virtio-serial: Flow control

index 3bd2b79..7c12230 100644
@@ -20,11 +20,11 @@ typedef struct VirtConsole {
 
 
 /* Callback function that's called when the guest sends us data */
-static void flush_buf(VirtIOSerialPort *port, const uint8_t *buf, size_t len)
+static ssize_t flush_buf(VirtIOSerialPort *port, const uint8_t *buf, size_t len)
 {
     VirtConsole *vcon = DO_UPCAST(VirtConsole, port, port);
 
-    qemu_chr_write(vcon->chr, buf, len);
+    return qemu_chr_write(vcon->chr, buf, len);
 }
 
 /* Readiness of the guest to accept data on a port */

index ff08c40..9cc0fb3 100644
@@ -137,10 +137,11 @@ struct VirtIOSerialPortInfo {
 
     /*
      * Guest wrote some data to the port. This data is handed over to
-     * the app via this callback.  The app is supposed to consume all
-     * the data that is presented to it.
+     * the app via this callback.  The app can return a size less than
+     * 'len'.  In this case, throttling will be enabled for this port.
      */
-    void (*have_data)(VirtIOSerialPort *port, const uint8_t *buf, size_t len);
+    ssize_t (*have_data)(VirtIOSerialPort *port, const uint8_t *buf,
+                         size_t len);
 };
 
 /* Interface to the virtio-serial bus */

index 6726f72..c5ce60b 100644
@@ -126,24 +126,49 @@ static void discard_vq_data(VirtQueue *vq, VirtIODevice *vdev)
 static void do_flush_queued_data(VirtIOSerialPort *port, VirtQueue *vq,
                                  VirtIODevice *vdev)
 {
-    VirtQueueElement elem;
-
     assert(port);
     assert(virtio_queue_ready(vq));
 
-    while (!port->throttled && virtqueue_pop(vq, &elem)) {
+    while (!port->throttled) {
         unsigned int i;
 
-        for (i = 0; i < elem.out_num; i++) {
-            size_t buf_size;
-
-            buf_size = elem.out_sg[i].iov_len;
+        /* Pop an elem only if we haven't left off a previous one mid-way */
+        if (!port->elem.out_num) {
+            if (!virtqueue_pop(vq, &port->elem)) {
+                break;
+            }
+            port->iov_idx = 0;
+            port->iov_offset = 0;
+        }
 
-            port->info->have_data(port,
-                                  elem.out_sg[i].iov_base,
-                                  buf_size);
+        for (i = port->iov_idx; i < port->elem.out_num; i++) {
+            size_t buf_size;
+            ssize_t ret;
+
+            buf_size = port->elem.out_sg[i].iov_len - port->iov_offset;
+            ret = port->info->have_data(port,
+                                        port->elem.out_sg[i].iov_base
+                                          + port->iov_offset,
+                                        buf_size);
+            if (ret < 0 && ret != -EAGAIN) {
+                /* We don't handle any other type of errors here */
+                abort();
+            }
+            if (ret == -EAGAIN || (ret >= 0 && ret < buf_size)) {
+                virtio_serial_throttle_port(port, true);
+                port->iov_idx = i;
+                if (ret > 0) {
+                    port->iov_offset += ret;
+                }
+                break;
+            }
+            port->iov_offset = 0;
         }
-        virtqueue_push(vq, &elem, 0);
+        if (port->throttled) {
+            break;
+        }
+        virtqueue_push(vq, &port->elem, 0);
+        port->elem.out_num = 0;
     }
     virtio_notify(vdev, vq);
 }

@@ -505,6 +530,7 @@ static void virtio_serial_save(QEMUFile *f, void *opaque)
         qemu_put_be32s(f, &port->id);
         qemu_put_byte(f, port->guest_connected);
         qemu_put_byte(f, port->host_connected);
+        /* TODO: Save & load port->elem and port->iov* */
     }
 }
 

@@ -709,6 +735,8 @@ static int virtser_port_qdev_init(DeviceState *qdev, DeviceInfo *base)
         port->guest_connected = true;
     }
 
+    port->elem.out_num = 0;
+
     QTAILQ_INSERT_TAIL(&port->vser->ports, port, next);
     port->ivq = port->vser->ivqs[port->id];
     port->ovq = port->vser->ovqs[port->id];

index 9cc0fb3..5f41706 100644
@@ -102,6 +102,23 @@ struct VirtIOSerialPort {
      */
     uint32_t id;
 
+    /*
+     * This is the elem that we pop from the virtqueue.  A slow
+     * backend that consumes guest data (e.g. the file backend for
+     * qemu chardevs) can cause the guest to block till all the output
+     * is flushed.  This isn't desired, so we keep a note of the last
+     * element popped and continue consuming it once the backend
+     * becomes writable again.
+     */
+    VirtQueueElement elem;
+
+    /*
+     * The index and the offset into the iov buffer that was popped in
+     * elem above.
+     */
+    unsigned int iov_idx;
+    size_t iov_offset;
+
     /* Identify if this is a port that binds with hvc in the guest */
     uint8_t is_console;
 

index 7c12230..4c27e4c 100644
@@ -18,6 +18,16 @@ typedef struct VirtConsole {
     CharDriverState *chr;
 } VirtConsole;
 
+/*
+ * Callback function that's called from chardevs when backend becomes
+ * writable.
+ */
+static void chr_write_unblocked(void *opaque)
+{
+    VirtConsole *vcon = opaque;
+
+    virtio_serial_throttle_port(&vcon->port, false);
+}
 
 /* Callback function that's called when the guest sends us data */
 static ssize_t flush_buf(VirtIOSerialPort *port, const uint8_t *buf, size_t len)

@@ -61,6 +71,7 @@ static QemuChrHandlers chr_handlers = {
     .fd_can_read = chr_can_read,
     .fd_read = chr_read,
     .fd_event = chr_event,
+    .fd_write_unblocked = chr_write_unblocked,
 };
 
 static int generic_port_init(VirtConsole *vcon, VirtIOSerialDevice *dev)