qemu_get_pid_name(pid_t pid)
 }
 
 
-bool qemu_socket_select(SOCKET s, WSAEVENT hEventObject,
+bool qemu_socket_select(int sockfd, WSAEVENT hEventObject,
                         long lNetworkEvents, Error **errp)
 {
+    SOCKET s = _get_osfhandle(sockfd);
+
     if (errp == NULL) {
         errp = &error_warn;
     }
 
+    if (s == INVALID_SOCKET) {
+        error_setg(errp, "invalid socket fd=%d", sockfd);
+        return false;
+    }
+
     if (WSAEventSelect(s, hEventObject, lNetworkEvents) != 0) {
         error_setg_win32(errp, WSAGetLastError(), "failed to WSAEventSelect()");
         return false;
@@ -298,9 +305,9 @@ bool qemu_socket_select(SOCKET s, WSAEVENT hEventObject,
     return true;
 }
 
-bool qemu_socket_unselect(SOCKET s, Error **errp)
+bool qemu_socket_unselect(int sockfd, Error **errp)
 {
-    return qemu_socket_select(s, NULL, 0, errp);
+    return qemu_socket_select(sockfd, NULL, 0, errp);
 }
 
 #undef connect
@@ -308,7 +315,13 @@ int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
                       socklen_t addrlen)
 {
     int ret;
-    ret = connect(sockfd, addr, addrlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = connect(s, addr, addrlen);
     if (ret < 0) {
         if (WSAGetLastError() == WSAEWOULDBLOCK) {
             errno = EINPROGRESS;
@@ -324,7 +337,13 @@ int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
 int qemu_listen_wrap(int sockfd, int backlog)
 {
     int ret;
-    ret = listen(sockfd, backlog);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = listen(s, backlog);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -337,7 +356,13 @@ int qemu_bind_wrap(int sockfd, const struct sockaddr *addr,
                    socklen_t addrlen)
 {
     int ret;
-    ret = bind(sockfd, addr, addrlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = bind(s, addr, addrlen);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -345,28 +370,108 @@ int qemu_bind_wrap(int sockfd, const struct sockaddr *addr,
 }
 
 
-#undef socket
-int qemu_socket_wrap(int domain, int type, int protocol)
+#undef closesocket
+int qemu_closesocket_wrap(int fd)
 {
     int ret;
-    ret = socket(domain, type, protocol);
+    DWORD flags = 0;
+    SOCKET s = _get_osfhandle(fd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    /*
+     * If we were to just call _close on the descriptor, it would close the
+     * HANDLE, but it wouldn't free any of the resources associated to the
+     * SOCKET, and we can't call _close after calling closesocket, because
+     * closesocket has already closed the HANDLE, and _close would attempt to
+     * close the HANDLE again, resulting in a double free. We can however
+     * protect the HANDLE from actually being closed long enough to close the
+     * file descriptor, then close the socket itself.
+     */
+    if (!GetHandleInformation((HANDLE)s, &flags)) {
+        errno = EACCES;
+        return -1;
+    }
+
+    if (!SetHandleInformation((HANDLE)s, HANDLE_FLAG_PROTECT_FROM_CLOSE, HANDLE_FLAG_PROTECT_FROM_CLOSE)) {
+        errno = EACCES;
+        return -1;
+    }
+
+    ret = close(fd);
+
+    if (!SetHandleInformation((HANDLE)s, flags, flags)) {
+        errno = EACCES;
+        return -1;
+    }
+
+    /*
+     * close() returns EBADF since we PROTECT_FROM_CLOSE the underlying handle,
+     * but the FD is actually freed
+     */
+    if (ret < 0 && errno != EBADF) {
+        return ret;
+    }
+
+    ret = closesocket(s);
     if (ret < 0) {
         errno = socket_error();
     }
+
     return ret;
 }
 
 
+#undef socket
+int qemu_socket_wrap(int domain, int type, int protocol)
+{
+    SOCKET s;
+    int fd;
+
+    s = socket(domain, type, protocol);
+    if (s == -1) {
+        errno = socket_error();
+        return -1;
+    }
+
+    fd = _open_osfhandle(s, _O_BINARY);
+    if (fd < 0) {
+        closesocket(s);
+        /* _open_osfhandle may not set errno, and closesocket() may override it */
+        errno = ENOMEM;
+    }
+
+    return fd;
+}
+
+
 #undef accept
 int qemu_accept_wrap(int sockfd, struct sockaddr *addr,
                      socklen_t *addrlen)
 {
-    int ret;
-    ret = accept(sockfd, addr, addrlen);
-    if (ret < 0) {
+    int fd;
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    s = accept(s, addr, addrlen);
+    if (s == -1) {
         errno = socket_error();
+        return -1;
     }
-    return ret;
+
+    fd = _open_osfhandle(s, _O_BINARY);
+    if (fd < 0) {
+        closesocket(s);
+        /* _open_osfhandle may not set errno, and closesocket() may override it */
+        errno = ENOMEM;
+    }
+
+    return fd;
 }
 
 
@@ -374,7 +479,13 @@ int qemu_accept_wrap(int sockfd, struct sockaddr *addr,
 int qemu_shutdown_wrap(int sockfd, int how)
 {
     int ret;
-    ret = shutdown(sockfd, how);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = shutdown(s, how);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -386,19 +497,13 @@ int qemu_shutdown_wrap(int sockfd, int how)
 int qemu_ioctlsocket_wrap(int fd, int req, void *val)
 {
     int ret;
-    ret = ioctlsocket(fd, req, val);
-    if (ret < 0) {
-        errno = socket_error();
-    }
-    return ret;
-}
+    SOCKET s = _get_osfhandle(fd);
 
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
 
-#undef closesocket
-int qemu_closesocket_wrap(int fd)
-{
-    int ret;
-    ret = closesocket(fd);
+    ret = ioctlsocket(s, req, val);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -411,7 +516,13 @@ int qemu_getsockopt_wrap(int sockfd, int level, int optname,
                          void *optval, socklen_t *optlen)
 {
     int ret;
-    ret = getsockopt(sockfd, level, optname, optval, optlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = getsockopt(s, level, optname, optval, optlen);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -424,7 +535,13 @@ int qemu_setsockopt_wrap(int sockfd, int level, int optname,
                          const void *optval, socklen_t optlen)
 {
     int ret;
-    ret = setsockopt(sockfd, level, optname, optval, optlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = setsockopt(s, level, optname, optval, optlen);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -437,7 +554,13 @@ int qemu_getpeername_wrap(int sockfd, struct sockaddr *addr,
                           socklen_t *addrlen)
 {
     int ret;
-    ret = getpeername(sockfd, addr, addrlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = getpeername(s, addr, addrlen);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -450,7 +573,13 @@ int qemu_getsockname_wrap(int sockfd, struct sockaddr *addr,
                           socklen_t *addrlen)
 {
     int ret;
-    ret = getsockname(sockfd, addr, addrlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = getsockname(s, addr, addrlen);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -462,7 +591,13 @@ int qemu_getsockname_wrap(int sockfd, struct sockaddr *addr,
 ssize_t qemu_send_wrap(int sockfd, const void *buf, size_t len, int flags)
 {
     int ret;
-    ret = send(sockfd, buf, len, flags);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = send(s, buf, len, flags);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -475,7 +610,13 @@ ssize_t qemu_sendto_wrap(int sockfd, const void *buf, size_t len, int flags,
                          const struct sockaddr *addr, socklen_t addrlen)
 {
     int ret;
-    ret = sendto(sockfd, buf, len, flags, addr, addrlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = sendto(s, buf, len, flags, addr, addrlen);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -487,7 +628,13 @@ ssize_t qemu_sendto_wrap(int sockfd, const void *buf, size_t len, int flags,
 ssize_t qemu_recv_wrap(int sockfd, void *buf, size_t len, int flags)
 {
     int ret;
-    ret = recv(sockfd, buf, len, flags);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = recv(s, buf, len, flags);
     if (ret < 0) {
         errno = socket_error();
     }
@@ -500,7 +647,13 @@ ssize_t qemu_recvfrom_wrap(int sockfd, void *buf, size_t len, int flags,
                            struct sockaddr *addr, socklen_t *addrlen)
 {
     int ret;
-    ret = recvfrom(sockfd, buf, len, flags, addr, addrlen);
+    SOCKET s = _get_osfhandle(sockfd);
+
+    if (s == INVALID_SOCKET) {
+        return -1;
+    }
+
+    ret = recvfrom(s, buf, len, flags, addr, addrlen);
     if (ret < 0) {
         errno = socket_error();
     }
-- 
2.39.2

[PATCH v3 15/16] os-posix: remove useless ioctlsocket() define

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

The API is specific to win32.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
---
 include/sysemu/os-posix.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/include/sysemu/os-posix.h b/include/sysemu/os-posix.h
index 58de7c994d..378213fc86 100644
--- a/include/sysemu/os-posix.h
+++ b/include/sysemu/os-posix.h
@@ -52,7 +52,6 @@ void os_setup_post(void);
 int os_mlock(void);
 
 #define closesocket(s) close(s)
-#define ioctlsocket(s, r, v) ioctl(s, r, v)
 
 int os_set_daemonize(bool d);
 bool is_daemonized(void);
-- 
2.39.2

[PATCH v3 16/16] win32: replace closesocket() with close() wrapper

[marcandre.lureau]

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Use a close() wrapper instead, so that we don't need to worry about
closesocket() vs close() anymore, let's hope.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/sysemu/os-posix.h   |  2 --
 include/sysemu/os-win32.h   |  8 +++---
 backends/tpm/tpm_emulator.c |  6 ++--
 crypto/afalg.c              |  6 ++--
 hw/hyperv/syndbg.c          |  4 +--
 io/channel-socket.c         | 10 +++----
 net/dgram.c                 | 14 +++++-----
 net/socket.c                | 22 +++++++--------
 tests/qtest/libqtest.c      |  8 +++---
 tests/qtest/microbit-test.c |  2 +-
 tests/qtest/netdev-socket.c | 10 +++----
 tests/unit/socket-helpers.c |  8 +++---
 util/oslib-win32.c          | 56 +++++++++++++++++++------------------
 util/qemu-sockets.c         | 22 +++++++--------
 14 files changed, 89 insertions(+), 89 deletions(-)

diff --git a/include/sysemu/os-posix.h b/include/sysemu/os-posix.h
index 378213fc86..1030d39904 100644
--- a/include/sysemu/os-posix.h
+++ b/include/sysemu/os-posix.h
@@ -51,8 +51,6 @@ void os_daemonize(void);
 void os_setup_post(void);
 int os_mlock(void);
 
-#define closesocket(s) close(s)
-
 int os_set_daemonize(bool d);
 bool is_daemonized(void);
 
diff --git a/include/sysemu/os-win32.h b/include/sysemu/os-win32.h
index d440f4e03e..f9a6f71b5a 100644
--- a/include/sysemu/os-win32.h
+++ b/include/sysemu/os-win32.h
@@ -155,6 +155,10 @@ bool qemu_socket_unselect(int sockfd, Error **errp);
  * set errno based on WSAGetLastError()
  */
 
+#undef close
+#define close qemu_close_wrap
+int qemu_close_wrap(int fd);
+
 #undef connect
 #define connect qemu_connect_wrap
 int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
@@ -186,10 +190,6 @@ int qemu_shutdown_wrap(int sockfd, int how);
 #define ioctlsocket qemu_ioctlsocket_wrap
 int qemu_ioctlsocket_wrap(int fd, int req, void *val);
 
-#undef closesocket
-#define closesocket qemu_closesocket_wrap
-int qemu_closesocket_wrap(int fd);
-
 #undef getsockopt
 #define getsockopt qemu_getsockopt_wrap
 int qemu_getsockopt_wrap(int sockfd, int level, int optname,
diff --git a/backends/tpm/tpm_emulator.c b/backends/tpm/tpm_emulator.c
index d18144b92e..402a2d6312 100644
--- a/backends/tpm/tpm_emulator.c
+++ b/backends/tpm/tpm_emulator.c
@@ -573,13 +573,13 @@ static int tpm_emulator_prepare_data_fd(TPMEmulator *tpm_emu)
         goto err_exit;
     }
 
-    closesocket(fds[1]);
+    close(fds[1]);
 
     return 0;
 
 err_exit:
-    closesocket(fds[0]);
-    closesocket(fds[1]);
+    close(fds[0]);
+    close(fds[1]);
     return -1;
 }
 
diff --git a/crypto/afalg.c b/crypto/afalg.c
index 10046bb0ae..348301e703 100644
--- a/crypto/afalg.c
+++ b/crypto/afalg.c
@@ -59,7 +59,7 @@ qcrypto_afalg_socket_bind(const char *type, const char *name,
 
     if (bind(sbind, (const struct sockaddr *)&salg, sizeof(salg)) != 0) {
         error_setg_errno(errp, errno, "Failed to bind socket");
-        closesocket(sbind);
+        close(sbind);
         return -1;
     }
 
@@ -105,11 +105,11 @@ void qcrypto_afalg_comm_free(QCryptoAFAlg *afalg)
     }
 
     if (afalg->tfmfd != -1) {
-        closesocket(afalg->tfmfd);
+        close(afalg->tfmfd);
     }
 
     if (afalg->opfd != -1) {
-        closesocket(afalg->opfd);
+        close(afalg->opfd);
     }
 
     g_free(afalg);
diff --git a/hw/hyperv/syndbg.c b/hw/hyperv/syndbg.c
index 94fe1b534b..065e12fb1e 100644
--- a/hw/hyperv/syndbg.c
+++ b/hw/hyperv/syndbg.c
@@ -340,7 +340,7 @@ static void hv_syndbg_realize(DeviceState *dev, Error **errp)
     syndbg->servaddr.sin_family = AF_INET;
     if (connect(syndbg->socket, (struct sockaddr *)&syndbg->servaddr,
                 sizeof(syndbg->servaddr)) < 0) {
-        closesocket(syndbg->socket);
+        close(syndbg->socket);
         error_setg(errp, "%s failed to connect to socket", TYPE_HV_SYNDBG);
         return;
     }
@@ -357,7 +357,7 @@ static void hv_syndbg_unrealize(DeviceState *dev)
 
     if (syndbg->socket > 0) {
         qemu_set_fd_handler(syndbg->socket, NULL, NULL, NULL);
-        closesocket(syndbg->socket);
+        close(syndbg->socket);
     }
 }
 
diff --git a/io/channel-socket.c b/io/channel-socket.c
index 03757c7a7e..b0ea7d48b3 100644
--- a/io/channel-socket.c
+++ b/io/channel-socket.c
@@ -159,7 +159,7 @@ int qio_channel_socket_connect_sync(QIOChannelSocket *ioc,
 
     trace_qio_channel_socket_connect_complete(ioc, fd);
     if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
-        closesocket(fd);
+        close(fd);
         return -1;
     }
 
@@ -233,7 +233,7 @@ int qio_channel_socket_listen_sync(QIOChannelSocket *ioc,
 
     trace_qio_channel_socket_listen_complete(ioc, fd);
     if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
-        closesocket(fd);
+        close(fd);
         return -1;
     }
     qio_channel_set_feature(QIO_CHANNEL(ioc), QIO_CHANNEL_FEATURE_LISTEN);
@@ -310,7 +310,7 @@ int qio_channel_socket_dgram_sync(QIOChannelSocket *ioc,
 
     trace_qio_channel_socket_dgram_complete(ioc, fd);
     if (qio_channel_socket_set_fd(ioc, fd, errp) < 0) {
-        closesocket(fd);
+        close(fd);
         return -1;
     }
 
@@ -444,7 +444,7 @@ static void qio_channel_socket_finalize(Object *obj)
 #ifdef WIN32
         qemu_socket_unselect(ioc->fd, NULL);
 #endif
-        closesocket(ioc->fd);
+        close(ioc->fd);
         ioc->fd = -1;
     }
 }
@@ -852,7 +852,7 @@ qio_channel_socket_close(QIOChannel *ioc,
             socket_listen_cleanup(sioc->fd, errp);
         }
 
-        if (closesocket(sioc->fd) < 0) {
+        if (close(sioc->fd) < 0) {
             sioc->fd = -1;
             error_setg_errno(&err, errno, "Unable to close socket");
             error_propagate(errp, err);
diff --git a/net/dgram.c b/net/dgram.c
index 9f7bf38376..48f653bceb 100644
--- a/net/dgram.c
+++ b/net/dgram.c
@@ -230,7 +230,7 @@ static int net_dgram_mcast_create(struct sockaddr_in *mcastaddr,
     return fd;
 fail:
     if (fd >= 0) {
-        closesocket(fd);
+        close(fd);
     }
     return -1;
 }
@@ -352,7 +352,7 @@ static int net_dgram_mcast_init(NetClientState *peer,
             if (convert_host_port(saddr, local->u.inet.host, local->u.inet.port,
                                   errp) < 0) {
                 g_free(saddr);
-                closesocket(fd);
+                close(fd);
                 return -1;
             }
 
@@ -360,14 +360,14 @@ static int net_dgram_mcast_init(NetClientState *peer,
             if (saddr->sin_addr.s_addr == 0) {
                 error_setg(errp, "can't setup multicast destination address");
                 g_free(saddr);
-                closesocket(fd);
+                close(fd);
                 return -1;
             }
             /* clone dgram socket */
             newfd = net_dgram_mcast_create(saddr, NULL, errp);
             if (newfd < 0) {
                 g_free(saddr);
-                closesocket(fd);
+                close(fd);
                 return -1;
             }
             /* clone newfd to fd, close newfd */
@@ -494,14 +494,14 @@ int net_init_dgram(const Netdev *netdev, const char *name,
         if (ret < 0) {
             error_setg_errno(errp, errno,
                              "can't set socket option SO_REUSEADDR");
-            closesocket(fd);
+            close(fd);
             return -1;
         }
         ret = bind(fd, (struct sockaddr *)&laddr_in, sizeof(laddr_in));
         if (ret < 0) {
             error_setg_errno(errp, errno, "can't bind ip=%s to socket",
                              inet_ntoa(laddr_in.sin_addr));
-            closesocket(fd);
+            close(fd);
             return -1;
         }
         qemu_socket_set_nonblock(fd);
@@ -548,7 +548,7 @@ int net_init_dgram(const Netdev *netdev, const char *name,
         if (ret < 0) {
             error_setg_errno(errp, errno, "can't bind unix=%s to socket",
                              laddr_un.sun_path);
-            closesocket(fd);
+            close(fd);
             return -1;
         }
         qemu_socket_set_nonblock(fd);
diff --git a/net/socket.c b/net/socket.c
index 2fc5696755..ba6e5b0b00 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -172,7 +172,7 @@ static void net_socket_send(void *opaque)
         if (s->listen_fd != -1) {
             qemu_set_fd_handler(s->listen_fd, net_socket_accept, NULL, s);
         }
-        closesocket(s->fd);
+        close(s->fd);
 
         s->fd = -1;
         net_socket_rs_init(&s->rs, net_socket_rs_finalize, false);
@@ -299,7 +299,7 @@ static int net_socket_mcast_create(struct sockaddr_in *mcastaddr,
     return fd;
 fail:
     if (fd >= 0)
-        closesocket(fd);
+        close(fd);
     return -1;
 }
 
@@ -314,7 +314,7 @@ static void net_socket_cleanup(NetClientState *nc)
     }
     if (s->listen_fd != -1) {
         qemu_set_fd_handler(s->listen_fd, NULL, NULL, NULL);
-        closesocket(s->listen_fd);
+        close(s->listen_fd);
         s->listen_fd = -1;
     }
 }
@@ -399,7 +399,7 @@ static NetSocketState *net_socket_fd_init_dgram(NetClientState *peer,
     return s;
 
 err:
-    closesocket(fd);
+    close(fd);
     return NULL;
 }
 
@@ -456,7 +456,7 @@ static NetSocketState *net_socket_fd_init(NetClientState *peer,
     if(getsockopt(fd, SOL_SOCKET, SO_TYPE, (char *)&so_type,
         (socklen_t *)&optlen)< 0) {
         error_setg(errp, "can't get socket option SO_TYPE");
-        closesocket(fd);
+        close(fd);
         return NULL;
     }
     switch(so_type) {
@@ -468,7 +468,7 @@ static NetSocketState *net_socket_fd_init(NetClientState *peer,
     default:
         error_setg(errp, "socket type=%d for fd=%d must be either"
                    " SOCK_DGRAM or SOCK_STREAM", so_type, fd);
-        closesocket(fd);
+        close(fd);
     }
     return NULL;
 }
@@ -526,13 +526,13 @@ static int net_socket_listen_init(NetClientState *peer,
     if (ret < 0) {
         error_setg_errno(errp, errno, "can't bind ip=%s to socket",
                          inet_ntoa(saddr.sin_addr));
-        closesocket(fd);
+        close(fd);
         return -1;
     }
     ret = listen(fd, 0);
     if (ret < 0) {
         error_setg_errno(errp, errno, "can't listen on socket");
-        closesocket(fd);
+        close(fd);
         return -1;
     }
 
@@ -579,7 +579,7 @@ static int net_socket_connect_init(NetClientState *peer,
                 break;
             } else {
                 error_setg_errno(errp, errno, "can't connect socket");
-                closesocket(fd);
+                close(fd);
                 return -1;
             }
         } else {
@@ -671,14 +671,14 @@ static int net_socket_udp_init(NetClientState *peer,
     if (ret < 0) {
         error_setg_errno(errp, errno,
                          "can't set socket option SO_REUSEADDR");
-        closesocket(fd);
+        close(fd);
         return -1;
     }
     ret = bind(fd, (struct sockaddr *)&laddr, sizeof(laddr));
     if (ret < 0) {
         error_setg_errno(errp, errno, "can't bind ip=%s to socket",
                          inet_ntoa(laddr.sin_addr));
-        closesocket(fd);
+        close(fd);
         return -1;
     }
     qemu_socket_set_nonblock(fd);
diff --git a/tests/qtest/libqtest.c b/tests/qtest/libqtest.c
index 2bfd460531..dee2032331 100644
--- a/tests/qtest/libqtest.c
+++ b/tests/qtest/libqtest.c
@@ -124,7 +124,7 @@ static int socket_accept(int sock)
                    (void *)&timeout, sizeof(timeout))) {
         fprintf(stderr, "%s failed to set SO_RCVTIMEO: %s\n",
                 __func__, strerror(errno));
-        closesocket(sock);
+        close(sock);
         return -1;
     }
 
@@ -135,7 +135,7 @@ static int socket_accept(int sock)
     if (ret == -1) {
         fprintf(stderr, "%s failed: %s\n", __func__, strerror(errno));
     }
-    closesocket(sock);
+    close(sock);
 
     return ret;
 }
@@ -564,8 +564,8 @@ void qtest_quit(QTestState *s)
     qtest_remove_abrt_handler(s);
 
     qtest_kill_qemu(s);
-    closesocket(s->fd);
-    closesocket(s->qmp_fd);
+    close(s->fd);
+    close(s->qmp_fd);
     g_string_free(s->rx, true);
 
     for (GList *it = s->pending_events; it != NULL; it = it->next) {
diff --git a/tests/qtest/microbit-test.c b/tests/qtest/microbit-test.c
index 4bc267020b..6022a92b6a 100644
--- a/tests/qtest/microbit-test.c
+++ b/tests/qtest/microbit-test.c
@@ -107,7 +107,7 @@ static void test_nrf51_uart(void)
     g_assert_true(recv(sock_fd, s, 10, 0) == 5);
     g_assert_true(memcmp(s, "world", 5) == 0);
 
-    closesocket(sock_fd);
+    close(sock_fd);
 
     qtest_quit(qts);
 }
diff --git a/tests/qtest/netdev-socket.c b/tests/qtest/netdev-socket.c
index 1d98dca821..614b9bce28 100644
--- a/tests/qtest/netdev-socket.c
+++ b/tests/qtest/netdev-socket.c
@@ -95,7 +95,7 @@ static int inet_get_free_port_multiple(int nb, int *port, bool ipv6)
 
     nb = i;
     for (i = 0; i < nb; i++) {
-        closesocket(sock[i]);
+        close(sock[i]);
     }
 
     return nb;
@@ -262,8 +262,8 @@ static void test_stream_fd(void)
     qtest_quit(qts1);
     qtest_quit(qts0);
 
-    closesocket(sock[0]);
-    closesocket(sock[1]);
+    close(sock[0]);
+    close(sock[1]);
 }
 #endif
 
@@ -388,8 +388,8 @@ static void test_dgram_fd(void)
     qtest_quit(qts1);
     qtest_quit(qts0);
 
-    closesocket(sv[0]);
-    closesocket(sv[1]);
+    close(sv[0]);
+    close(sv[1]);
 }
 #endif
 
diff --git a/tests/unit/socket-helpers.c b/tests/unit/socket-helpers.c
index 914b3aa0cf..6de27baee2 100644
--- a/tests/unit/socket-helpers.c
+++ b/tests/unit/socket-helpers.c
@@ -117,13 +117,13 @@ static int socket_can_bind_connect(const char *hostname, int family)
 
  cleanup:
     if (afd != -1) {
-        closesocket(afd);
+        close(afd);
     }
     if (cfd != -1) {
-        closesocket(cfd);
+        close(cfd);
     }
     if (lfd != -1) {
-        closesocket(lfd);
+        close(lfd);
     }
     if (res) {
         freeaddrinfo(res);
@@ -160,7 +160,7 @@ void socket_check_afunix_support(bool *has_afunix)
     int fd;
 
     fd = socket(PF_UNIX, SOCK_STREAM, 0);
-    closesocket(fd);
+    close(fd);
 
 #ifdef _WIN32
     *has_afunix = (fd != (int)INVALID_SOCKET);
diff --git a/util/oslib-win32.c b/util/oslib-win32.c
index 7836fb0be3..29a667ae3d 100644
--- a/util/oslib-win32.c
+++ b/util/oslib-win32.c
@@ -370,39 +370,39 @@ int qemu_bind_wrap(int sockfd, const struct sockaddr *addr,
 }
 
 
-#undef closesocket
-int qemu_closesocket_wrap(int fd)
+#undef close
+int qemu_close_wrap(int fd)
 {
     int ret;
     DWORD flags = 0;
-    SOCKET s = _get_osfhandle(fd);
+    SOCKET s = INVALID_SOCKET;
 
-    if (s == INVALID_SOCKET) {
-        return -1;
-    }
+    if (fd_is_socket(fd)) {
+        s = _get_osfhandle(fd);
 
-    /*
-     * If we were to just call _close on the descriptor, it would close the
-     * HANDLE, but it wouldn't free any of the resources associated to the
-     * SOCKET, and we can't call _close after calling closesocket, because
-     * closesocket has already closed the HANDLE, and _close would attempt to
-     * close the HANDLE again, resulting in a double free. We can however
-     * protect the HANDLE from actually being closed long enough to close the
-     * file descriptor, then close the socket itself.
-     */
-    if (!GetHandleInformation((HANDLE)s, &flags)) {
-        errno = EACCES;
-        return -1;
-    }
+        /*
+         * If we were to just call _close on the descriptor, it would close the
+         * HANDLE, but it wouldn't free any of the resources associated to the
+         * SOCKET, and we can't call _close after calling closesocket, because
+         * closesocket has already closed the HANDLE, and _close would attempt to
+         * close the HANDLE again, resulting in a double free. We can however
+         * protect the HANDLE from actually being closed long enough to close the
+         * file descriptor, then close the socket itself.
+         */
+        if (!GetHandleInformation((HANDLE)s, &flags)) {
+            errno = EACCES;
+            return -1;
+        }
 
-    if (!SetHandleInformation((HANDLE)s, HANDLE_FLAG_PROTECT_FROM_CLOSE, HANDLE_FLAG_PROTECT_FROM_CLOSE)) {
-        errno = EACCES;
-        return -1;
+        if (!SetHandleInformation((HANDLE)s, HANDLE_FLAG_PROTECT_FROM_CLOSE, HANDLE_FLAG_PROTECT_FROM_CLOSE)) {
+            errno = EACCES;
+            return -1;
+        }
     }
 
     ret = close(fd);
 
-    if (!SetHandleInformation((HANDLE)s, flags, flags)) {
+    if (s != INVALID_SOCKET && !SetHandleInformation((HANDLE)s, flags, flags)) {
         errno = EACCES;
         return -1;
     }
@@ -411,13 +411,15 @@ int qemu_closesocket_wrap(int fd)
      * close() returns EBADF since we PROTECT_FROM_CLOSE the underlying handle,
      * but the FD is actually freed
      */
-    if (ret < 0 && errno != EBADF) {
+    if (ret < 0 && (s == INVALID_SOCKET || errno != EBADF)) {
         return ret;
     }
 
-    ret = closesocket(s);
-    if (ret < 0) {
-        errno = socket_error();
+    if (s != INVALID_SOCKET) {
+        ret = closesocket(s);
+        if (ret < 0) {
+            errno = socket_error();
+        }
     }
 
     return ret;
diff --git a/util/qemu-sockets.c b/util/qemu-sockets.c
index 6538859b87..c06a4dce77 100644
--- a/util/qemu-sockets.c
+++ b/util/qemu-sockets.c
@@ -326,7 +326,7 @@ static int inet_listen_saddr(InetSocketAddress *saddr,
              * recover from this situation, so we need to recreate the
              * socket to allow bind attempts for subsequent ports:
              */
-            closesocket(slisten);
+            close(slisten);
             slisten = -1;
         }
     }
@@ -337,7 +337,7 @@ static int inet_listen_saddr(InetSocketAddress *saddr,
 listen_failed:
     saved_errno = errno;
     if (slisten >= 0) {
-        closesocket(slisten);
+        close(slisten);
     }
     freeaddrinfo(res);
     errno = saved_errno;
@@ -380,7 +380,7 @@ static int inet_connect_addr(const InetSocketAddress *saddr,
     if (rc < 0) {
         error_setg_errno(errp, errno, "Failed to connect to '%s:%s'",
                          saddr->host, saddr->port);
-        closesocket(sock);
+        close(sock);
         return -1;
     }
 
@@ -483,7 +483,7 @@ int inet_connect_saddr(InetSocketAddress *saddr, Error **errp)
 
         if (ret < 0) {
             error_setg_errno(errp, errno, "Unable to set KEEPALIVE");
-            closesocket(sock);
+            close(sock);
             return -1;
         }
     }
@@ -580,7 +580,7 @@ static int inet_dgram_saddr(InetSocketAddress *sraddr,
 
 err:
     if (sock != -1) {
-        closesocket(sock);
+        close(sock);
     }
     if (local) {
         freeaddrinfo(local);
@@ -777,7 +777,7 @@ static int vsock_connect_addr(const VsockSocketAddress *vaddr,
     if (rc < 0) {
         error_setg_errno(errp, errno, "Failed to connect to '%s:%s'",
                          vaddr->cid, vaddr->port);
-        closesocket(sock);
+        close(sock);
         return -1;
     }
 
@@ -814,13 +814,13 @@ static int vsock_listen_saddr(VsockSocketAddress *vaddr,
 
     if (bind(slisten, (const struct sockaddr *)&svm, sizeof(svm)) != 0) {
         error_setg_errno(errp, errno, "Failed to bind socket");
-        closesocket(slisten);
+        close(slisten);
         return -1;
     }
 
     if (listen(slisten, num) != 0) {
         error_setg_errno(errp, errno, "Failed to listen on socket");
-        closesocket(slisten);
+        close(slisten);
         return -1;
     }
     return slisten;
@@ -978,7 +978,7 @@ static int unix_listen_saddr(UnixSocketAddress *saddr,
 
 err:
     g_free(pathbuf);
-    closesocket(sock);
+    close(sock);
     return -1;
 }
 
@@ -1041,7 +1041,7 @@ static int unix_connect_saddr(UnixSocketAddress *saddr, Error **errp)
     return sock;
 
  err:
-    closesocket(sock);
+    close(sock);
     return -1;
 }
 
@@ -1238,7 +1238,7 @@ int socket_listen(SocketAddress *addr, int num, Error **errp)
          */
         if (listen(fd, num) != 0) {
             error_setg_errno(errp, errno, "Failed to listen on fd socket");
-            closesocket(fd);
+            close(fd);
             return -1;
         }
         break;
-- 
2.39.2

Re: [PATCH v3 10/16] RFC: build-sys: add slirp.wrap

[Paolo Bonzini]

On 2/21/23 13:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> This allows to build with --enable-slirp / -D slirp=enabled, even when
> libslirp is not installed on the system. Meson will pull it from git in
> that case.
> 
> RFC because this is very convenient, for a developper targetting
> different environments, but might not be considered appropriate, as it
> is "a kind of" git submodule (without git submodule integration issues
> though, afaik, experience should tell).
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

I have no objections to using wraps for slirp and especially libfdt 
(though the latter is not yet part of wrapdb).

However, right now meson will do network access without respecting 
--with-git-submodule= or something like that, i.e. with no way to avoid 
the fallback.

As a start, configure could hardcode --wrap-mode=nodownload, so that 
wraps would be used only after a conscious decision of the user to use 
"meson subprojects download".

Paolo

> ---
>   .gitignore             | 2 ++
>   subprojects/slirp.wrap | 6 ++++++
>   2 files changed, 8 insertions(+)
>   create mode 100644 subprojects/slirp.wrap
> 
> diff --git a/.gitignore b/.gitignore
> index 61fa39967b..1ea59f4819 100644
> --- a/.gitignore
> +++ b/.gitignore
> @@ -20,3 +20,5 @@ GTAGS
>   *.swp
>   *.patch
>   *.gcov
> +
> +/subprojects/slirp
> diff --git a/subprojects/slirp.wrap b/subprojects/slirp.wrap
> new file mode 100644
> index 0000000000..87cdd8dcd8
> --- /dev/null
> +++ b/subprojects/slirp.wrap
> @@ -0,0 +1,6 @@
> +[wrap-git]
> +url = https://gitlab.freedesktop.org/slirp/libslirp
> +revision = 15c52d69
> +
> +[provide]
> +slirp = libslirp_dep

Re: [PATCH v3 15/16] os-posix: remove useless ioctlsocket() define

[Paolo Bonzini]

On 2/21/23 13:48, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> The API is specific to win32.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

The idea was to use for socket ioctls that are common to POSIX and 
Windows, but it turns out there's no such usecase.

Paolo

> ---
>   include/sysemu/os-posix.h | 1 -
>   1 file changed, 1 deletion(-)
> 
> diff --git a/include/sysemu/os-posix.h b/include/sysemu/os-posix.h
> index 58de7c994d..378213fc86 100644
> --- a/include/sysemu/os-posix.h
> +++ b/include/sysemu/os-posix.h
> @@ -52,7 +52,6 @@ void os_setup_post(void);
>   int os_mlock(void);
>   
>   #define closesocket(s) close(s)
> -#define ioctlsocket(s, r, v) ioctl(s, r, v)
>   
>   int os_set_daemonize(bool d);
>   bool is_daemonized(void);

Re: [PATCH v3 00/16] win32: do not mix SOCKET and fd space

[Marc-André Lureau]

[-- Attachment #1: Type: text/plain, Size: 3719 bytes --]

Hi

On Tue, Feb 21, 2023 at 4:48 PM <marcandre.lureau@redhat.com> wrote:

> From: Marc-André Lureau <marcandre.lureau@redhat.com>
>
> Hi,
>
> A win32 SOCKET handle is often cast to an int file descriptor, as this is
> what
> other OS use for sockets. When necessary, QEMU eventually queries whether
> it's a
> socket with the help of fd_is_socket(). However, there is no guarantee of
> conflict between the fd and SOCKET space. Such conflict would have
> surprising
> consequences. We can fix this by using FDs only.
>
> After fixing a few missed closesocket(), this patch series makes the win32
> socket API wrappers take FDs. It finally get rid of closesocket() usage by
> using
> a close() wrapper instead. (note that fdopen/fclose would not be enough
> either
> to close the underlying socket appropriately)
>
> v3:
> - fix closesocket() to prevent CloseHandle() from close()
> - fix direct closesocket() usage (#undef closesocket before)
> - add a test for &error_warn
> - add r-b tags
>
>
ping  (I am missing reviews, thanks)


> v2:
> - add clean up patch "util: drop qemu_fork()"
> - add a "&error_warn", to help with basic error reporting
> - fix errno handling after _get_osfhandle()
> - introduce qemu_socket_(un)select() helpers
> - add patch "aio_set_fd_handler() only supports SOCKET"
> - add meson slirp.wrap RFC
> - various misc cleanups
> - add r-b tags
>
> Marc-André Lureau (16):
>   util: drop qemu_fork()
>   tests: use closesocket()
>   io: use closesocket()
>   tests: add test-error-report
>   error: add global &error_warn destination
>   win32/socket: introduce qemu_socket_select() helper
>   win32/socket: introduce qemu_socket_unselect() helper
>   aio: make aio_set_fd_poll() static to aio-posix.c
>   aio/win32: aio_set_fd_handler() only supports SOCKET
>   RFC: build-sys: add slirp.wrap
>   main-loop: remove qemu_fd_register(), win32/slirp/socket specific
>   slirp: unregister the win32 SOCKET
>   slirp: open-code qemu_socket_(un)select()
>   win32: avoid mixing SOCKET and file descriptor space
>   os-posix: remove useless ioctlsocket() define
>   win32: replace closesocket() with close() wrapper
>
>  include/block/aio.h            |   8 --
>  include/qapi/error.h           |   6 +
>  include/qemu/main-loop.h       |   2 -
>  include/qemu/osdep.h           |  14 --
>  include/sysemu/os-posix.h      |   3 -
>  include/sysemu/os-win32.h      |  15 ++-
>  backends/tpm/tpm_emulator.c    |   6 +-
>  crypto/afalg.c                 |   6 +-
>  hw/hyperv/syndbg.c             |   4 +-
>  io/channel-socket.c            |   8 +-
>  io/channel-watch.c             |  10 +-
>  net/dgram.c                    |  14 +-
>  net/slirp.c                    |  16 ++-
>  net/socket.c                   |  22 +--
>  tests/qtest/libqtest.c         |   8 +-
>  tests/qtest/microbit-test.c    |   2 +-
>  tests/qtest/netdev-socket.c    |  10 +-
>  tests/unit/socket-helpers.c    |   2 +-
>  tests/unit/test-error-report.c | 139 +++++++++++++++++++
>  util/aio-posix.c               |   6 +-
>  util/aio-win32.c               |  23 ++--
>  util/error.c                   |  10 +-
>  util/main-loop.c               |  11 --
>  util/oslib-posix.c             |  70 ----------
>  util/oslib-win32.c             | 240 ++++++++++++++++++++++++++++-----
>  util/qemu-sockets.c            |  22 +--
>  .gitignore                     |   2 +
>  subprojects/slirp.wrap         |   6 +
>  tests/unit/meson.build         |   1 +
>  29 files changed, 461 insertions(+), 225 deletions(-)
>  create mode 100644 tests/unit/test-error-report.c
>  create mode 100644 subprojects/slirp.wrap
>
> --
> 2.39.2
>
>

[-- Attachment #2: Type: text/html, Size: 4742 bytes --]

Re: [PATCH v3 04/16] tests: add test-error-report

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 05/16] error: add global &error_warn destination

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> This can help debugging issues or develop, when error handling is
> introduced.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 06/16] win32/socket: introduce qemu_socket_select() helper

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> This is a wrapper for WSAEventSelect, with Error handling. By default,
> it will produce a warning, so callers don't have to be modified
> now, and yet we can spot potential mis-use.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 07/16] win32/socket: introduce qemu_socket_unselect() helper

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> A more explicit version of qemu_socket_select() with no events.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 08/16] aio: make aio_set_fd_poll() static to aio-posix.c

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 09/16] aio/win32: aio_set_fd_handler() only supports SOCKET

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Let's check if the argument is actually a SOCKET, else report an error
> and return.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 11/16] main-loop: remove qemu_fd_register(), win32/slirp/socket specific

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Open-code the socket registration where it's needed, to avoid
> artificially used or unclear generic interface.
> 
> Furthermore, the following patches are going to make socket handling use
> FD-only inside QEMU, but we need to handle win32 SOCKET from libslirp.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>   include/qemu/main-loop.h |  2 --
>   net/slirp.c              |  8 +++++++-
>   util/main-loop.c         | 11 -----------
>   3 files changed, 7 insertions(+), 14 deletions(-)
> 
> diff --git a/include/qemu/main-loop.h b/include/qemu/main-loop.h
> index c25f390696..b3e54e00bc 100644
> --- a/include/qemu/main-loop.h
> +++ b/include/qemu/main-loop.h
> @@ -387,8 +387,6 @@ void qemu_cond_timedwait_iothread(QemuCond *cond, int ms);
> 
>   /* internal interfaces */
> 
> -void qemu_fd_register(int fd);
> -
>   #define qemu_bh_new(cb, opaque) \
>       qemu_bh_new_full((cb), (opaque), (stringify(cb)))
>   QEMUBH *qemu_bh_new_full(QEMUBHFunc *cb, void *opaque, const char *name);
> diff --git a/net/slirp.c b/net/slirp.c
> index 2ee3f1a0d7..0730a935ba 100644
> --- a/net/slirp.c
> +++ b/net/slirp.c
> @@ -248,7 +248,13 @@ static void net_slirp_timer_mod(void *timer, int64_t expire_timer,
> 
>   static void net_slirp_register_poll_fd(int fd, void *opaque)
>   {
> -    qemu_fd_register(fd);
> +#ifdef WIN32

_WIN32 ?

With this fixed:

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 12/16] slirp: unregister the win32 SOCKET

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Presumably, this is what should happen when the SOCKET is to be removed.
> (it probably worked until now because closesocket() does it implicitly,
> but we never now how the slirp library could use the SOCKET later)
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>   net/slirp.c | 4 +++-
>   1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/net/slirp.c b/net/slirp.c
> index 0730a935ba..a7c35778a6 100644
> --- a/net/slirp.c
> +++ b/net/slirp.c
> @@ -259,7 +259,9 @@ static void net_slirp_register_poll_fd(int fd, void *opaque)
> 
>   static void net_slirp_unregister_poll_fd(int fd, void *opaque)
>   {
> -    /* no qemu_fd_unregister */
> +#ifdef WIN32
The majority of code seems to use _WIN32. Not sure what is 'right'.

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

> +    qemu_socket_unselect(fd, NULL);
> +#endif
>   }
> 
>   static void net_slirp_notify(void *opaque)

Re: [PATCH v3 14/16] win32: avoid mixing SOCKET and file descriptor space

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Until now, a win32 SOCKET handle is often cast to an int file
> descriptor, as this is what other OS use for sockets. When necessary,
> QEMU eventually queries whether it's a socket with the help of
> fd_is_socket(). However, there is no guarantee of conflict between the
> fd and SOCKET space. Such conflict would have surprising consequences,
> we shouldn't mix them.
> 
> Also, it is often forgotten that SOCKET must be closed with
> closesocket(), and not close().
> 
> Instead, let's make the win32 socket wrapper functions return and take a
> file descriptor, and let util/ wrappers do the fd/SOCKET conversion as
> necessary. A bit of adaptation is necessary in io/ as well.
> 
> Unfortunately, we can't drop closesocket() usage, despite
> _open_osfhandle() documentation claiming transfer of ownership, testing
> shows bad behaviour if you forget to call closesocket().
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>   include/sysemu/os-win32.h |   4 +-
>   io/channel-watch.c        |   6 +-
>   util/aio-win32.c          |   9 +-
>   util/oslib-win32.c        | 219 ++++++++++++++++++++++++++++++++------
>   4 files changed, 197 insertions(+), 41 deletions(-)

>   #undef connect
> @@ -308,7 +315,13 @@ int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
>                         socklen_t addrlen)
>   {
>       int ret;
> -    ret = connect(sockfd, addr, addrlen);
> +    SOCKET s = _get_osfhandle(sockfd);
> +
> +    if (s == INVALID_SOCKET) {
> +        return -1;
> +    }
> +
> +    ret = connect(s, addr, addrlen);


Previously you passed int sockfd and now you convert this sockfd to a SOCKET s and you can pass this as an alternative? Or was sockfd before this patch a SOCKET??
    Stefan

Re: [PATCH v3 14/16] win32: avoid mixing SOCKET and file descriptor space

[Marc-André Lureau]

Hi

On Fri, Mar 3, 2023 at 12:54 AM Stefan Berger <stefanb@linux.ibm.com> wrote:
>
>
>
> On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> > From: Marc-André Lureau <marcandre.lureau@redhat.com>
> >
> > Until now, a win32 SOCKET handle is often cast to an int file
> > descriptor, as this is what other OS use for sockets. When necessary,
> > QEMU eventually queries whether it's a socket with the help of
> > fd_is_socket(). However, there is no guarantee of conflict between the
> > fd and SOCKET space. Such conflict would have surprising consequences,
> > we shouldn't mix them.
> >
> > Also, it is often forgotten that SOCKET must be closed with
> > closesocket(), and not close().
> >
> > Instead, let's make the win32 socket wrapper functions return and take a
> > file descriptor, and let util/ wrappers do the fd/SOCKET conversion as
> > necessary. A bit of adaptation is necessary in io/ as well.
> >
> > Unfortunately, we can't drop closesocket() usage, despite
> > _open_osfhandle() documentation claiming transfer of ownership, testing
> > shows bad behaviour if you forget to call closesocket().
> >
> > Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > ---
> >   include/sysemu/os-win32.h |   4 +-
> >   io/channel-watch.c        |   6 +-
> >   util/aio-win32.c          |   9 +-
> >   util/oslib-win32.c        | 219 ++++++++++++++++++++++++++++++++------
> >   4 files changed, 197 insertions(+), 41 deletions(-)
>
> >   #undef connect
> > @@ -308,7 +315,13 @@ int qemu_connect_wrap(int sockfd, const struct sockaddr *addr,
> >                         socklen_t addrlen)
> >   {
> >       int ret;
> > -    ret = connect(sockfd, addr, addrlen);
> > +    SOCKET s = _get_osfhandle(sockfd);
> > +
> > +    if (s == INVALID_SOCKET) {
> > +        return -1;
> > +    }
> > +
> > +    ret = connect(s, addr, addrlen);
>
>
> Previously you passed int sockfd and now you convert this sockfd to a SOCKET s and you can pass this as an alternative? Or was sockfd before this patch a SOCKET??

yes, sockfd was in fact a SOCKET.

Previous to this patch, a SOCKET is cast to int, as a fake fd, and
back to SOCKET as necessary. The whole point of this patch is to avoid
mixing SOCKET & fd space, instead a SOCKET is associated with a real
CRT fd.

thanks

-- 
Marc-André Lureau

Re: [PATCH v3 12/16] slirp: unregister the win32 SOCKET

[Marc-André Lureau]

[-- Attachment #1: Type: text/plain, Size: 1445 bytes --]

Hi

On Thu, Mar 2, 2023 at 10:45 PM Stefan Berger <stefanb@linux.ibm.com> wrote:

>
>
> On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> > From: Marc-André Lureau <marcandre.lureau@redhat.com>
> >
> > Presumably, this is what should happen when the SOCKET is to be removed.
> > (it probably worked until now because closesocket() does it implicitly,
> > but we never now how the slirp library could use the SOCKET later)
> >
> > Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > ---
> >   net/slirp.c | 4 +++-
> >   1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/net/slirp.c b/net/slirp.c
> > index 0730a935ba..a7c35778a6 100644
> > --- a/net/slirp.c
> > +++ b/net/slirp.c
> > @@ -259,7 +259,9 @@ static void net_slirp_register_poll_fd(int fd, void
> *opaque)
> >
> >   static void net_slirp_unregister_poll_fd(int fd, void *opaque)
> >   {
> > -    /* no qemu_fd_unregister */
> > +#ifdef WIN32
> The majority of code seems to use _WIN32. Not sure what is 'right'.
>
>
Both should be correct. I think I like the "WIN32" version better though
(see also
https://stackoverflow.com/questions/662084/whats-the-difference-between-the-win32-and-win32-defines-in-c
)


> Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
>
>
thanks


> > +    qemu_socket_unselect(fd, NULL);
> > +#endif
> >   }
> >
> >   static void net_slirp_notify(void *opaque)
>
>

[-- Attachment #2: Type: text/html, Size: 2652 bytes --]

Re: [PATCH v3 00/16] win32: do not mix SOCKET and fd space

[Paolo Bonzini]

On 3/2/23 15:09, Marc-André Lureau wrote:
> 
> 
>     v3:
>     - fix closesocket() to prevent CloseHandle() from close()
>     - fix direct closesocket() usage (#undef closesocket before)
>     - add a test for &error_warn
>     - add r-b tags
> 
> ping  (I am missing reviews, thanks)

I'm going to queue this series today if that's fine for you---thanks for 
reminding!

Paolo

Re: [PATCH v3 00/16] win32: do not mix SOCKET and fd space

[Marc-André Lureau]

[-- Attachment #1: Type: text/plain, Size: 635 bytes --]

Hi Paolo

On Mon, Mar 6, 2023 at 12:05 PM Paolo Bonzini <pbonzini@redhat.com> wrote:

> On 3/2/23 15:09, Marc-André Lureau wrote:
> >
> >
> >     v3:
> >     - fix closesocket() to prevent CloseHandle() from close()
> >     - fix direct closesocket() usage (#undef closesocket before)
> >     - add a test for &error_warn
> >     - add r-b tags
> >
> > ping  (I am missing reviews, thanks)
>
> I'm going to queue this series today if that's fine for you---thanks for
> reminding!
>
>
Great, thanks! (I suppose you'll drop "RFC: build-sys: add slirp.wrap", and
perhaps queue the other meson/wrap series instead)

[-- Attachment #2: Type: text/html, Size: 1056 bytes --]

Re: [PATCH v3 00/16] win32: do not mix SOCKET and fd space

[Paolo Bonzini]

On 3/6/23 09:08, Marc-André Lureau wrote:
> Great, thanks! (I suppose you'll drop "RFC: build-sys: add slirp.wrap", 
> and perhaps queue the other meson/wrap series instead)

I don't have time to test the dtc fallback in CI, so I'll queue only the 
first three patches and note it as experimental in the release notes.

In the next release we'll probably have pip support in configure, so 
perhaps we can make the --wrap-mode option conditional.

Paolo

Re: [PATCH v3 13/16] slirp: open-code qemu_socket_(un)select()

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> We are about to make the QEMU socket API use file-descriptor space only,
> but libslirp gives us SOCKET as fd, still.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>   net/slirp.c | 10 +++++++---
>   1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/net/slirp.c b/net/slirp.c
> index a7c35778a6..c33b3e02e7 100644
> --- a/net/slirp.c
> +++ b/net/slirp.c
> @@ -251,16 +251,20 @@ static void net_slirp_register_poll_fd(int fd, void *opaque)

Shouldn't this int fd rather be a SOCKET s instead? Or do you get compiler warnings then?

>   #ifdef WIN32
>       AioContext *ctxt = qemu_get_aio_context();
> 
> -    qemu_socket_select(fd, event_notifier_get_handle(&ctxt->notifier),
> +    if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
>                          FD_READ | FD_ACCEPT | FD_CLOSE |
> -                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
> +                       FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
> +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
> +    }
>   #endif
>   }
> >   static void net_slirp_unregister_poll_fd(int fd, void *opaque)

Same here.

>   {
>   #ifdef WIN32
> -    qemu_socket_unselect(fd, NULL);
> +    if (WSAEventSelect(fd, NULL, 0) != 0) {
> +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
> +    }
>   #endif
>   }
> 

Re: [PATCH v3 13/16] slirp: open-code qemu_socket_(un)select()

[Marc-André Lureau]

[-- Attachment #1: Type: text/plain, Size: 1893 bytes --]

Hi

On Mon, Mar 6, 2023 at 5:59 PM Stefan Berger <stefanb@linux.ibm.com> wrote:

>
>
> On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> > From: Marc-André Lureau <marcandre.lureau@redhat.com>
> >
> > We are about to make the QEMU socket API use file-descriptor space only,
> > but libslirp gives us SOCKET as fd, still.
> >
> > Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > ---
> >   net/slirp.c | 10 +++++++---
> >   1 file changed, 7 insertions(+), 3 deletions(-)
> >
> > diff --git a/net/slirp.c b/net/slirp.c
> > index a7c35778a6..c33b3e02e7 100644
> > --- a/net/slirp.c
> > +++ b/net/slirp.c
> > @@ -251,16 +251,20 @@ static void net_slirp_register_poll_fd(int fd,
> void *opaque)
>
> Shouldn't this int fd rather be a SOCKET s instead? Or do you get compiler
> warnings then?
>
>
Yes, you would get compiler warnings, because the "int fd" argument is from
the slirp API, whether it is posix or win32.


> >   #ifdef WIN32
> >       AioContext *ctxt = qemu_get_aio_context();
> >
> > -    qemu_socket_select(fd, event_notifier_get_handle(&ctxt->notifier),
> > +    if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
> >                          FD_READ | FD_ACCEPT | FD_CLOSE |
> > -                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
> > +                       FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
> > +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to
> WSAEventSelect()");
> > +    }
> >   #endif
> >   }
> > >   static void net_slirp_unregister_poll_fd(int fd, void *opaque)
>
> Same here.
>
> >   {
> >   #ifdef WIN32
> > -    qemu_socket_unselect(fd, NULL);
> > +    if (WSAEventSelect(fd, NULL, 0) != 0) {
> > +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to
> WSAEventSelect()");
> > +    }
> >   #endif
> >   }
> >
>
>

[-- Attachment #2: Type: text/html, Size: 2912 bytes --]

Re: [PATCH v3 13/16] slirp: open-code qemu_socket_(un)select()

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> We are about to make the QEMU socket API use file-descriptor space only,
> but libslirp gives us SOCKET as fd, still.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>   net/slirp.c | 10 +++++++---
>   1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/net/slirp.c b/net/slirp.c
> index a7c35778a6..c33b3e02e7 100644
> --- a/net/slirp.c
> +++ b/net/slirp.c
> @@ -251,16 +251,20 @@ static void net_slirp_register_poll_fd(int fd, void *opaque)
>   #ifdef WIN32
>       AioContext *ctxt = qemu_get_aio_context();
> 
> -    qemu_socket_select(fd, event_notifier_get_handle(&ctxt->notifier),
> +    if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
>                          FD_READ | FD_ACCEPT | FD_CLOSE |
> -                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
> +                       FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
> +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
> +    }
>   #endif
>   }
> 
>   static void net_slirp_unregister_poll_fd(int fd, void *opaque)
>   {
>   #ifdef WIN32
> -    qemu_socket_unselect(fd, NULL);
> +    if (WSAEventSelect(fd, NULL, 0) != 0) {
> +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
> +    }
>   #endif
>   }
> 

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 14/16] win32: avoid mixing SOCKET and file descriptor space

[Stefan Berger]

On 2/21/23 07:47, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Until now, a win32 SOCKET handle is often cast to an int file
> descriptor, as this is what other OS use for sockets. When necessary,
> QEMU eventually queries whether it's a socket with the help of
> fd_is_socket(). However, there is no guarantee of conflict between the
> fd and SOCKET space. Such conflict would have surprising consequences,
> we shouldn't mix them.
> 
> Also, it is often forgotten that SOCKET must be closed with
> closesocket(), and not close().
> 
> Instead, let's make the win32 socket wrapper functions return and take a
> file descriptor, and let util/ wrappers do the fd/SOCKET conversion as
> necessary. A bit of adaptation is necessary in io/ as well.
> 
> Unfortunately, we can't drop closesocket() usage, despite
> _open_osfhandle() documentation claiming transfer of ownership, testing
> shows bad behaviour if you forget to call closesocket().
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 16/16] win32: replace closesocket() with close() wrapper

[Stefan Berger]

On 2/21/23 07:48, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Use a close() wrapper instead, so that we don't need to worry about
> closesocket() vs close() anymore, let's hope.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

> diff --git a/util/oslib-win32.c b/util/oslib-win32.c
> index 7836fb0be3..29a667ae3d 100644
> --- a/util/oslib-win32.c
> +++ b/util/oslib-win32.c
> @@ -370,39 +370,39 @@ int qemu_bind_wrap(int sockfd, const struct sockaddr *addr,
>   }
> 
> 
> -#undef closesocket
> -int qemu_closesocket_wrap(int fd)
> +#undef close
> +int qemu_close_wrap(int fd)
>   {
>       int ret;
>       DWORD flags = 0;
> -    SOCKET s = _get_osfhandle(fd);
> +    SOCKET s = INVALID_SOCKET;
> 
> -    if (s == INVALID_SOCKET) {
> -        return -1;
> -    }
> +    if (fd_is_socket(fd)) {
> +        s = _get_osfhandle(fd);
> 
> -    /*
> -     * If we were to just call _close on the descriptor, it would close the
> -     * HANDLE, but it wouldn't free any of the resources associated to the
> -     * SOCKET, and we can't call _close after calling closesocket, because
> -     * closesocket has already closed the HANDLE, and _close would attempt to
> -     * close the HANDLE again, resulting in a double free. We can however
> -     * protect the HANDLE from actually being closed long enough to close the
> -     * file descriptor, then close the socket itself.
> -     */
> -    if (!GetHandleInformation((HANDLE)s, &flags)) {
> -        errno = EACCES;
> -        return -1;
> -    }
> +        /*
> +         * If we were to just call _close on the descriptor, it would close the
> +         * HANDLE, but it wouldn't free any of the resources associated to the
> +         * SOCKET, and we can't call _close after calling closesocket, because
> +         * closesocket has already closed the HANDLE, and _close would attempt to
> +         * close the HANDLE again, resulting in a double free. We can however
> +         * protect the HANDLE from actually being closed long enough to close the
> +         * file descriptor, then close the socket itself.
> +         */
> +        if (!GetHandleInformation((HANDLE)s, &flags)) {
> +            errno = EACCES;
> +            return -1;
> +        }
> 
> -    if (!SetHandleInformation((HANDLE)s, HANDLE_FLAG_PROTECT_FROM_CLOSE, HANDLE_FLAG_PROTECT_FROM_CLOSE)) {
> -        errno = EACCES;
> -        return -1;
> +        if (!SetHandleInformation((HANDLE)s, HANDLE_FLAG_PROTECT_FROM_CLOSE, HANDLE_FLAG_PROTECT_FROM_CLOSE)) {
> +            errno = EACCES;
> +            return -1;
> +        }
>       }
> 
>       ret = close(fd);
> 
> -    if (!SetHandleInformation((HANDLE)s, flags, flags)) {
> +    if (s != INVALID_SOCKET && !SetHandleInformation((HANDLE)s, flags, flags)) {
>           errno = EACCES;
>           return -1;
>       }
> @@ -411,13 +411,15 @@ int qemu_closesocket_wrap(int fd)
>        * close() returns EBADF since we PROTECT_FROM_CLOSE the underlying handle,
>        * but the FD is actually freed
>        */
> -    if (ret < 0 && errno != EBADF) {
> +    if (ret < 0 && (s == INVALID_SOCKET || errno != EBADF)) {
>           return ret;
>       }
> 
> -    ret = closesocket(s);
> -    if (ret < 0) {
> -        errno = socket_error();
> +    if (s != INVALID_SOCKET) {
> +        ret = closesocket(s);
> +        if (ret < 0) {
> +            errno = socket_error();
> +        }
>       }


if (fs_is_socket()) {{
     ...
     close()
     ...
     closesocket()
     ...
} else {
     ...
     close()
     ...
}

would avoid the threading and make this function look a bit simpler.

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Re: [PATCH v3 13/16] slirp: open-code qemu_socket_(un)select()

[Stefan Berger]

On 3/6/23 09:03, Marc-André Lureau wrote:

> 
> On Mon, Mar 6, 2023 at 5:59 PM Stefan Berger <stefanb@linux.ibm.com <mailto:stefanb@linux.ibm.com>> wrote:
> 
> 
> 
>     On 2/21/23 07:47, marcandre.lureau@redhat.com <mailto:marcandre.lureau@redhat.com> wrote:
>      > From: Marc-André Lureau <marcandre.lureau@redhat.com <mailto:marcandre.lureau@redhat.com>>
>      >
>      > We are about to make the QEMU socket API use file-descriptor space only,
>      > but libslirp gives us SOCKET as fd, still.
>      >
>      > Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com <mailto:marcandre.lureau@redhat.com>>
>      > ---
>      >   net/slirp.c | 10 +++++++---
>      >   1 file changed, 7 insertions(+), 3 deletions(-)
>      >
>      > diff --git a/net/slirp.c b/net/slirp.c
>      > index a7c35778a6..c33b3e02e7 100644
>      > --- a/net/slirp.c
>      > +++ b/net/slirp.c
>      > @@ -251,16 +251,20 @@ static void net_slirp_register_poll_fd(int fd, void *opaque)
> 
>     Shouldn't this int fd rather be a SOCKET s instead? Or do you get compiler warnings then?
> 
> 
> Yes, you would get compiler warnings, because the "int fd" argument is from the slirp API, whether it is posix or win32.


Right, this is shared code.

> 
>      >   #ifdef WIN32
>      >       AioContext *ctxt = qemu_get_aio_context();
>      >
>      > -    qemu_socket_select(fd, event_notifier_get_handle(&ctxt->notifier),
>      > +    if (WSAEventSelect(fd, event_notifier_get_handle(&ctxt->notifier),
>      >                          FD_READ | FD_ACCEPT | FD_CLOSE |
>      > -                       FD_CONNECT | FD_WRITE | FD_OOB, NULL);
>      > +                       FD_CONNECT | FD_WRITE | FD_OOB) != 0) {
>      > +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
>      > +    }
>      >   #endif
>      >   }
>      > >   static void net_slirp_unregister_poll_fd(int fd, void *opaque)
> 
>     Same here.
> 
>      >   {
>      >   #ifdef WIN32
>      > -    qemu_socket_unselect(fd, NULL);
>      > +    if (WSAEventSelect(fd, NULL, 0) != 0) {
>      > +        error_setg_win32(&error_warn, WSAGetLastError(), "failed to WSAEventSelect()");
>      > +    }
>      >   #endif
>      >   }
>      >
>