[Qemu-devel] [PATCH v3 00/16] chardev: refactoring & many bugfixes related tcp_chr_wait_connected

[Qemu-devel] [PATCH v3 00/16] chardev: refactoring & many bugfixes related tcp_chr_wait_connected

[Daniel P. Berrangé]

This is a followup to

  v1: https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg03344.html
  v2: http://lists.nongnu.org/archive/html/qemu-devel/2019-01/msg05947.html

This series comes out of a discussion between myself & Yongji Xie in:

  https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg01881.html

I eventually understood that the problem faced was that
tcp_chr_wait_connected was racing with the background connection attempt
previously started, causing two connections to be established. This
broke because some vhost user servers only allow a single connection.

After messing around with the code alot the final solution was in fact
very easy. We simply have to delay the first background connection
attempt until the main loop is running. It will then automatically
turn into a no-op if tcp_chr_wait_connected has been run. This is
dealt with in the last patch in this series

I believe this should solve the problem Yongji Xie faced, and thus not
require us to add support for "nowait" option with client sockets at
all. The reconnect=1 option effectively already implements nowait
semantics, and now plays nicely with tcp_chr_wait_connected.

In investigating this I found various other bugs that needed fixing and
identified some useful refactoring to simplify / clarify the code, hence
this very long series.

This series passes make check-unit & check-qtest (for x86_64 target).

I did a test with tests/vhost-user-bridge too, which was in fact already
broken for the same reason Yongji Xie found - it only accepts a single
connection. This series fixes use of reconnect=1 with vhost-user-bridge.

Changed in v3:

 - Fix thread between qio_task_thread_worker unlocking mutex
   and qio_task_thread_result free'ing mutex (Peter)
 - Ensure tcp_chr_wait_connected destroys any pending
   reconnect_timer GSource
 - Use g_setenv/g_unsetenv instead of setenv/unsetenv for
   platform portability

Changed in v2:

 - Rewrite the way we synchronize in tcp_chr_wait_connected again
   to cope with chardev+device hotplug scenario
 - Add extensive unit test coverage
 - Resolve conflicts with git master

Daniel P. Berrangé (16):
  io: store reference to thread information in the QIOTask struct
  io: add qio_task_wait_thread to join with a background thread
  chardev: fix validation of options for QMP created chardevs
  chardev: forbid 'reconnect' option with server sockets
  chardev: forbid 'wait' option with client sockets
  chardev: remove many local variables in qemu_chr_parse_socket
  chardev: ensure qemu_chr_parse_compat reports missing driver error
  chardev: remove unused 'sioc' variable & cleanup paths
  chardev: split tcp_chr_wait_connected into two methods
  chardev: split up qmp_chardev_open_socket connection code
  chardev: use a state machine for socket connection state
  chardev: honour the reconnect setting in tcp_chr_wait_connected
  chardev: disallow TLS/telnet/websocket with tcp_chr_wait_connected
  chardev: fix race with client connections in tcp_chr_wait_connected
  tests: expand coverage of socket chardev test
  chardev: ensure termios is fully initialized

 chardev/char-serial.c          |   2 +-
 chardev/char-socket.c          | 490 ++++++++++++++++++-------
 chardev/char.c                 |   2 +
 include/io/task.h              |  29 +-
 io/task.c                      | 101 ++++--
 io/trace-events                |   2 +
 tests/ivshmem-test.c           |   2 +-
 tests/libqtest.c               |   4 +-
 tests/test-char.c              | 643 ++++++++++++++++++++++++---------
 tests/test-filter-redirector.c |   4 +-
 10 files changed, 939 insertions(+), 340 deletions(-)

-- 
2.20.1

[Qemu-devel] [PATCH v3 01/16] io: store reference to thread information in the QIOTask struct

[Daniel P. Berrangé]

Currently the struct QIOTaskThreadData is only needed by the worker
thread, but a subsequent patch will need to access it from another
context.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 io/task.c | 64 +++++++++++++++++++++++++++++--------------------------
 1 file changed, 34 insertions(+), 30 deletions(-)

diff --git a/io/task.c b/io/task.c
index 2886a2c1bc..396866b10f 100644
--- a/io/task.c
+++ b/io/task.c
@@ -24,6 +24,14 @@
 #include "qemu/thread.h"
 #include "trace.h"
 
+struct QIOTaskThreadData {
+    QIOTaskWorker worker;
+    gpointer opaque;
+    GDestroyNotify destroy;
+    GMainContext *context;
+};
+
+
 struct QIOTask {
     Object *source;
     QIOTaskFunc func;
@@ -32,6 +40,7 @@ struct QIOTask {
     Error *err;
     gpointer result;
     GDestroyNotify destroyResult;
+    struct QIOTaskThreadData *thread;
 };
 
 
@@ -57,6 +66,18 @@ QIOTask *qio_task_new(Object *source,
 
 static void qio_task_free(QIOTask *task)
 {
+    if (task->thread) {
+        if (task->thread->destroy) {
+            task->thread->destroy(task->thread->opaque);
+        }
+
+        if (task->thread->context) {
+            g_main_context_unref(task->thread->context);
+        }
+
+        g_free(task->thread);
+    }
+
     if (task->destroy) {
         task->destroy(task->opaque);
     }
@@ -72,31 +93,12 @@ static void qio_task_free(QIOTask *task)
 }
 
 
-struct QIOTaskThreadData {
-    QIOTask *task;
-    QIOTaskWorker worker;
-    gpointer opaque;
-    GDestroyNotify destroy;
-    GMainContext *context;
-};
-
-
 static gboolean qio_task_thread_result(gpointer opaque)
 {
-    struct QIOTaskThreadData *data = opaque;
+    QIOTask *task = opaque;
 
-    trace_qio_task_thread_result(data->task);
-    qio_task_complete(data->task);
-
-    if (data->destroy) {
-        data->destroy(data->opaque);
-    }
-
-    if (data->context) {
-        g_main_context_unref(data->context);
-    }
-
-    g_free(data);
+    trace_qio_task_thread_result(task);
+    qio_task_complete(task);
 
     return FALSE;
 }
@@ -104,22 +106,23 @@ static gboolean qio_task_thread_result(gpointer opaque)
 
 static gpointer qio_task_thread_worker(gpointer opaque)
 {
-    struct QIOTaskThreadData *data = opaque;
+    QIOTask *task = opaque;
     GSource *idle;
 
-    trace_qio_task_thread_run(data->task);
-    data->worker(data->task, data->opaque);
+    trace_qio_task_thread_run(task);
+
+    task->thread->worker(task, task->thread->opaque);
 
     /* We're running in the background thread, and must only
      * ever report the task results in the main event loop
      * thread. So we schedule an idle callback to report
      * the worker results
      */
-    trace_qio_task_thread_exit(data->task);
+    trace_qio_task_thread_exit(task);
 
     idle = g_idle_source_new();
-    g_source_set_callback(idle, qio_task_thread_result, data, NULL);
-    g_source_attach(idle, data->context);
+    g_source_set_callback(idle, qio_task_thread_result, task, NULL);
+    g_source_attach(idle, task->thread->context);
 
     return NULL;
 }
@@ -138,17 +141,18 @@ void qio_task_run_in_thread(QIOTask *task,
         g_main_context_ref(context);
     }
 
-    data->task = task;
     data->worker = worker;
     data->opaque = opaque;
     data->destroy = destroy;
     data->context = context;
 
+    task->thread = data;
+
     trace_qio_task_thread_start(task, worker, opaque);
     qemu_thread_create(&thread,
                        "io-task-worker",
                        qio_task_thread_worker,
-                       data,
+                       task,
                        QEMU_THREAD_DETACHED);
 }
 
-- 
2.20.1

[Qemu-devel] [PATCH v3 02/16] io: add qio_task_wait_thread to join with a background thread

[Daniel P. Berrangé]

Add the ability for a caller to wait for completion of the
background thread to synchronously dispatch its result, without
needing to wait for the main loop to run the idle callback.

This method needs very careful usage to avoid a dangerous
race condition with the free'ing of the task. The completion
callback is normally invoked from an idle callback registered
with the main loop context. The qio_task_wait_thread method
must only be called if the completion callback has not yet
run. The only safe way to achieve this is to run the
qio_task_wait_thread method from the thread that executes
the main loop.

It is generally a bad idea to use this method since it will
block execution of the main loop, however, the design of
the character devices and its usage from vhostuser already
requires blocking execution.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 include/io/task.h | 29 ++++++++++++++++++++++++++++-
 io/task.c         | 41 +++++++++++++++++++++++++++++++++++++----
 io/trace-events   |  2 ++
 3 files changed, 67 insertions(+), 5 deletions(-)

diff --git a/include/io/task.h b/include/io/task.h
index 9e09b95b2e..57d8ba835e 100644
--- a/include/io/task.h
+++ b/include/io/task.h
@@ -232,7 +232,8 @@ QIOTask *qio_task_new(Object *source,
  *
  * Run a task in a background thread. When @worker
  * returns it will call qio_task_complete() in
- * the event thread context that provided.
+ * the thread that is running the main loop associated
+ * with @context.
  */
 void qio_task_run_in_thread(QIOTask *task,
                             QIOTaskWorker worker,
@@ -240,6 +241,32 @@ void qio_task_run_in_thread(QIOTask *task,
                             GDestroyNotify destroy,
                             GMainContext *context);
 
+
+/**
+ * qio_task_wait_thread:
+ * @task: the task struct
+ *
+ * Wait for completion of a task that was previously
+ * invoked using qio_task_run_in_thread. This MUST
+ * ONLY be invoked if the task has not already
+ * completed, since after the completion callback
+ * is invoked, @task will have been freed.
+ *
+ * To avoid racing with execution of the completion
+ * callback provided with qio_task_new, this method
+ * MUST ONLY be invoked from the thread that is
+ * running the main loop associated with @context
+ * parameter to qio_task_run_in_thread.
+ *
+ * When the thread has completed, the completion
+ * callback provided to qio_task_new will be invoked.
+ * When that callback returns @task will be freed,
+ * so @task must not be referenced after this
+ * method completes.
+ */
+void qio_task_wait_thread(QIOTask *task);
+
+
 /**
  * qio_task_complete:
  * @task: the task struct
diff --git a/io/task.c b/io/task.c
index 396866b10f..64c4c7126a 100644
--- a/io/task.c
+++ b/io/task.c
@@ -29,6 +29,7 @@ struct QIOTaskThreadData {
     gpointer opaque;
     GDestroyNotify destroy;
     GMainContext *context;
+    GSource *completion;
 };
 
 
@@ -40,6 +41,8 @@ struct QIOTask {
     Error *err;
     gpointer result;
     GDestroyNotify destroyResult;
+    QemuMutex thread_lock;
+    QemuCond thread_cond;
     struct QIOTaskThreadData *thread;
 };
 
@@ -58,6 +61,8 @@ QIOTask *qio_task_new(Object *source,
     task->func = func;
     task->opaque = opaque;
     task->destroy = destroy;
+    qemu_mutex_init(&task->thread_lock);
+    qemu_cond_init(&task->thread_cond);
 
     trace_qio_task_new(task, source, func, opaque);
 
@@ -66,6 +71,7 @@ QIOTask *qio_task_new(Object *source,
 
 static void qio_task_free(QIOTask *task)
 {
+    qemu_mutex_lock(&task->thread_lock);
     if (task->thread) {
         if (task->thread->destroy) {
             task->thread->destroy(task->thread->opaque);
@@ -89,6 +95,10 @@ static void qio_task_free(QIOTask *task)
     }
     object_unref(task->source);
 
+    qemu_mutex_unlock(&task->thread_lock);
+    qemu_mutex_destroy(&task->thread_lock);
+    qemu_cond_destroy(&task->thread_cond);
+
     g_free(task);
 }
 
@@ -107,7 +117,6 @@ static gboolean qio_task_thread_result(gpointer opaque)
 static gpointer qio_task_thread_worker(gpointer opaque)
 {
     QIOTask *task = opaque;
-    GSource *idle;
 
     trace_qio_task_thread_run(task);
 
@@ -120,9 +129,17 @@ static gpointer qio_task_thread_worker(gpointer opaque)
      */
     trace_qio_task_thread_exit(task);
 
-    idle = g_idle_source_new();
-    g_source_set_callback(idle, qio_task_thread_result, task, NULL);
-    g_source_attach(idle, task->thread->context);
+    qemu_mutex_lock(&task->thread_lock);
+
+    task->thread->completion = g_idle_source_new();
+    g_source_set_callback(task->thread->completion,
+                          qio_task_thread_result, task, NULL);
+    g_source_attach(task->thread->completion,
+                    task->thread->context);
+    trace_qio_task_thread_source_attach(task, task->thread->completion);
+
+    qemu_cond_signal(&task->thread_cond);
+    qemu_mutex_unlock(&task->thread_lock);
 
     return NULL;
 }
@@ -157,6 +174,22 @@ void qio_task_run_in_thread(QIOTask *task,
 }
 
 
+void qio_task_wait_thread(QIOTask *task)
+{
+    qemu_mutex_lock(&task->thread_lock);
+    g_assert(task->thread != NULL);
+    while (task->thread->completion == NULL) {
+        qemu_cond_wait(&task->thread_cond, &task->thread_lock);
+    }
+
+    trace_qio_task_thread_source_cancel(task, task->thread->completion);
+    g_source_destroy(task->thread->completion);
+    qemu_mutex_unlock(&task->thread_lock);
+
+    qio_task_thread_result(task);
+}
+
+
 void qio_task_complete(QIOTask *task)
 {
     task->func(task, task->opaque);
diff --git a/io/trace-events b/io/trace-events
index f70bad7cbe..07a7bbec6a 100644
--- a/io/trace-events
+++ b/io/trace-events
@@ -7,6 +7,8 @@ qio_task_thread_start(void *task, void *worker, void *opaque) "Task thread start
 qio_task_thread_run(void *task) "Task thread run task=%p"
 qio_task_thread_exit(void *task) "Task thread exit task=%p"
 qio_task_thread_result(void *task) "Task thread result task=%p"
+qio_task_thread_source_attach(void *task, void *source) "Task thread source attach task=%p source=%p"
+qio_task_thread_source_cancel(void *task, void *source) "Task thread source cancel task=%p source=%p"
 
 # io/channel-socket.c
 qio_channel_socket_new(void *ioc) "Socket new ioc=%p"
-- 
2.20.1

[Qemu-devel] [PATCH v3 03/16] chardev: fix validation of options for QMP created chardevs

[Daniel P. Berrangé]

The TLS creds option is not valid with certain address types. The user
config was only checked for errors when parsing legacy QemuOpts, thus
the user could pass unsupported values via QMP.

Pull all code for validating options out into a new method
qmp_chardev_validate_socket, that is called from the main
qmp_chardev_open_socket method. This adds a missing check for rejecting
TLS creds with the vsock address type.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 92 +++++++++++++++++++++++++++++++------------
 1 file changed, 66 insertions(+), 26 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index eaa8e8b68f..e85250b624 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -987,6 +987,65 @@ static gboolean socket_reconnect_timeout(gpointer opaque)
     return false;
 }
 
+
+static bool qmp_chardev_validate_socket(ChardevSocket *sock,
+                                        SocketAddress *addr,
+                                        Error **errp)
+{
+    /* Validate any options which have a dependency on address type */
+    switch (addr->type) {
+    case SOCKET_ADDRESS_TYPE_FD:
+        if (sock->has_reconnect) {
+            error_setg(errp,
+                       "'reconnect' option is incompatible with "
+                       "'fd' address type");
+            return false;
+        }
+        if (sock->has_tls_creds &&
+            !(sock->has_server && sock->server)) {
+            error_setg(errp,
+                       "'tls_creds' option is incompatible with "
+                       "'fd' address type as client");
+            return false;
+        }
+        break;
+
+    case SOCKET_ADDRESS_TYPE_UNIX:
+        if (sock->has_tls_creds) {
+            error_setg(errp,
+                       "'tls_creds' option is incompatible with "
+                       "'unix' address type");
+            return false;
+        }
+        break;
+
+    case SOCKET_ADDRESS_TYPE_INET:
+        break;
+
+    case SOCKET_ADDRESS_TYPE_VSOCK:
+        if (sock->has_tls_creds) {
+            error_setg(errp,
+                       "'tls_creds' option is incompatible with "
+                       "'vsock' address type");
+            return false;
+        }
+
+    default:
+        break;
+    }
+
+    /* Validate any options which have a dependancy on client vs server */
+    if (!(sock->has_server && sock->server)) {
+        if (sock->has_websocket && sock->websocket) {
+            error_setg(errp, "%s", "Websocket client is not implemented");
+            return false;
+        }
+    }
+
+    return true;
+}
+
+
 static void qmp_chardev_open_socket(Chardev *chr,
                                     ChardevBackend *backend,
                                     bool *be_opened,
@@ -1004,11 +1063,6 @@ static void qmp_chardev_open_socket(Chardev *chr,
     QIOChannelSocket *sioc = NULL;
     SocketAddress *addr;
 
-    if (!is_listen && is_websock) {
-        error_setg(errp, "%s", "Websocket client is not implemented");
-        goto error;
-    }
-
     s->is_listen = is_listen;
     s->is_telnet = is_telnet;
     s->is_tn3270 = is_tn3270;
@@ -1049,10 +1103,10 @@ static void qmp_chardev_open_socket(Chardev *chr,
 
     s->addr = addr = socket_address_flatten(sock->addr);
 
-    if (sock->has_reconnect && addr->type == SOCKET_ADDRESS_TYPE_FD) {
-        error_setg(errp, "'reconnect' option is incompatible with 'fd'");
+    if (!qmp_chardev_validate_socket(sock, addr, errp)) {
         goto error;
     }
+
     qemu_chr_set_feature(chr, QEMU_CHAR_FEATURE_RECONNECTABLE);
     /* TODO SOCKET_ADDRESS_FD where fd has AF_UNIX */
     if (addr->type == SOCKET_ADDRESS_TYPE_UNIX) {
@@ -1140,27 +1194,12 @@ static void qemu_chr_parse_socket(QemuOpts *opts, ChardevBackend *backend,
         return;
     }
 
-    backend->type = CHARDEV_BACKEND_KIND_SOCKET;
-    if (path) {
-        if (tls_creds) {
-            error_setg(errp, "TLS can only be used over TCP socket");
-            return;
-        }
-    } else if (host) {
-        if (!port) {
-            error_setg(errp, "chardev: socket: no port given");
-            return;
-        }
-    } else if (fd) {
-        /* We don't know what host to validate against when in client mode */
-        if (tls_creds && !is_listen) {
-            error_setg(errp, "TLS can not be used with pre-opened client FD");
-            return;
-        }
-    } else {
-        g_assert_not_reached();
+    if (host && !port) {
+        error_setg(errp, "chardev: socket: no port given");
+        return;
     }
 
+    backend->type = CHARDEV_BACKEND_KIND_SOCKET;
     sock = backend->u.socket.data = g_new0(ChardevSocket, 1);
     qemu_chr_parse_common(opts, qapi_ChardevSocket_base(sock));
 
@@ -1178,6 +1217,7 @@ static void qemu_chr_parse_socket(QemuOpts *opts, ChardevBackend *backend,
     sock->wait = is_waitconnect;
     sock->has_reconnect = qemu_opt_find(opts, "reconnect");
     sock->reconnect = reconnect;
+    sock->has_tls_creds = tls_creds;
     sock->tls_creds = g_strdup(tls_creds);
 
     addr = g_new0(SocketAddressLegacy, 1);
-- 
2.20.1

[Qemu-devel] [PATCH v3 04/16] chardev: forbid 'reconnect' option with server sockets

[Daniel P. Berrangé]

The 'reconnect' option is used to give the sleep time, in seconds,
before a client socket attempts to re-establish a connection to the
server. It does not make sense to set this for server sockets, as they
will always accept a new client connection immediately after the
previous one went away.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index e85250b624..743b7b11cd 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -1035,7 +1035,14 @@ static bool qmp_chardev_validate_socket(ChardevSocket *sock,
     }
 
     /* Validate any options which have a dependancy on client vs server */
-    if (!(sock->has_server && sock->server)) {
+    if (!sock->has_server || sock->server) {
+        if (sock->has_reconnect) {
+            error_setg(errp,
+                       "'reconnect' option is incompatible with "
+                       "socket in server listen mode");
+            return false;
+        }
+    } else {
         if (sock->has_websocket && sock->websocket) {
             error_setg(errp, "%s", "Websocket client is not implemented");
             return false;
-- 
2.20.1

[Qemu-devel] [PATCH v3 05/16] chardev: forbid 'wait' option with client sockets

[Daniel P. Berrangé]

The 'wait'/'nowait' parameter is used to tell server sockets whether to
block until a client is accepted during initialization. Client chardevs
have always silently ignored this option. Various tests were mistakenly
passing this option for their client chardevs.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c          | 12 +++++++++++-
 tests/ivshmem-test.c           |  2 +-
 tests/libqtest.c               |  4 ++--
 tests/test-filter-redirector.c |  4 ++--
 4 files changed, 16 insertions(+), 6 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 743b7b11cd..728342dc9f 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -1047,6 +1047,12 @@ static bool qmp_chardev_validate_socket(ChardevSocket *sock,
             error_setg(errp, "%s", "Websocket client is not implemented");
             return false;
         }
+        if (sock->has_wait) {
+            error_setg(errp, "%s",
+                       "'wait' option is incompatible with "
+                       "socket in client connect mode");
+            return false;
+        }
     }
 
     return true;
@@ -1220,7 +1226,11 @@ static void qemu_chr_parse_socket(QemuOpts *opts, ChardevBackend *backend,
     sock->tn3270 = is_tn3270;
     sock->has_websocket = true;
     sock->websocket = is_websock;
-    sock->has_wait = true;
+    /*
+     * We have different default to QMP for 'wait' when 'server'
+     * is set, hence we can't just check for existence of 'wait'
+     */
+    sock->has_wait = qemu_opt_find(opts, "wait") || is_listen;
     sock->wait = is_waitconnect;
     sock->has_reconnect = qemu_opt_find(opts, "reconnect");
     sock->reconnect = reconnect;
diff --git a/tests/ivshmem-test.c b/tests/ivshmem-test.c
index 4911b69317..942ddc9192 100644
--- a/tests/ivshmem-test.c
+++ b/tests/ivshmem-test.c
@@ -295,7 +295,7 @@ static void setup_vm_with_server(IVState *s, int nvectors)
 {
     char *cmd;
 
-    cmd = g_strdup_printf("-chardev socket,id=chr0,path=%s,nowait "
+    cmd = g_strdup_printf("-chardev socket,id=chr0,path=%s "
                           "-device ivshmem-doorbell,chardev=chr0,vectors=%d",
                           tmpserver, nvectors);
 
diff --git a/tests/libqtest.c b/tests/libqtest.c
index 6fb30855fa..c49b85482d 100644
--- a/tests/libqtest.c
+++ b/tests/libqtest.c
@@ -232,9 +232,9 @@ QTestState *qtest_init_without_qmp_handshake(const char *extra_args)
     qtest_add_abrt_handler(kill_qemu_hook_func, s);
 
     command = g_strdup_printf("exec %s "
-                              "-qtest unix:%s,nowait "
+                              "-qtest unix:%s "
                               "-qtest-log %s "
-                              "-chardev socket,path=%s,nowait,id=char0 "
+                              "-chardev socket,path=%s,id=char0 "
                               "-mon chardev=char0,mode=control "
                               "-machine accel=qtest "
                               "-display none "
diff --git a/tests/test-filter-redirector.c b/tests/test-filter-redirector.c
index 9ca9feabf8..6dc21dd4fb 100644
--- a/tests/test-filter-redirector.c
+++ b/tests/test-filter-redirector.c
@@ -96,7 +96,7 @@ static void test_redirector_tx(void)
         "-device %s,netdev=qtest-bn0,id=qtest-e0 "
         "-chardev socket,id=redirector0,path=%s,server,nowait "
         "-chardev socket,id=redirector1,path=%s,server,nowait "
-        "-chardev socket,id=redirector2,path=%s,nowait "
+        "-chardev socket,id=redirector2,path=%s "
         "-object filter-redirector,id=qtest-f0,netdev=qtest-bn0,"
         "queue=tx,outdev=redirector0 "
         "-object filter-redirector,id=qtest-f1,netdev=qtest-bn0,"
@@ -166,7 +166,7 @@ static void test_redirector_rx(void)
         "-device %s,netdev=qtest-bn0,id=qtest-e0 "
         "-chardev socket,id=redirector0,path=%s,server,nowait "
         "-chardev socket,id=redirector1,path=%s,server,nowait "
-        "-chardev socket,id=redirector2,path=%s,nowait "
+        "-chardev socket,id=redirector2,path=%s "
         "-object filter-redirector,id=qtest-f0,netdev=qtest-bn0,"
         "queue=rx,indev=redirector0 "
         "-object filter-redirector,id=qtest-f1,netdev=qtest-bn0,"
-- 
2.20.1

[Qemu-devel] [PATCH v3 06/16] chardev: remove many local variables in qemu_chr_parse_socket

[Daniel P. Berrangé]

Now that all validation is separated off into a separate method,
we can directly populate the ChardevSocket struct from the
QemuOpts values, avoiding many local variables.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 40 ++++++++++++++++++----------------------
 1 file changed, 18 insertions(+), 22 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 728342dc9f..8a6e203da7 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -1186,18 +1186,10 @@ error:
 static void qemu_chr_parse_socket(QemuOpts *opts, ChardevBackend *backend,
                                   Error **errp)
 {
-    bool is_listen      = qemu_opt_get_bool(opts, "server", false);
-    bool is_waitconnect = is_listen && qemu_opt_get_bool(opts, "wait", true);
-    bool is_telnet      = qemu_opt_get_bool(opts, "telnet", false);
-    bool is_tn3270      = qemu_opt_get_bool(opts, "tn3270", false);
-    bool is_websock     = qemu_opt_get_bool(opts, "websocket", false);
-    bool do_nodelay     = !qemu_opt_get_bool(opts, "delay", true);
-    int64_t reconnect   = qemu_opt_get_number(opts, "reconnect", 0);
     const char *path = qemu_opt_get(opts, "path");
     const char *host = qemu_opt_get(opts, "host");
     const char *port = qemu_opt_get(opts, "port");
     const char *fd = qemu_opt_get(opts, "fd");
-    const char *tls_creds = qemu_opt_get(opts, "tls-creds");
     SocketAddressLegacy *addr;
     ChardevSocket *sock;
 
@@ -1216,26 +1208,30 @@ static void qemu_chr_parse_socket(QemuOpts *opts, ChardevBackend *backend,
     sock = backend->u.socket.data = g_new0(ChardevSocket, 1);
     qemu_chr_parse_common(opts, qapi_ChardevSocket_base(sock));
 
-    sock->has_nodelay = true;
-    sock->nodelay = do_nodelay;
+    sock->has_nodelay = qemu_opt_get(opts, "delay");
+    sock->nodelay = !qemu_opt_get_bool(opts, "delay", true);
+    /*
+     * We have different default to QMP for 'server', hence
+     * we can't just check for existence of 'server'
+     */
     sock->has_server = true;
-    sock->server = is_listen;
-    sock->has_telnet = true;
-    sock->telnet = is_telnet;
-    sock->has_tn3270 = true;
-    sock->tn3270 = is_tn3270;
-    sock->has_websocket = true;
-    sock->websocket = is_websock;
+    sock->server = qemu_opt_get_bool(opts, "server", false);
+    sock->has_telnet = qemu_opt_get(opts, "telnet");
+    sock->telnet = qemu_opt_get_bool(opts, "telnet", false);
+    sock->has_tn3270 = qemu_opt_get(opts, "tn3270");
+    sock->tn3270 = qemu_opt_get_bool(opts, "tn3270", false);
+    sock->has_websocket = qemu_opt_get(opts, "websocket");
+    sock->websocket = qemu_opt_get_bool(opts, "websocket", false);
     /*
      * We have different default to QMP for 'wait' when 'server'
      * is set, hence we can't just check for existence of 'wait'
      */
-    sock->has_wait = qemu_opt_find(opts, "wait") || is_listen;
-    sock->wait = is_waitconnect;
+    sock->has_wait = qemu_opt_find(opts, "wait") || sock->server;
+    sock->wait = qemu_opt_get_bool(opts, "wait", true);
     sock->has_reconnect = qemu_opt_find(opts, "reconnect");
-    sock->reconnect = reconnect;
-    sock->has_tls_creds = tls_creds;
-    sock->tls_creds = g_strdup(tls_creds);
+    sock->reconnect = qemu_opt_get_number(opts, "reconnect", 0);
+    sock->has_tls_creds = qemu_opt_get(opts, "tls-creds");
+    sock->tls_creds = g_strdup(qemu_opt_get(opts, "tls-creds"));
 
     addr = g_new0(SocketAddressLegacy, 1);
     if (path) {
-- 
2.20.1

[Qemu-devel] [PATCH v3 07/16] chardev: ensure qemu_chr_parse_compat reports missing driver error

[Daniel P. Berrangé]

If no valid char driver was identified the qemu_chr_parse_compat method
was silent, leaving callers no clue what failed.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char.c    | 2 ++
 tests/test-char.c | 3 ++-
 2 files changed, 4 insertions(+), 1 deletion(-)

diff --git a/chardev/char.c b/chardev/char.c
index ccba36bafb..b99f3692f7 100644
--- a/chardev/char.c
+++ b/chardev/char.c
@@ -490,6 +490,8 @@ QemuOpts *qemu_chr_parse_compat(const char *label, const char *filename,
         return opts;
     }
 
+    error_report("'%s' is not a valid char driver", filename);
+
 fail:
     qemu_opts_del(opts);
     return NULL;
diff --git a/tests/test-char.c b/tests/test-char.c
index 19c3efad72..89c43e4ada 100644
--- a/tests/test-char.c
+++ b/tests/test-char.c
@@ -856,9 +856,10 @@ static void char_null_test(void)
 static void char_invalid_test(void)
 {
     Chardev *chr;
-
+    g_setenv("QTEST_SILENT_ERRORS", "1", 1);
     chr = qemu_chr_new("label-invalid", "invalid");
     g_assert_null(chr);
+    g_unsetenv("QTEST_SILENT_ERRORS");
 }
 
 static int chardev_change(void *opaque)
-- 
2.20.1

[Qemu-devel] [PATCH v3 08/16] chardev: remove unused 'sioc' variable & cleanup paths

[Daniel P. Berrangé]

The 'sioc' variable in qmp_chardev_open_socket was unused since

  commit 3e7d4d20d3a528b1ed10b1dc3d83119bfb0c5f24
  Author: Peter Xu <peterx@redhat.com>
  Date:   Tue Mar 6 13:33:17 2018 +0800

    chardev: use chardev's gcontext for async connect

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 22 +++++++---------------
 1 file changed, 7 insertions(+), 15 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 8a6e203da7..8a5e5c2fe7 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -1073,7 +1073,6 @@ static void qmp_chardev_open_socket(Chardev *chr,
     bool is_waitconnect = sock->has_wait    ? sock->wait    : false;
     bool is_websock     = sock->has_websocket ? sock->websocket : false;
     int64_t reconnect   = sock->has_reconnect ? sock->reconnect : 0;
-    QIOChannelSocket *sioc = NULL;
     SocketAddress *addr;
 
     s->is_listen = is_listen;
@@ -1088,7 +1087,7 @@ static void qmp_chardev_open_socket(Chardev *chr,
         if (!creds) {
             error_setg(errp, "No TLS credentials with id '%s'",
                        sock->tls_creds);
-            goto error;
+            return;
         }
         s->tls_creds = (QCryptoTLSCreds *)
             object_dynamic_cast(creds,
@@ -1096,20 +1095,20 @@ static void qmp_chardev_open_socket(Chardev *chr,
         if (!s->tls_creds) {
             error_setg(errp, "Object with id '%s' is not TLS credentials",
                        sock->tls_creds);
-            goto error;
+            return;
         }
         object_ref(OBJECT(s->tls_creds));
         if (is_listen) {
             if (s->tls_creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
                 error_setg(errp, "%s",
                            "Expected TLS credentials for server endpoint");
-                goto error;
+                return;
             }
         } else {
             if (s->tls_creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT) {
                 error_setg(errp, "%s",
                            "Expected TLS credentials for client endpoint");
-                goto error;
+                return;
             }
         }
     }
@@ -1117,7 +1116,7 @@ static void qmp_chardev_open_socket(Chardev *chr,
     s->addr = addr = socket_address_flatten(sock->addr);
 
     if (!qmp_chardev_validate_socket(sock, addr, errp)) {
-        goto error;
+        return;
     }
 
     qemu_chr_set_feature(chr, QEMU_CHAR_FEATURE_RECONNECTABLE);
@@ -1153,7 +1152,7 @@ static void qmp_chardev_open_socket(Chardev *chr,
             if (qio_net_listener_open_sync(s->listener, s->addr, errp) < 0) {
                 object_unref(OBJECT(s->listener));
                 s->listener = NULL;
-                goto error;
+                return;
             }
 
             qapi_free_SocketAddress(s->addr);
@@ -1171,16 +1170,9 @@ static void qmp_chardev_open_socket(Chardev *chr,
                                                       chr->gcontext);
             }
         } else if (qemu_chr_wait_connected(chr, errp) < 0) {
-            goto error;
+            return;
         }
     }
-
-    return;
-
-error:
-    if (sioc) {
-        object_unref(OBJECT(sioc));
-    }
 }
 
 static void qemu_chr_parse_socket(QemuOpts *opts, ChardevBackend *backend,
-- 
2.20.1

[Qemu-devel] [PATCH v3 09/16] chardev: split tcp_chr_wait_connected into two methods

[Daniel P. Berrangé]

The tcp_chr_wait_connected method can deal with either server or client
chardevs, but some callers only care about one of these possibilities.
The tcp_chr_wait_connected method will also need some refactoring to
reliably deal with its primary goal of allowing a device frontend to
wait for an established connection, which will interfere with other
callers.

Split it into two methods, one responsible for server initiated
connections, the other responsible for client initiated connections.
In doing this split the tcp_char_connect_async() method is renamed
to become consistent with naming of the new methods.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 59 +++++++++++++++++++++++++++----------------
 1 file changed, 37 insertions(+), 22 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 8a5e5c2fe7..222adbbad3 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -886,30 +886,47 @@ static void tcp_chr_accept(QIONetListener *listener,
     tcp_chr_new_client(chr, cioc);
 }
 
-static int tcp_chr_wait_connected(Chardev *chr, Error **errp)
+
+static int tcp_chr_connect_client_sync(Chardev *chr, Error **errp)
+{
+    SocketChardev *s = SOCKET_CHARDEV(chr);
+    QIOChannelSocket *sioc = qio_channel_socket_new();
+    tcp_chr_set_client_ioc_name(chr, sioc);
+    if (qio_channel_socket_connect_sync(sioc, s->addr, errp) < 0) {
+        object_unref(OBJECT(sioc));
+        return -1;
+    }
+    tcp_chr_new_client(chr, sioc);
+    object_unref(OBJECT(sioc));
+    return 0;
+}
+
+
+static void tcp_chr_accept_server_sync(Chardev *chr)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
     QIOChannelSocket *sioc;
+    info_report("QEMU waiting for connection on: %s",
+                chr->filename);
+    sioc = qio_net_listener_wait_client(s->listener);
+    tcp_chr_set_client_ioc_name(chr, sioc);
+    tcp_chr_new_client(chr, sioc);
+    object_unref(OBJECT(sioc));
+}
+
 
+static int tcp_chr_wait_connected(Chardev *chr, Error **errp)
+{
+    SocketChardev *s = SOCKET_CHARDEV(chr);
     /* It can't wait on s->connected, since it is set asynchronously
      * in TLS and telnet cases, only wait for an accepted socket */
     while (!s->ioc) {
         if (s->is_listen) {
-            info_report("QEMU waiting for connection on: %s",
-                        chr->filename);
-            sioc = qio_net_listener_wait_client(s->listener);
-            tcp_chr_set_client_ioc_name(chr, sioc);
-            tcp_chr_new_client(chr, sioc);
-            object_unref(OBJECT(sioc));
+            tcp_chr_accept_server_sync(chr);
         } else {
-            sioc = qio_channel_socket_new();
-            tcp_chr_set_client_ioc_name(chr, sioc);
-            if (qio_channel_socket_connect_sync(sioc, s->addr, errp) < 0) {
-                object_unref(OBJECT(sioc));
+            if (tcp_chr_connect_client_sync(chr, errp) < 0) {
                 return -1;
             }
-            tcp_chr_new_client(chr, sioc);
-            object_unref(OBJECT(sioc));
         }
     }
 
@@ -958,7 +975,7 @@ cleanup:
     object_unref(OBJECT(sioc));
 }
 
-static void tcp_chr_connect_async(Chardev *chr)
+static void tcp_chr_connect_client_async(Chardev *chr)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
     QIOChannelSocket *sioc;
@@ -982,7 +999,7 @@ static gboolean socket_reconnect_timeout(gpointer opaque)
         return false;
     }
 
-    tcp_chr_connect_async(chr);
+    tcp_chr_connect_client_async(chr);
 
     return false;
 }
@@ -1139,7 +1156,7 @@ static void qmp_chardev_open_socket(Chardev *chr,
     }
 
     if (s->reconnect_time) {
-        tcp_chr_connect_async(chr);
+        tcp_chr_connect_client_async(chr);
     } else {
         if (s->is_listen) {
             char *name;
@@ -1159,17 +1176,15 @@ static void qmp_chardev_open_socket(Chardev *chr,
             s->addr = socket_local_address(s->listener->sioc[0]->fd, errp);
             update_disconnected_filename(s);
 
-            if (is_waitconnect &&
-                qemu_chr_wait_connected(chr, errp) < 0) {
-                return;
-            }
-            if (!s->ioc) {
+            if (is_waitconnect) {
+                tcp_chr_accept_server_sync(chr);
+            } else {
                 qio_net_listener_set_client_func_full(s->listener,
                                                       tcp_chr_accept,
                                                       chr, NULL,
                                                       chr->gcontext);
             }
-        } else if (qemu_chr_wait_connected(chr, errp) < 0) {
+        } else if (tcp_chr_connect_client_sync(chr, errp) < 0) {
             return;
         }
     }
-- 
2.20.1

[Qemu-devel] [PATCH v3 10/16] chardev: split up qmp_chardev_open_socket connection code

[Daniel P. Berrangé]

In qmp_chardev_open_socket the code for connecting client chardevs is
split across two conditionals far apart with some server chardev code in
the middle. Split up the method so that code for client connection setup
is separate from code for server connection setup.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 96 +++++++++++++++++++++++++++----------------
 1 file changed, 60 insertions(+), 36 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 222adbbad3..90dafef7d4 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -1005,6 +1005,61 @@ static gboolean socket_reconnect_timeout(gpointer opaque)
 }
 
 
+static int qmp_chardev_open_socket_server(Chardev *chr,
+                                          bool is_telnet,
+                                          bool is_waitconnect,
+                                          Error **errp)
+{
+    SocketChardev *s = SOCKET_CHARDEV(chr);
+    char *name;
+    if (is_telnet) {
+        s->do_telnetopt = 1;
+    }
+    s->listener = qio_net_listener_new();
+
+    name = g_strdup_printf("chardev-tcp-listener-%s", chr->label);
+    qio_net_listener_set_name(s->listener, name);
+    g_free(name);
+
+    if (qio_net_listener_open_sync(s->listener, s->addr, errp) < 0) {
+        object_unref(OBJECT(s->listener));
+        s->listener = NULL;
+        return -1;
+    }
+
+    qapi_free_SocketAddress(s->addr);
+    s->addr = socket_local_address(s->listener->sioc[0]->fd, errp);
+    update_disconnected_filename(s);
+
+    if (is_waitconnect) {
+        tcp_chr_accept_server_sync(chr);
+    } else {
+        qio_net_listener_set_client_func_full(s->listener,
+                                              tcp_chr_accept,
+                                              chr, NULL,
+                                              chr->gcontext);
+    }
+
+    return 0;
+}
+
+
+static int qmp_chardev_open_socket_client(Chardev *chr,
+                                          int64_t reconnect,
+                                          Error **errp)
+{
+    SocketChardev *s = SOCKET_CHARDEV(chr);
+
+    if (reconnect > 0) {
+        s->reconnect_time = reconnect;
+        tcp_chr_connect_client_async(chr);
+        return 0;
+    } else {
+        return tcp_chr_connect_client_sync(chr, errp);
+    }
+}
+
+
 static bool qmp_chardev_validate_socket(ChardevSocket *sock,
                                         SocketAddress *addr,
                                         Error **errp)
@@ -1147,44 +1202,13 @@ static void qmp_chardev_open_socket(Chardev *chr,
 
     update_disconnected_filename(s);
 
-    if (is_listen) {
-        if (is_telnet || is_tn3270) {
-            s->do_telnetopt = 1;
+    if (s->is_listen) {
+        if (qmp_chardev_open_socket_server(chr, is_telnet || is_tn3270,
+                                           is_waitconnect, errp) < 0) {
+            return;
         }
-    } else if (reconnect > 0) {
-        s->reconnect_time = reconnect;
-    }
-
-    if (s->reconnect_time) {
-        tcp_chr_connect_client_async(chr);
     } else {
-        if (s->is_listen) {
-            char *name;
-            s->listener = qio_net_listener_new();
-
-            name = g_strdup_printf("chardev-tcp-listener-%s", chr->label);
-            qio_net_listener_set_name(s->listener, name);
-            g_free(name);
-
-            if (qio_net_listener_open_sync(s->listener, s->addr, errp) < 0) {
-                object_unref(OBJECT(s->listener));
-                s->listener = NULL;
-                return;
-            }
-
-            qapi_free_SocketAddress(s->addr);
-            s->addr = socket_local_address(s->listener->sioc[0]->fd, errp);
-            update_disconnected_filename(s);
-
-            if (is_waitconnect) {
-                tcp_chr_accept_server_sync(chr);
-            } else {
-                qio_net_listener_set_client_func_full(s->listener,
-                                                      tcp_chr_accept,
-                                                      chr, NULL,
-                                                      chr->gcontext);
-            }
-        } else if (tcp_chr_connect_client_sync(chr, errp) < 0) {
+        if (qmp_chardev_open_socket_client(chr, reconnect, errp) < 0) {
             return;
         }
     }
-- 
2.20.1

[Qemu-devel] [PATCH v3 11/16] chardev: use a state machine for socket connection state

[Daniel P. Berrangé]

The socket connection state is indicated via the 'bool connected' field
in the SocketChardev struct. This variable is somewhat misleading
though, as it is only set to true once the connection has completed all
required handshakes (eg for TLS, telnet or websockets). IOW there is a
period of time in which the socket is connected, but the "connected"
flag is still false.

The socket chardev really has three states that it can be in,
disconnected, connecting and connected and those should be tracked
explicitly.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 63 +++++++++++++++++++++++++++++++++----------
 1 file changed, 49 insertions(+), 14 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 90dafef7d4..d6de5d2305 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -46,6 +46,12 @@ typedef struct {
     size_t buflen;
 } TCPChardevTelnetInit;
 
+typedef enum {
+    TCP_CHARDEV_STATE_DISCONNECTED,
+    TCP_CHARDEV_STATE_CONNECTING,
+    TCP_CHARDEV_STATE_CONNECTED,
+} TCPChardevState;
+
 typedef struct {
     Chardev parent;
     QIOChannel *ioc; /* Client I/O channel */
@@ -53,7 +59,7 @@ typedef struct {
     QIONetListener *listener;
     GSource *hup_source;
     QCryptoTLSCreds *tls_creds;
-    int connected;
+    TCPChardevState state;
     int max_size;
     int do_telnetopt;
     int do_nodelay;
@@ -82,6 +88,21 @@ typedef struct {
 static gboolean socket_reconnect_timeout(gpointer opaque);
 static void tcp_chr_telnet_init(Chardev *chr);
 
+static void tcp_chr_change_state(SocketChardev *s, TCPChardevState state)
+{
+    switch (state) {
+    case TCP_CHARDEV_STATE_DISCONNECTED:
+        break;
+    case TCP_CHARDEV_STATE_CONNECTING:
+        assert(s->state == TCP_CHARDEV_STATE_DISCONNECTED);
+        break;
+    case TCP_CHARDEV_STATE_CONNECTED:
+        assert(s->state == TCP_CHARDEV_STATE_CONNECTING);
+        break;
+    }
+    s->state = state;
+}
+
 static void tcp_chr_reconn_timer_cancel(SocketChardev *s)
 {
     if (s->reconnect_timer) {
@@ -96,7 +117,7 @@ static void qemu_chr_socket_restart_timer(Chardev *chr)
     SocketChardev *s = SOCKET_CHARDEV(chr);
     char *name;
 
-    assert(s->connected == 0);
+    assert(s->state == TCP_CHARDEV_STATE_DISCONNECTED);
     name = g_strdup_printf("chardev-socket-reconnect-%s", chr->label);
     s->reconnect_timer = qemu_chr_timeout_add_ms(chr,
                                                  s->reconnect_time * 1000,
@@ -131,7 +152,7 @@ static int tcp_chr_write(Chardev *chr, const uint8_t *buf, int len)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
 
-    if (s->connected) {
+    if (s->state == TCP_CHARDEV_STATE_CONNECTED) {
         int ret =  io_channel_send_full(s->ioc, buf, len,
                                         s->write_msgfds,
                                         s->write_msgfds_num);
@@ -164,7 +185,7 @@ static int tcp_chr_read_poll(void *opaque)
 {
     Chardev *chr = CHARDEV(opaque);
     SocketChardev *s = SOCKET_CHARDEV(opaque);
-    if (!s->connected) {
+    if (s->state != TCP_CHARDEV_STATE_CONNECTED) {
         return 0;
     }
     s->max_size = qemu_chr_be_can_write(chr);
@@ -277,7 +298,7 @@ static int tcp_set_msgfds(Chardev *chr, int *fds, int num)
     s->write_msgfds = NULL;
     s->write_msgfds_num = 0;
 
-    if (!s->connected ||
+    if ((s->state != TCP_CHARDEV_STATE_CONNECTED) ||
         !qio_channel_has_feature(s->ioc,
                                  QIO_CHANNEL_FEATURE_FD_PASS)) {
         return -1;
@@ -389,7 +410,7 @@ static void tcp_chr_free_connection(Chardev *chr)
     s->ioc = NULL;
     g_free(chr->filename);
     chr->filename = NULL;
-    s->connected = 0;
+    tcp_chr_change_state(s, TCP_CHARDEV_STATE_DISCONNECTED);
 }
 
 static const char *qemu_chr_socket_protocol(SocketChardev *s)
@@ -442,12 +463,12 @@ static void update_disconnected_filename(SocketChardev *s)
 
 /* NB may be called even if tcp_chr_connect has not been
  * reached, due to TLS or telnet initialization failure,
- * so can *not* assume s->connected == true
+ * so can *not* assume s->state == TCP_CHARDEV_STATE_CONNECTED
  */
 static void tcp_chr_disconnect(Chardev *chr)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
-    bool emit_close = s->connected;
+    bool emit_close = s->state == TCP_CHARDEV_STATE_CONNECTED;
 
     tcp_chr_free_connection(chr);
 
@@ -471,7 +492,8 @@ static gboolean tcp_chr_read(QIOChannel *chan, GIOCondition cond, void *opaque)
     uint8_t buf[CHR_READ_BUF_LEN];
     int len, size;
 
-    if (!s->connected || s->max_size <= 0) {
+    if ((s->state != TCP_CHARDEV_STATE_CONNECTED) ||
+        s->max_size <= 0) {
         return TRUE;
     }
     len = sizeof(buf);
@@ -508,7 +530,7 @@ static int tcp_chr_sync_read(Chardev *chr, const uint8_t *buf, int len)
     SocketChardev *s = SOCKET_CHARDEV(chr);
     int size;
 
-    if (!s->connected) {
+    if (s->state != TCP_CHARDEV_STATE_CONNECTED) {
         return 0;
     }
 
@@ -564,7 +586,7 @@ static void update_ioc_handlers(SocketChardev *s)
 {
     Chardev *chr = CHARDEV(s);
 
-    if (!s->connected) {
+    if (s->state != TCP_CHARDEV_STATE_CONNECTED) {
         return;
     }
 
@@ -589,7 +611,7 @@ static void tcp_chr_connect(void *opaque)
     g_free(chr->filename);
     chr->filename = qemu_chr_compute_filename(s);
 
-    s->connected = 1;
+    tcp_chr_change_state(s, TCP_CHARDEV_STATE_CONNECTED);
     update_ioc_handlers(s);
     qemu_chr_be_event(chr, CHR_EVENT_OPENED);
 }
@@ -828,7 +850,7 @@ static int tcp_chr_new_client(Chardev *chr, QIOChannelSocket *sioc)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
 
-    if (s->ioc != NULL) {
+    if (s->state != TCP_CHARDEV_STATE_CONNECTING) {
         return -1;
     }
 
@@ -865,11 +887,17 @@ static int tcp_chr_add_client(Chardev *chr, int fd)
 {
     int ret;
     QIOChannelSocket *sioc;
+    SocketChardev *s = SOCKET_CHARDEV(chr);
+
+    if (s->state != TCP_CHARDEV_STATE_DISCONNECTED) {
+        return -1;
+    }
 
     sioc = qio_channel_socket_new_fd(fd, NULL);
     if (!sioc) {
         return -1;
     }
+    tcp_chr_change_state(s, TCP_CHARDEV_STATE_CONNECTING);
     tcp_chr_set_client_ioc_name(chr, sioc);
     ret = tcp_chr_new_client(chr, sioc);
     object_unref(OBJECT(sioc));
@@ -881,7 +909,9 @@ static void tcp_chr_accept(QIONetListener *listener,
                            void *opaque)
 {
     Chardev *chr = CHARDEV(opaque);
+    SocketChardev *s = SOCKET_CHARDEV(chr);
 
+    tcp_chr_change_state(s, TCP_CHARDEV_STATE_CONNECTING);
     tcp_chr_set_client_ioc_name(chr, cioc);
     tcp_chr_new_client(chr, cioc);
 }
@@ -891,8 +921,10 @@ static int tcp_chr_connect_client_sync(Chardev *chr, Error **errp)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
     QIOChannelSocket *sioc = qio_channel_socket_new();
+    tcp_chr_change_state(s, TCP_CHARDEV_STATE_CONNECTING);
     tcp_chr_set_client_ioc_name(chr, sioc);
     if (qio_channel_socket_connect_sync(sioc, s->addr, errp) < 0) {
+        tcp_chr_change_state(s, TCP_CHARDEV_STATE_DISCONNECTED);
         object_unref(OBJECT(sioc));
         return -1;
     }
@@ -908,6 +940,7 @@ static void tcp_chr_accept_server_sync(Chardev *chr)
     QIOChannelSocket *sioc;
     info_report("QEMU waiting for connection on: %s",
                 chr->filename);
+    tcp_chr_change_state(s, TCP_CHARDEV_STATE_CONNECTING);
     sioc = qio_net_listener_wait_client(s->listener);
     tcp_chr_set_client_ioc_name(chr, sioc);
     tcp_chr_new_client(chr, sioc);
@@ -963,6 +996,7 @@ static void qemu_chr_socket_connected(QIOTask *task, void *opaque)
     Error *err = NULL;
 
     if (qio_task_propagate_error(task, &err)) {
+        tcp_chr_change_state(s, TCP_CHARDEV_STATE_DISCONNECTED);
         check_report_connect_error(chr, err);
         error_free(err);
         goto cleanup;
@@ -980,6 +1014,7 @@ static void tcp_chr_connect_client_async(Chardev *chr)
     SocketChardev *s = SOCKET_CHARDEV(chr);
     QIOChannelSocket *sioc;
 
+    tcp_chr_change_state(s, TCP_CHARDEV_STATE_CONNECTING);
     sioc = qio_channel_socket_new();
     tcp_chr_set_client_ioc_name(chr, sioc);
     qio_channel_socket_connect_async(sioc, s->addr,
@@ -1307,7 +1342,7 @@ char_socket_get_connected(Object *obj, Error **errp)
 {
     SocketChardev *s = SOCKET_CHARDEV(obj);
 
-    return s->connected;
+    return s->state == TCP_CHARDEV_STATE_CONNECTED;
 }
 
 static void char_socket_class_init(ObjectClass *oc, void *data)
-- 
2.20.1

[Qemu-devel] [PATCH v3 12/16] chardev: honour the reconnect setting in tcp_chr_wait_connected

[Daniel P. Berrangé]

If establishing a client connection fails, the tcp_chr_wait_connected
method should sleep for the reconnect timeout and then retry the
attempt. This ensures the callers don't immediately abort with an
error when the initial connection fails.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index d6de5d2305..7db20ff0a0 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -957,8 +957,15 @@ static int tcp_chr_wait_connected(Chardev *chr, Error **errp)
         if (s->is_listen) {
             tcp_chr_accept_server_sync(chr);
         } else {
-            if (tcp_chr_connect_client_sync(chr, errp) < 0) {
-                return -1;
+            Error *err = NULL;
+            if (tcp_chr_connect_client_sync(chr, &err) < 0) {
+                if (s->reconnect_time) {
+                    error_free(err);
+                    g_usleep(s->reconnect_time * 1000ULL * 1000ULL);
+                } else {
+                    error_propagate(errp, err);
+                    return -1;
+                }
             }
         }
     }
-- 
2.20.1

[Qemu-devel] [PATCH v3 13/16] chardev: disallow TLS/telnet/websocket with tcp_chr_wait_connected

[Daniel P. Berrangé]

In the previous commit

    commit 1dc8a6695c731abb7461c637b2512c3670d82be4
    Author: Marc-André Lureau <marcandre.lureau@redhat.com>
    Date:   Tue Aug 16 12:33:32 2016 +0400

      char: fix waiting for TLS and telnet connection

the tcp_chr_wait_connected() method was changed to check for a non-NULL
's->ioc' as a sign that there is already a connection present, as
opposed to checking the "connected" flag to supposedly fix handling of
TLS/telnet connections.

The original code would repeatedly call tcp_chr_wait_connected creating
many connections as 'connected' would never become true. The changed
code would still repeatedly call tcp_chr_wait_connected busy waiting
because s->ioc is set but the chardev will never see CHR_EVENT_OPENED.
IOW, the code is still broken with TLS/telnet, but in a different way.

Checking for a non-NULL 's->ioc' does not mean that a CHR_EVENT_OPENED
will be ready for a TLS/telnet connection. These protocols (and the
websocket protocol) all require the main loop to be running in order
to complete the protocol handshake before emitting CHR_EVENT_OPENED.
The tcp_chr_wait_connected() method is only used during early startup
before a main loop is running, so TLS/telnet/websock connections can
never complete initialization.

Making this work would require changing tcp_chr_wait_connected to run
a main loop. This is quite complex since we must not allow GSource's
that other parts of QEMU have registered to run yet. The current callers
of tcp_chr_wait_connected do not require use of the TLS/telnet/websocket
protocols, so the simplest option is to just forbid this combination
completely for now.

Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 7db20ff0a0..86c1f502d6 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -951,8 +951,20 @@ static void tcp_chr_accept_server_sync(Chardev *chr)
 static int tcp_chr_wait_connected(Chardev *chr, Error **errp)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
-    /* It can't wait on s->connected, since it is set asynchronously
-     * in TLS and telnet cases, only wait for an accepted socket */
+    const char *opts[] = { "telnet", "tn3270", "websock", "tls-creds" };
+    bool optset[] = { s->is_telnet, s->is_tn3270, s->is_websock, s->tls_creds };
+    size_t i;
+
+    QEMU_BUILD_BUG_ON(G_N_ELEMENTS(opts) != G_N_ELEMENTS(optset));
+    for (i = 0; i < G_N_ELEMENTS(opts); i++) {
+        if (optset[i]) {
+            error_setg(errp,
+                       "'%s' option is incompatible with waiting for "
+                       "connection completion", opts[i]);
+            return -1;
+        }
+    }
+
     while (!s->ioc) {
         if (s->is_listen) {
             tcp_chr_accept_server_sync(chr);
-- 
2.20.1

[Qemu-devel] [PATCH v3 14/16] chardev: fix race with client connections in tcp_chr_wait_connected

[Daniel P. Berrangé]

When the 'reconnect' option is given for a client connection, the
qmp_chardev_open_socket_client method will run an asynchronous
connection attempt. The QIOChannel socket executes this is a single use
background thread, so the connection will succeed immediately (assuming
the server is listening). The chardev, however, won't get the result
from this background thread until the main loop starts running and
processes idle callbacks.

Thus when tcp_chr_wait_connected is run s->ioc will be NULL, but the
state will be TCP_CHARDEV_STATE_CONNECTING, and there may already
be an established connection that will be associated with the chardev
by the pending idle callback. tcp_chr_wait_connected doesn't check the
state, only s->ioc, so attempts to establish another connection
synchronously.

If the server allows multiple connections this is unhelpful but not a
fatal problem as the duplicate connection will get ignored by the
tcp_chr_new_client method when it sees the state is already connected.

If the server only supports a single connection, however, the
tcp_chr_wait_connected method will hang forever because the server will
not accept its synchronous connection attempt until the first connection
is closed.

To deal with this tcp_chr_wait_connected needs to synchronize with the
completion of the background connection task. To do this it needs to
create the QIOTask directly and use the qio_task_wait_thread method.
This will cancel the pending idle callback and directly dispatch the
task completion callback, allowing the connection to be associated
with the chardev. If the background connection failed, it can still
attempt a new synchronous connection.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-socket.c | 90 +++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 86 insertions(+), 4 deletions(-)

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 86c1f502d6..4fcdd8aedd 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -80,6 +80,8 @@ typedef struct {
     GSource *reconnect_timer;
     int64_t reconnect_time;
     bool connect_err_reported;
+
+    QIOTask *connect_task;
 } SocketChardev;
 
 #define SOCKET_CHARDEV(obj)                                     \
@@ -118,6 +120,7 @@ static void qemu_chr_socket_restart_timer(Chardev *chr)
     char *name;
 
     assert(s->state == TCP_CHARDEV_STATE_DISCONNECTED);
+    assert(!s->reconnect_timer);
     name = g_strdup_printf("chardev-socket-reconnect-%s", chr->label);
     s->reconnect_timer = qemu_chr_timeout_add_ms(chr,
                                                  s->reconnect_time * 1000,
@@ -965,7 +968,56 @@ static int tcp_chr_wait_connected(Chardev *chr, Error **errp)
         }
     }
 
-    while (!s->ioc) {
+    tcp_chr_reconn_timer_cancel(s);
+
+    /*
+     * We expect states to be as follows:
+     *
+     *  - server
+     *    - wait   -> CONNECTED
+     *    - nowait -> DISCONNECTED
+     *  - client
+     *    - reconnect == 0 -> CONNECTED
+     *    - reconnect != 0 -> CONNECTING
+     *
+     */
+    if (s->state == TCP_CHARDEV_STATE_CONNECTING) {
+        if (!s->connect_task) {
+            error_setg(errp,
+                       "Unexpected 'connecting' state without connect task "
+                       "while waiting for connection completion");
+            return -1;
+        }
+        /*
+         * tcp_chr_wait_connected should only ever be run from the
+         * main loop thread associated with chr->gcontext, otherwise
+         * qio_task_wait_thread has a dangerous race condition with
+         * free'ing of the s->connect_task object.
+         *
+         * Acquiring the main context doesn't 100% prove we're in
+         * the main loop thread, but it does at least guarantee
+         * that the main loop won't be executed by another thread
+         * avoiding the race condition with the task idle callback.
+         */
+        g_main_context_acquire(chr->gcontext);
+        qio_task_wait_thread(s->connect_task);
+        g_main_context_release(chr->gcontext);
+
+        /*
+         * The completion callback (qemu_chr_socket_connected) for
+         * s->connect_task should have set this to NULL by the time
+         * qio_task_wait_thread has returned.
+         */
+        assert(!s->connect_task);
+
+        /*
+         * NB we are *not* guaranteed to have "s->state == ..CONNECTED"
+         * at this point as this first connect may be failed, so
+         * allow the next loop to run regardless.
+         */
+    }
+
+    while (s->state != TCP_CHARDEV_STATE_CONNECTED) {
         if (s->is_listen) {
             tcp_chr_accept_server_sync(chr);
         } else {
@@ -1014,6 +1066,8 @@ static void qemu_chr_socket_connected(QIOTask *task, void *opaque)
     SocketChardev *s = SOCKET_CHARDEV(chr);
     Error *err = NULL;
 
+    s->connect_task = NULL;
+
     if (qio_task_propagate_error(task, &err)) {
         tcp_chr_change_state(s, TCP_CHARDEV_STATE_DISCONNECTED);
         check_report_connect_error(chr, err);
@@ -1028,6 +1082,20 @@ cleanup:
     object_unref(OBJECT(sioc));
 }
 
+
+static void tcp_chr_connect_client_task(QIOTask *task,
+                                        gpointer opaque)
+{
+    QIOChannelSocket *ioc = QIO_CHANNEL_SOCKET(qio_task_get_source(task));
+    SocketAddress *addr = opaque;
+    Error *err = NULL;
+
+    qio_channel_socket_connect_sync(ioc, addr, &err);
+
+    qio_task_set_error(task, err);
+}
+
+
 static void tcp_chr_connect_client_async(Chardev *chr)
 {
     SocketChardev *s = SOCKET_CHARDEV(chr);
@@ -1036,9 +1104,23 @@ static void tcp_chr_connect_client_async(Chardev *chr)
     tcp_chr_change_state(s, TCP_CHARDEV_STATE_CONNECTING);
     sioc = qio_channel_socket_new();
     tcp_chr_set_client_ioc_name(chr, sioc);
-    qio_channel_socket_connect_async(sioc, s->addr,
-                                     qemu_chr_socket_connected,
-                                     chr, NULL, chr->gcontext);
+    /*
+     * Normally code would use the qio_channel_socket_connect_async
+     * method which uses a QIOTask + qio_task_set_error internally
+     * to avoid blocking. The tcp_chr_wait_connected method, however,
+     * needs a way to synchronize with completion of the background
+     * connect task which can't be done with the QIOChannelSocket
+     * async APIs. Thus we must use QIOTask directly to implement
+     * the non-blocking concept locally.
+     */
+    s->connect_task = qio_task_new(OBJECT(sioc),
+                                   qemu_chr_socket_connected,
+                                   chr, NULL);
+    qio_task_run_in_thread(s->connect_task,
+                           tcp_chr_connect_client_task,
+                           s->addr,
+                           NULL,
+                           chr->gcontext);
 }
 
 static gboolean socket_reconnect_timeout(gpointer opaque)
-- 
2.20.1

[Qemu-devel] [PATCH v3 15/16] tests: expand coverage of socket chardev test

[Daniel P. Berrangé]

The current socket chardev tests try to exercise the chardev socket
driver in both server and client mode at the same time. The chardev API
is not very well designed to handle both ends of the connection being in
the same process so this approach makes the test case quite unpleasant
to deal with.

This splits the tests into distinct cases, one to test server socket
chardevs and one to test client socket chardevs. In each case the peer
is run in a background thread using the simpler QIOChannelSocket APIs.
The main test case code can now be written in a way that mirrors the
typical usage from within QEMU.

In doing this recfactoring it is possible to greatly expand the test
coverage for the socket chardevs to test all combinations except for a
server operating in blocking wait mode.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 tests/test-char.c | 640 ++++++++++++++++++++++++++++++++++------------
 1 file changed, 475 insertions(+), 165 deletions(-)

diff --git a/tests/test-char.c b/tests/test-char.c
index 89c43e4ada..8b478bb67d 100644
--- a/tests/test-char.c
+++ b/tests/test-char.c
@@ -11,6 +11,9 @@
 #include "qapi/qapi-commands-char.h"
 #include "qapi/qmp/qdict.h"
 #include "qom/qom-qobject.h"
+#include "io/channel-socket.h"
+#include "qapi/qobject-input-visitor.h"
+#include "qapi/qapi-visit-sockets.h"
 
 static bool quit;
 
@@ -257,168 +260,6 @@ static void char_mux_test(void)
     qemu_chr_fe_deinit(&chr_be2, true);
 }
 
-typedef struct SocketIdleData {
-    GMainLoop *loop;
-    Chardev *chr;
-    bool conn_expected;
-    CharBackend *be;
-    CharBackend *client_be;
-} SocketIdleData;
-
-static gboolean char_socket_test_idle(gpointer user_data)
-{
-    SocketIdleData *data = user_data;
-
-    if (object_property_get_bool(OBJECT(data->chr), "connected", NULL)
-        == data->conn_expected) {
-        quit = true;
-        return FALSE;
-    }
-
-    return TRUE;
-}
-
-static void socket_read(void *opaque, const uint8_t *buf, int size)
-{
-    SocketIdleData *data = opaque;
-
-    g_assert_cmpint(size, ==, 1);
-    g_assert_cmpint(*buf, ==, 'Z');
-
-    size = qemu_chr_fe_write(data->be, (const uint8_t *)"hello", 5);
-    g_assert_cmpint(size, ==, 5);
-}
-
-static int socket_can_read(void *opaque)
-{
-    return 10;
-}
-
-static void socket_read_hello(void *opaque, const uint8_t *buf, int size)
-{
-    g_assert_cmpint(size, ==, 5);
-    g_assert(strncmp((char *)buf, "hello", 5) == 0);
-
-    quit = true;
-}
-
-static int socket_can_read_hello(void *opaque)
-{
-    return 10;
-}
-
-static void char_socket_test_common(Chardev *chr, bool reconnect)
-{
-    Chardev *chr_client;
-    QObject *addr;
-    QDict *qdict;
-    const char *port;
-    SocketIdleData d = { .chr = chr };
-    CharBackend be;
-    CharBackend client_be;
-    char *tmp;
-
-    d.be = &be;
-    d.client_be = &be;
-
-    g_assert_nonnull(chr);
-    g_assert(!object_property_get_bool(OBJECT(chr), "connected", &error_abort));
-
-    addr = object_property_get_qobject(OBJECT(chr), "addr", &error_abort);
-    qdict = qobject_to(QDict, addr);
-    port = qdict_get_str(qdict, "port");
-    tmp = g_strdup_printf("tcp:127.0.0.1:%s%s", port,
-                          reconnect ? ",reconnect=1" : "");
-    qobject_unref(qdict);
-
-    qemu_chr_fe_init(&be, chr, &error_abort);
-    qemu_chr_fe_set_handlers(&be, socket_can_read, socket_read,
-                             NULL, NULL, &d, NULL, true);
-
-    chr_client = qemu_chr_new("client", tmp);
-    qemu_chr_fe_init(&client_be, chr_client, &error_abort);
-    qemu_chr_fe_set_handlers(&client_be, socket_can_read_hello,
-                             socket_read_hello,
-                             NULL, NULL, &d, NULL, true);
-    g_free(tmp);
-
-    d.conn_expected = true;
-    guint id = g_idle_add(char_socket_test_idle, &d);
-    g_source_set_name_by_id(id, "test-idle");
-    g_assert_cmpint(id, >, 0);
-    main_loop();
-
-    d.chr = chr_client;
-    id = g_idle_add(char_socket_test_idle, &d);
-    g_source_set_name_by_id(id, "test-idle");
-    g_assert_cmpint(id, >, 0);
-    main_loop();
-
-    g_assert(object_property_get_bool(OBJECT(chr), "connected", &error_abort));
-    g_assert(object_property_get_bool(OBJECT(chr_client),
-                                      "connected", &error_abort));
-
-    qemu_chr_write_all(chr_client, (const uint8_t *)"Z", 1);
-    main_loop();
-
-    object_unparent(OBJECT(chr_client));
-
-    d.chr = chr;
-    d.conn_expected = false;
-    g_idle_add(char_socket_test_idle, &d);
-    main_loop();
-
-    object_unparent(OBJECT(chr));
-}
-
-
-static void char_socket_basic_test(void)
-{
-    Chardev *chr = qemu_chr_new("server", "tcp:127.0.0.1:0,server,nowait");
-
-    char_socket_test_common(chr, false);
-}
-
-
-static void char_socket_reconnect_test(void)
-{
-    Chardev *chr = qemu_chr_new("server", "tcp:127.0.0.1:0,server,nowait");
-
-    char_socket_test_common(chr, true);
-}
-
-
-static void char_socket_fdpass_test(void)
-{
-    Chardev *chr;
-    char *optstr;
-    QemuOpts *opts;
-    int fd;
-    SocketAddress *addr = g_new0(SocketAddress, 1);
-
-    addr->type = SOCKET_ADDRESS_TYPE_INET;
-    addr->u.inet.host = g_strdup("127.0.0.1");
-    addr->u.inet.port = g_strdup("0");
-
-    fd = socket_listen(addr, &error_abort);
-    g_assert(fd >= 0);
-
-    qapi_free_SocketAddress(addr);
-
-    optstr = g_strdup_printf("socket,id=cdev,fd=%d,server,nowait", fd);
-
-    opts = qemu_opts_parse_noisily(qemu_find_opts("chardev"),
-                                   optstr, true);
-    g_free(optstr);
-    g_assert_nonnull(opts);
-
-    chr = qemu_chr_new_from_opts(opts, &error_abort);
-
-    qemu_opts_del(opts);
-
-    char_socket_test_common(chr, false);
-}
-
 
 static void websock_server_read(void *opaque, const uint8_t *buf, int size)
 {
@@ -610,6 +451,28 @@ static void char_pipe_test(void)
 }
 #endif
 
+typedef struct SocketIdleData {
+    GMainLoop *loop;
+    Chardev *chr;
+    bool conn_expected;
+    CharBackend *be;
+    CharBackend *client_be;
+} SocketIdleData;
+
+
+static void socket_read_hello(void *opaque, const uint8_t *buf, int size)
+{
+    g_assert_cmpint(size, ==, 5);
+    g_assert(strncmp((char *)buf, "hello", 5) == 0);
+
+    quit = true;
+}
+
+static int socket_can_read_hello(void *opaque)
+{
+    return 10;
+}
+
 static int make_udp_socket(int *port)
 {
     struct sockaddr_in addr = { 0, };
@@ -680,6 +543,391 @@ static void char_udp_test(void)
     char_udp_test_internal(NULL, 0);
 }
 
+
+typedef struct {
+    int event;
+    bool got_pong;
+} CharSocketTestData;
+
+
+#define SOCKET_PING "Hello"
+#define SOCKET_PONG "World"
+
+
+static void
+char_socket_event(void *opaque, int event)
+{
+    CharSocketTestData *data = opaque;
+    data->event = event;
+}
+
+
+static void
+char_socket_read(void *opaque, const uint8_t *buf, int size)
+{
+    CharSocketTestData *data = opaque;
+    g_assert_cmpint(size, ==, sizeof(SOCKET_PONG));
+    g_assert(memcmp(buf, SOCKET_PONG, size) == 0);
+    data->got_pong = true;
+}
+
+
+static int
+char_socket_can_read(void *opaque)
+{
+    return sizeof(SOCKET_PONG);
+}
+
+
+static char *
+char_socket_addr_to_opt_str(SocketAddress *addr, bool fd_pass,
+                            const char *reconnect, bool is_listen)
+{
+    if (fd_pass) {
+        QIOChannelSocket *ioc = qio_channel_socket_new();
+        int fd;
+        char *optstr;
+        g_assert(!reconnect);
+        if (is_listen) {
+            qio_channel_socket_listen_sync(ioc, addr, &error_abort);
+        } else {
+            qio_channel_socket_connect_sync(ioc, addr, &error_abort);
+        }
+        fd = ioc->fd;
+        ioc->fd = -1;
+        optstr = g_strdup_printf("socket,id=cdev0,fd=%d%s",
+                                 fd, is_listen ? ",server,nowait" : "");
+        object_unref(OBJECT(ioc));
+        return optstr;
+    } else {
+        switch (addr->type) {
+        case SOCKET_ADDRESS_TYPE_INET:
+            return g_strdup_printf("socket,id=cdev0,host=%s,port=%s%s%s",
+                                   addr->u.inet.host,
+                                   addr->u.inet.port,
+                                   reconnect ? reconnect : "",
+                                   is_listen ? ",server,nowait" : "");
+
+        case SOCKET_ADDRESS_TYPE_UNIX:
+            return g_strdup_printf("socket,id=cdev0,path=%s%s%s",
+                                   addr->u.q_unix.path,
+                                   reconnect ? reconnect : "",
+                                   is_listen ? ",server,nowait" : "");
+
+        default:
+            g_assert_not_reached();
+        }
+    }
+}
+
+
+static void
+char_socket_ping_pong(QIOChannel *ioc)
+{
+    char greeting[sizeof(SOCKET_PING)];
+    const char *response = SOCKET_PONG;
+
+    qio_channel_read_all(ioc, greeting, sizeof(greeting), &error_abort);
+
+    g_assert(memcmp(greeting, SOCKET_PING, sizeof(greeting)) == 0);
+
+    qio_channel_write_all(ioc, response, sizeof(SOCKET_PONG), &error_abort);
+
+    object_unref(OBJECT(ioc));
+}
+
+
+static gpointer
+char_socket_server_client_thread(gpointer data)
+{
+    SocketAddress *addr = data;
+    QIOChannelSocket *ioc = qio_channel_socket_new();
+
+    qio_channel_socket_connect_sync(ioc, addr, &error_abort);
+
+    char_socket_ping_pong(QIO_CHANNEL(ioc));
+
+    return NULL;
+}
+
+
+typedef struct {
+    SocketAddress *addr;
+    bool wait_connected;
+    bool fd_pass;
+} CharSocketServerTestConfig;
+
+
+static void char_socket_server_test(gconstpointer opaque)
+{
+    const CharSocketServerTestConfig *config = opaque;
+    Chardev *chr;
+    CharBackend be = {0};
+    CharSocketTestData data = {0};
+    QObject *qaddr;
+    SocketAddress *addr;
+    Visitor *v;
+    QemuThread thread;
+    int ret;
+    bool reconnected;
+    char *optstr;
+    QemuOpts *opts;
+
+    g_setenv("QTEST_SILENT_ERRORS", "1", 1);
+    /*
+     * We rely on config->addr containing "nowait", otherwise
+     * qemu_chr_new() will block until a client connects. We
+     * can't spawn our client thread though, because until
+     * qemu_chr_new() returns we don't know what TCP port was
+     * allocated by the OS
+     */
+    optstr = char_socket_addr_to_opt_str(config->addr,
+                                         config->fd_pass,
+                                         NULL,
+                                         true);
+    opts = qemu_opts_parse_noisily(qemu_find_opts("chardev"),
+                                   optstr, true);
+    g_assert_nonnull(opts);
+    chr = qemu_chr_new_from_opts(opts, &error_abort);
+    qemu_opts_del(opts);
+    g_assert_nonnull(chr);
+    g_assert(!object_property_get_bool(OBJECT(chr), "connected", &error_abort));
+
+    qaddr = object_property_get_qobject(OBJECT(chr), "addr", &error_abort);
+    g_assert_nonnull(qaddr);
+
+    v = qobject_input_visitor_new(qaddr);
+    visit_type_SocketAddress(v, "addr", &addr, &error_abort);
+    visit_free(v);
+    qobject_unref(qaddr);
+
+    qemu_chr_fe_init(&be, chr, &error_abort);
+
+ reconnect:
+    data.event = -1;
+    qemu_chr_fe_set_handlers(&be, NULL, NULL,
+                             char_socket_event, NULL,
+                             &data, NULL, true);
+    g_assert(data.event == -1);
+
+    /*
+     * Kick off a thread to act as the "remote" client
+     * which just plays ping-pong with us
+     */
+    qemu_thread_create(&thread, "client",
+                       char_socket_server_client_thread,
+                       addr, QEMU_THREAD_JOINABLE);
+    g_assert(data.event == -1);
+
+    if (config->wait_connected) {
+        /* Synchronously accept a connection */
+        qemu_chr_wait_connected(chr, &error_abort);
+    } else {
+        /*
+         * Asynchronously accept a connection when the evnt
+         * loop reports the listener socket as readable
+         */
+        while (data.event == -1) {
+            main_loop_wait(false);
+        }
+    }
+    g_assert(object_property_get_bool(OBJECT(chr), "connected", &error_abort));
+    g_assert(data.event == CHR_EVENT_OPENED);
+    data.event = -1;
+
+    /* Send a greeting to the client */
+    ret = qemu_chr_fe_write_all(&be, (const uint8_t *)SOCKET_PING,
+                                sizeof(SOCKET_PING));
+    g_assert_cmpint(ret, ==, sizeof(SOCKET_PING));
+    g_assert(data.event == -1);
+
+    /* Setup a callback to receive the reply to our greeting */
+    qemu_chr_fe_set_handlers(&be, char_socket_can_read,
+                             char_socket_read,
+                             char_socket_event, NULL,
+                             &data, NULL, true);
+    g_assert(data.event == CHR_EVENT_OPENED);
+    data.event = -1;
+
+    /* Wait for the client to go away */
+    while (data.event == -1) {
+        main_loop_wait(false);
+    }
+    g_assert(!object_property_get_bool(OBJECT(chr), "connected", &error_abort));
+    g_assert(data.event == CHR_EVENT_CLOSED);
+    g_assert(data.got_pong);
+
+    qemu_thread_join(&thread);
+
+    if (!reconnected) {
+        reconnected = true;
+        goto reconnect;
+    }
+
+    qapi_free_SocketAddress(addr);
+    object_unparent(OBJECT(chr));
+    g_free(optstr);
+    g_unsetenv("QTEST_SILENT_ERRORS");
+}
+
+
+static gpointer
+char_socket_client_server_thread(gpointer data)
+{
+    QIOChannelSocket *ioc = data;
+    QIOChannelSocket *cioc;
+
+    cioc = qio_channel_socket_accept(ioc, &error_abort);
+    g_assert_nonnull(cioc);
+
+    char_socket_ping_pong(QIO_CHANNEL(cioc));
+
+    return NULL;
+}
+
+
+typedef struct {
+    SocketAddress *addr;
+    const char *reconnect;
+    bool wait_connected;
+    bool fd_pass;
+} CharSocketClientTestConfig;
+
+
+static void char_socket_client_test(gconstpointer opaque)
+{
+    const CharSocketClientTestConfig *config = opaque;
+    QIOChannelSocket *ioc;
+    char *optstr;
+    Chardev *chr;
+    CharBackend be = {0};
+    CharSocketTestData data = {0};
+    SocketAddress *addr;
+    QemuThread thread;
+    int ret;
+    bool reconnected = false;
+    QemuOpts *opts;
+
+    /*
+     * Setup a listener socket and determine get its address
+     * so we know the TCP port for the client later
+     */
+    ioc = qio_channel_socket_new();
+    g_assert_nonnull(ioc);
+    qio_channel_socket_listen_sync(ioc, config->addr, &error_abort);
+    addr = qio_channel_socket_get_local_address(ioc, &error_abort);
+    g_assert_nonnull(addr);
+
+    /*
+     * Kick off a thread to act as the "remote" client
+     * which just plays ping-pong with us
+     */
+    qemu_thread_create(&thread, "client",
+                       char_socket_client_server_thread,
+                       ioc, QEMU_THREAD_JOINABLE);
+
+    /*
+     * Populate the chardev address based on what the server
+     * is actually listening on
+     */
+    optstr = char_socket_addr_to_opt_str(addr,
+                                         config->fd_pass,
+                                         config->reconnect,
+                                         false);
+
+    opts = qemu_opts_parse_noisily(qemu_find_opts("chardev"),
+                                   optstr, true);
+    g_assert_nonnull(opts);
+    chr = qemu_chr_new_from_opts(opts, &error_abort);
+    qemu_opts_del(opts);
+    g_assert_nonnull(chr);
+
+    if (config->reconnect) {
+        /*
+         * If reconnect is set, the connection will be
+         * established in a background thread and we won't
+         * see the "connected" status updated until we
+         * run the main event loop, or call qemu_chr_wait_connected
+         */
+        g_assert(!object_property_get_bool(OBJECT(chr), "connected",
+                                           &error_abort));
+    } else {
+        g_assert(object_property_get_bool(OBJECT(chr), "connected",
+                                          &error_abort));
+    }
+
+    qemu_chr_fe_init(&be, chr, &error_abort);
+
+ reconnect:
+    data.event = -1;
+    qemu_chr_fe_set_handlers(&be, NULL, NULL,
+                             char_socket_event, NULL,
+                             &data, NULL, true);
+    if (config->reconnect) {
+        g_assert(data.event == -1);
+    } else {
+        g_assert(data.event == CHR_EVENT_OPENED);
+    }
+
+    if (config->wait_connected) {
+        /*
+         * Synchronously wait for the connection to complete
+         * This should be a no-op if reconnect is not set.
+         */
+        qemu_chr_wait_connected(chr, &error_abort);
+    } else {
+        /*
+         * Asynchronously wait for the connection to be reported
+         * as complete when the background thread reports its
+         * status.
+         * The loop will short-circuit if reconnect was set
+         */
+        while (data.event == -1) {
+            main_loop_wait(false);
+        }
+    }
+    g_assert(data.event == CHR_EVENT_OPENED);
+    data.event = -1;
+    g_assert(object_property_get_bool(OBJECT(chr), "connected", &error_abort));
+
+    /* Send a greeting to the server */
+    ret = qemu_chr_fe_write_all(&be, (const uint8_t *)SOCKET_PING,
+                                sizeof(SOCKET_PING));
+    g_assert_cmpint(ret, ==, sizeof(SOCKET_PING));
+    g_assert(data.event == -1);
+
+    /* Setup a callback to receive the reply to our greeting */
+    qemu_chr_fe_set_handlers(&be, char_socket_can_read,
+                             char_socket_read,
+                             char_socket_event, NULL,
+                             &data, NULL, true);
+    g_assert(data.event == CHR_EVENT_OPENED);
+    data.event = -1;
+
+    /* Wait for the server to go away */
+    while (data.event == -1) {
+        main_loop_wait(false);
+    }
+    g_assert(data.event == CHR_EVENT_CLOSED);
+    g_assert(!object_property_get_bool(OBJECT(chr), "connected", &error_abort));
+    g_assert(data.got_pong);
+    qemu_thread_join(&thread);
+
+    if (config->reconnect && !reconnected) {
+        reconnected = true;
+        qemu_thread_create(&thread, "client",
+                           char_socket_client_server_thread,
+                           ioc, QEMU_THREAD_JOINABLE);
+        goto reconnect;
+    }
+
+    object_unref(OBJECT(ioc));
+    object_unparent(OBJECT(chr));
+    qapi_free_SocketAddress(addr);
+    g_free(optstr);
+}
+
+
 #ifdef HAVE_CHARDEV_SERIAL
 static void char_serial_test(void)
 {
@@ -959,9 +1207,71 @@ int main(int argc, char **argv)
 #ifndef _WIN32
     g_test_add_func("/char/file-fifo", char_file_fifo_test);
 #endif
-    g_test_add_func("/char/socket/basic", char_socket_basic_test);
-    g_test_add_func("/char/socket/reconnect", char_socket_reconnect_test);
-    g_test_add_func("/char/socket/fdpass", char_socket_fdpass_test);
+
+    SocketAddress tcpaddr = {
+        .type = SOCKET_ADDRESS_TYPE_INET,
+        .u.inet.host = (char *)"127.0.0.1",
+        .u.inet.port = (char *)"0",
+    };
+#ifndef WIN32
+    SocketAddress unixaddr = {
+        .type = SOCKET_ADDRESS_TYPE_UNIX,
+        .u.q_unix.path = (char *)"test-char.sock",
+    };
+#endif
+
+#define SOCKET_SERVER_TEST(name, addr)                                  \
+    CharSocketServerTestConfig server1 ## name =                        \
+        { addr, false, false };                                         \
+    CharSocketServerTestConfig server2 ## name =                        \
+        { addr, true, false };                                          \
+    CharSocketServerTestConfig server3 ## name =                        \
+        { addr, false, true };                                          \
+    CharSocketServerTestConfig server4 ## name =                        \
+        { addr, true, true };                                           \
+    g_test_add_data_func("/char/socket/server/mainloop/" # name,        \
+                         &server1 ##name, char_socket_server_test);     \
+    g_test_add_data_func("/char/socket/server/wait-conn/" # name,       \
+                         &server2 ##name, char_socket_server_test);     \
+    g_test_add_data_func("/char/socket/server/mainloop-fdpass/" # name, \
+                         &server3 ##name, char_socket_server_test);     \
+    g_test_add_data_func("/char/socket/server/wait-conn-fdpass/" # name, \
+                         &server4 ##name, char_socket_server_test)
+
+#define SOCKET_CLIENT_TEST(name, addr)                                  \
+    CharSocketClientTestConfig client1 ## name =                        \
+        { addr, NULL, false, false };                                   \
+    CharSocketClientTestConfig client2 ## name =                        \
+        { addr, NULL, true, false };                                    \
+    CharSocketClientTestConfig client3 ## name =                        \
+        { addr, ",reconnect=1", false };                                \
+    CharSocketClientTestConfig client4 ## name =                        \
+        { addr, ",reconnect=1", true };                                 \
+    CharSocketClientTestConfig client5 ## name =                        \
+        { addr, NULL, false, true };                                    \
+    CharSocketClientTestConfig client6 ## name =                        \
+        { addr, NULL, true, true };                                     \
+    g_test_add_data_func("/char/socket/client/mainloop/" # name,        \
+                         &client1 ##name, char_socket_client_test);     \
+    g_test_add_data_func("/char/socket/client/wait-conn/" # name,       \
+                         &client2 ##name, char_socket_client_test);     \
+    g_test_add_data_func("/char/socket/client/mainloop-reconnect/" # name, \
+                         &client3 ##name, char_socket_client_test);     \
+    g_test_add_data_func("/char/socket/client/wait-conn-reconnect/" # name, \
+                         &client4 ##name, char_socket_client_test);     \
+    g_test_add_data_func("/char/socket/client/mainloop-fdpass/" # name, \
+                         &client5 ##name, char_socket_client_test);     \
+    g_test_add_data_func("/char/socket/client/wait-conn-fdpass/" # name, \
+                         &client6 ##name, char_socket_client_test)
+
+    SOCKET_SERVER_TEST(tcp, &tcpaddr);
+    SOCKET_CLIENT_TEST(tcp, &tcpaddr);
+#ifndef WIN32
+    SOCKET_SERVER_TEST(unix, &unixaddr);
+    SOCKET_CLIENT_TEST(unix, &unixaddr);
+#endif
+
+
     g_test_add_func("/char/udp", char_udp_test);
 #ifdef HAVE_CHARDEV_SERIAL
     g_test_add_func("/char/serial", char_serial_test);
-- 
2.20.1

[Qemu-devel] [PATCH v3 16/16] chardev: ensure termios is fully initialized

[Daniel P. Berrangé]

valgrind on the test-char.c code reports that 'struct termios' contains
uninitialized memory.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 chardev/char-serial.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/chardev/char-serial.c b/chardev/char-serial.c
index 3299b46853..a8bae31b8d 100644
--- a/chardev/char-serial.c
+++ b/chardev/char-serial.c
@@ -57,7 +57,7 @@ static void qmp_chardev_open_serial(Chardev *chr,
 static void tty_serial_init(int fd, int speed,
                             int parity, int data_bits, int stop_bits)
 {
-    struct termios tty;
+    struct termios tty = {0};
     speed_t spd;
 
 #if 0
-- 
2.20.1

Re: [Qemu-devel] [PATCH v3 00/16] chardev: refactoring & many bugfixes related tcp_chr_wait_connected

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 2938 bytes --]

On 2/11/19 12:24 PM, Daniel P. Berrangé wrote:
> This is a followup to
> 
>   v1: https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg03344.html
>   v2: http://lists.nongnu.org/archive/html/qemu-devel/2019-01/msg05947.html
> 
> This series comes out of a discussion between myself & Yongji Xie in:
> 
>   https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg01881.html
> 
> I eventually understood that the problem faced was that
> tcp_chr_wait_connected was racing with the background connection attempt
> previously started, causing two connections to be established. This
> broke because some vhost user servers only allow a single connection.
> 
> After messing around with the code alot the final solution was in fact
> very easy. We simply have to delay the first background connection
> attempt until the main loop is running. It will then automatically
> turn into a no-op if tcp_chr_wait_connected has been run. This is
> dealt with in the last patch in this series
> 
> I believe this should solve the problem Yongji Xie faced, and thus not
> require us to add support for "nowait" option with client sockets at
> all. The reconnect=1 option effectively already implements nowait
> semantics, and now plays nicely with tcp_chr_wait_connected.
> 
> In investigating this I found various other bugs that needed fixing and
> identified some useful refactoring to simplify / clarify the code, hence
> this very long series.

Even with this series applied, I'm still seeing sporadic failures of
iotest 169. Max posted a hack patch a while back that tries to work
around the race:

https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg05907.html

which he originally diagnosed in iotest 147:
https://lists.nongnu.org/archive/html/qemu-devel/2018-12/msg05579.html

but as it was a hack, he has not pursued it further, and so the symptoms
are still there, although not completely reproducible:

169 10s ... - output mismatch (see 169.out.bad)
--- /home/eblake/qemu/tests/qemu-iotests/169.out	2018-11-16
15:48:12.018526748 -0600
+++ /home/eblake/qemu/tests/qemu-iotests/169.out.bad	2019-04-22
09:38:45.481517132 -0500
@@ -1,3 +1,5 @@
+WARNING:qemu:qemu received signal 11:
/home/eblake/qemu/tests/qemu-iotests/../../x86_64-softmmu/qemu-system-x86_64
-chardev
socket,id=mon,path=/home/eblake/qemu/tests/qemu-iotests/scratch/tmp4clmPF/qemua-26803-monitor.sock
-mon chardev=mon,mode=control -display none -vga none -qtest
unix:path=/home/eblake/qemu/tests/qemu-iotests/scratch/qemua-26803-qtest.sock
-machine accel=qtest -nodefaults -machine accel=qtest -drive
if=virtio,id=drive0,file=/home/eblake/qemu/tests/qemu-iotests/scratch/disk_a,format=qcow2,cache=writeback

Any chance you can take a look as to what a non-hack fix should be?

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [Qemu-devel] [PATCH v3 00/16] chardev: refactoring & many bugfixes related tcp_chr_wait_connected

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 2938 bytes --]

On 2/11/19 12:24 PM, Daniel P. Berrangé wrote:
> This is a followup to
> 
>   v1: https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg03344.html
>   v2: http://lists.nongnu.org/archive/html/qemu-devel/2019-01/msg05947.html
> 
> This series comes out of a discussion between myself & Yongji Xie in:
> 
>   https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg01881.html
> 
> I eventually understood that the problem faced was that
> tcp_chr_wait_connected was racing with the background connection attempt
> previously started, causing two connections to be established. This
> broke because some vhost user servers only allow a single connection.
> 
> After messing around with the code alot the final solution was in fact
> very easy. We simply have to delay the first background connection
> attempt until the main loop is running. It will then automatically
> turn into a no-op if tcp_chr_wait_connected has been run. This is
> dealt with in the last patch in this series
> 
> I believe this should solve the problem Yongji Xie faced, and thus not
> require us to add support for "nowait" option with client sockets at
> all. The reconnect=1 option effectively already implements nowait
> semantics, and now plays nicely with tcp_chr_wait_connected.
> 
> In investigating this I found various other bugs that needed fixing and
> identified some useful refactoring to simplify / clarify the code, hence
> this very long series.

Even with this series applied, I'm still seeing sporadic failures of
iotest 169. Max posted a hack patch a while back that tries to work
around the race:

https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg05907.html

which he originally diagnosed in iotest 147:
https://lists.nongnu.org/archive/html/qemu-devel/2018-12/msg05579.html

but as it was a hack, he has not pursued it further, and so the symptoms
are still there, although not completely reproducible:

169 10s ... - output mismatch (see 169.out.bad)
--- /home/eblake/qemu/tests/qemu-iotests/169.out	2018-11-16
15:48:12.018526748 -0600
+++ /home/eblake/qemu/tests/qemu-iotests/169.out.bad	2019-04-22
09:38:45.481517132 -0500
@@ -1,3 +1,5 @@
+WARNING:qemu:qemu received signal 11:
/home/eblake/qemu/tests/qemu-iotests/../../x86_64-softmmu/qemu-system-x86_64
-chardev
socket,id=mon,path=/home/eblake/qemu/tests/qemu-iotests/scratch/tmp4clmPF/qemua-26803-monitor.sock
-mon chardev=mon,mode=control -display none -vga none -qtest
unix:path=/home/eblake/qemu/tests/qemu-iotests/scratch/qemua-26803-qtest.sock
-machine accel=qtest -nodefaults -machine accel=qtest -drive
if=virtio,id=drive0,file=/home/eblake/qemu/tests/qemu-iotests/scratch/disk_a,format=qcow2,cache=writeback

Any chance you can take a look as to what a non-hack fix should be?

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [Qemu-devel] [PATCH v3 00/16] chardev: refactoring & many bugfixes related tcp_chr_wait_connected

[Daniel P. Berrangé]

On Mon, Apr 22, 2019 at 09:51:17AM -0500, Eric Blake wrote:
> On 2/11/19 12:24 PM, Daniel P. Berrangé wrote:
> > This is a followup to
> > 
> >   v1: https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg03344.html
> >   v2: http://lists.nongnu.org/archive/html/qemu-devel/2019-01/msg05947.html
> > 
> > This series comes out of a discussion between myself & Yongji Xie in:
> > 
> >   https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg01881.html
> > 
> > I eventually understood that the problem faced was that
> > tcp_chr_wait_connected was racing with the background connection attempt
> > previously started, causing two connections to be established. This
> > broke because some vhost user servers only allow a single connection.
> > 
> > After messing around with the code alot the final solution was in fact
> > very easy. We simply have to delay the first background connection
> > attempt until the main loop is running. It will then automatically
> > turn into a no-op if tcp_chr_wait_connected has been run. This is
> > dealt with in the last patch in this series
> > 
> > I believe this should solve the problem Yongji Xie faced, and thus not
> > require us to add support for "nowait" option with client sockets at
> > all. The reconnect=1 option effectively already implements nowait
> > semantics, and now plays nicely with tcp_chr_wait_connected.
> > 
> > In investigating this I found various other bugs that needed fixing and
> > identified some useful refactoring to simplify / clarify the code, hence
> > this very long series.
> 
> Even with this series applied, I'm still seeing sporadic failures of
> iotest 169. Max posted a hack patch a while back that tries to work
> around the race:
> 
> https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg05907.html
> 
> which he originally diagnosed in iotest 147:
> https://lists.nongnu.org/archive/html/qemu-devel/2018-12/msg05579.html
> 
> but as it was a hack, he has not pursued it further, and so the symptoms
> are still there, although not completely reproducible:
> 
> 169 10s ... - output mismatch (see 169.out.bad)
> --- /home/eblake/qemu/tests/qemu-iotests/169.out	2018-11-16
> 15:48:12.018526748 -0600
> +++ /home/eblake/qemu/tests/qemu-iotests/169.out.bad	2019-04-22
> 09:38:45.481517132 -0500
> @@ -1,3 +1,5 @@
> +WARNING:qemu:qemu received signal 11:
> /home/eblake/qemu/tests/qemu-iotests/../../x86_64-softmmu/qemu-system-x86_64
> -chardev
> socket,id=mon,path=/home/eblake/qemu/tests/qemu-iotests/scratch/tmp4clmPF/qemua-26803-monitor.sock
> -mon chardev=mon,mode=control -display none -vga none -qtest
> unix:path=/home/eblake/qemu/tests/qemu-iotests/scratch/qemua-26803-qtest.sock
> -machine accel=qtest -nodefaults -machine accel=qtest -drive
> if=virtio,id=drive0,file=/home/eblake/qemu/tests/qemu-iotests/scratch/disk_a,format=qcow2,cache=writeback
> 
> Any chance you can take a look as to what a non-hack fix should be?

Oh, it looks like we dropped the ball here. We have a fix already but
it doesn't appear to have been merged for 4.0 :-(

  https://lists.gnu.org/archive/html/qemu-devel/2019-02/msg06174.html

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|

Re: [Qemu-devel] [PATCH v3 00/16] chardev: refactoring & many bugfixes related tcp_chr_wait_connected

[Daniel P. Berrangé]

On Mon, Apr 22, 2019 at 09:51:17AM -0500, Eric Blake wrote:
> On 2/11/19 12:24 PM, Daniel P. Berrangé wrote:
> > This is a followup to
> > 
> >   v1: https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg03344.html
> >   v2: http://lists.nongnu.org/archive/html/qemu-devel/2019-01/msg05947.html
> > 
> > This series comes out of a discussion between myself & Yongji Xie in:
> > 
> >   https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg01881.html
> > 
> > I eventually understood that the problem faced was that
> > tcp_chr_wait_connected was racing with the background connection attempt
> > previously started, causing two connections to be established. This
> > broke because some vhost user servers only allow a single connection.
> > 
> > After messing around with the code alot the final solution was in fact
> > very easy. We simply have to delay the first background connection
> > attempt until the main loop is running. It will then automatically
> > turn into a no-op if tcp_chr_wait_connected has been run. This is
> > dealt with in the last patch in this series
> > 
> > I believe this should solve the problem Yongji Xie faced, and thus not
> > require us to add support for "nowait" option with client sockets at
> > all. The reconnect=1 option effectively already implements nowait
> > semantics, and now plays nicely with tcp_chr_wait_connected.
> > 
> > In investigating this I found various other bugs that needed fixing and
> > identified some useful refactoring to simplify / clarify the code, hence
> > this very long series.
> 
> Even with this series applied, I'm still seeing sporadic failures of
> iotest 169. Max posted a hack patch a while back that tries to work
> around the race:
> 
> https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg05907.html
> 
> which he originally diagnosed in iotest 147:
> https://lists.nongnu.org/archive/html/qemu-devel/2018-12/msg05579.html
> 
> but as it was a hack, he has not pursued it further, and so the symptoms
> are still there, although not completely reproducible:
> 
> 169 10s ... - output mismatch (see 169.out.bad)
> --- /home/eblake/qemu/tests/qemu-iotests/169.out	2018-11-16
> 15:48:12.018526748 -0600
> +++ /home/eblake/qemu/tests/qemu-iotests/169.out.bad	2019-04-22
> 09:38:45.481517132 -0500
> @@ -1,3 +1,5 @@
> +WARNING:qemu:qemu received signal 11:
> /home/eblake/qemu/tests/qemu-iotests/../../x86_64-softmmu/qemu-system-x86_64
> -chardev
> socket,id=mon,path=/home/eblake/qemu/tests/qemu-iotests/scratch/tmp4clmPF/qemua-26803-monitor.sock
> -mon chardev=mon,mode=control -display none -vga none -qtest
> unix:path=/home/eblake/qemu/tests/qemu-iotests/scratch/qemua-26803-qtest.sock
> -machine accel=qtest -nodefaults -machine accel=qtest -drive
> if=virtio,id=drive0,file=/home/eblake/qemu/tests/qemu-iotests/scratch/disk_a,format=qcow2,cache=writeback
> 
> Any chance you can take a look as to what a non-hack fix should be?

Oh, it looks like we dropped the ball here. We have a fix already but
it doesn't appear to have been merged for 4.0 :-(

  https://lists.gnu.org/archive/html/qemu-devel/2019-02/msg06174.html

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|