char-socket: grow timeout from smaller value

Commit Message

Vladimir Sementsov-Ogievskiy Aug. 1, 2024, 8:44 a.m. UTC

The smallest possible timeout for reconnect to set in QAPI is 1 second.
But for system where reconnect is considered a regular situation and
backend is guaranteed to start in tens of milliseconds this timeout of
1 second becomes an extra throttling.

On the other hand, just allowing smaller timeout would be ineffective
when we consider unplanned disconnect for larger time.

Let's do a simple thing: without any change to API, start reconnecting
with some small constant timeout (5 ms), increasing it exponentially up
to user given value.

Signed-off-by: Vladimir Sementsov-Ogievskiy <[e-mail]>
---

Hi all! We faced too long (1 second) pauses in IO during reconnect to
vhost server, and this 1 second is exactly reconnect=1 parameter of
char-socket. Here is our solution for the problem.

 chardev/char-socket.c         | 21 ++++++++++++++++-----
 include/chardev/char-socket.h |  5 ++++-
 2 files changed, 20 insertions(+), 6 deletions(-)

Patch

diff --git a/chardev/char-socket.c b/chardev/char-socket.c
index 1ca9441b1b..4a35f75ebe 100644
--- a/chardev/char-socket.c
+++ b/chardev/char-socket.c
@@ -74,9 +74,10 @@  static void qemu_chr_socket_restart_timer(Chardev *chr)
     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,
+                                                 s->reconnect_ms,
                                                  socket_reconnect_timeout,
                                                  chr);
+    s->reconnect_ms = MIN(s->reconnect_ms * 2, s->reconnect_ms_max);
     g_source_set_name(s->reconnect_timer, name);
     g_free(name);
 }
@@ -481,7 +482,7 @@  static void tcp_chr_disconnect_locked(Chardev *chr)
     if (emit_close) {
         qemu_chr_be_event(chr, CHR_EVENT_CLOSED);
     }
-    if (s->reconnect_time && !s->reconnect_timer) {
+    if (s->reconnect_ms && !s->reconnect_timer) {
         qemu_chr_socket_restart_timer(chr);
     }
 }
@@ -1080,9 +1081,11 @@  static int tcp_chr_wait_connected(Chardev *chr, Error **errp)
         } else {
             Error *err = NULL;
             if (tcp_chr_connect_client_sync(chr, &err) < 0) {
-                if (s->reconnect_time) {
+                if (s->reconnect_ms) {
                     error_free(err);
-                    g_usleep(s->reconnect_time * 1000ULL * 1000ULL);
+                    g_usleep(s->reconnect_ms * 1000ULL);
+                    s->reconnect_ms = MIN(s->reconnect_ms * 2,
+                                          s->reconnect_ms_max);
                 } else {
                     error_propagate(errp, err);
                     return -1;
@@ -1091,6 +1094,10 @@  static int tcp_chr_wait_connected(Chardev *chr, Error **errp)
         }
     }
 
+    if (s->reconnect_ms) {
+        s->reconnect_ms = SOCKET_CHARDEV_RECONNECT_MS_MIN;
+    }
+
     return 0;
 }
 
@@ -1147,6 +1154,9 @@  static void qemu_chr_socket_connected(QIOTask *task, void *opaque)
         goto cleanup;
     }
 
+    if (s->reconnect_ms) {
+        s->reconnect_ms = SOCKET_CHARDEV_RECONNECT_MS_MIN;
+    }
     s->connect_err_reported = false;
     tcp_chr_new_client(chr, sioc);
 
@@ -1273,7 +1283,8 @@  static int qmp_chardev_open_socket_client(Chardev *chr,
     SocketChardev *s = SOCKET_CHARDEV(chr);
 
     if (reconnect > 0) {
-        s->reconnect_time = reconnect;
+        s->reconnect_ms_max = reconnect * 1000;
+        s->reconnect_ms = SOCKET_CHARDEV_RECONNECT_MS_MIN;
         tcp_chr_connect_client_async(chr);
         return 0;
     } else {
diff --git a/include/chardev/char-socket.h b/include/chardev/char-socket.h
index 0708ca6fa9..f464c4c1c7 100644
--- a/include/chardev/char-socket.h
+++ b/include/chardev/char-socket.h
@@ -33,6 +33,8 @@ 
 
 #define TCP_MAX_FDS 16
 
+#define SOCKET_CHARDEV_RECONNECT_MS_MIN 5
+
 typedef struct {
     char buf[21];
     size_t buflen;
@@ -74,7 +76,8 @@  struct SocketChardev {
     bool is_websock;
 
     GSource *reconnect_timer;
-    int64_t reconnect_time;
+    int64_t reconnect_ms_max;
+    int64_t reconnect_ms;
     bool connect_err_reported;
 
     QIOTask *connect_task;