[Spice-devel] [spice PATCH v1 1/3] char-device: fix usage of free/unref on WriteBuffer

[Fabiano Fidêncio]

On Thu, Nov 12, 2015 at 5:00 PM, Victor Toso <victortoso at redhat.com> wrote:
> There are places were the could should definetly free the
> SpiceCharDeviceWriteBuffer and places that it should only unref it. The
> current use of spice_char_device_write_buffer_free was missleading.
>
> This patch creates the spice_char_device_write_buffer_unref and properly
> call these two functions.
>
> Related: https://bugs.freedesktop.org/show_bug.cgi?id=91350
> ---
>  server/char_device.c | 34 ++++++++++++++++++++++------------
>  1 file changed, 22 insertions(+), 12 deletions(-)
>
> diff --git a/server/char_device.c b/server/char_device.c
> index 54357f0..ee52e90 100644
> --- a/server/char_device.c
> +++ b/server/char_device.c
> @@ -81,6 +81,7 @@ enum {
>   * destroyed during a callback */
>  static void spice_char_device_state_ref(SpiceCharDeviceState *char_dev);
>  static void spice_char_device_state_unref(SpiceCharDeviceState *char_dev);
> +static void spice_char_device_write_buffer_unref(SpiceCharDeviceWriteBuffer *write_buf);
>
>  static void spice_char_dev_write_retry(void *opaque);
>
> @@ -91,10 +92,11 @@ typedef struct SpiceCharDeviceMsgToClientItem {
>
>  static void spice_char_device_write_buffer_free(SpiceCharDeviceWriteBuffer *buf)
>  {
> -    if (--buf->refs == 0) {
> -        free(buf->buf);
> -        free(buf);
> -    }
> +    if (!buf)
> +        return;

Please, keep the check explicit.

> +
> +    free(buf->buf);
> +    free(buf);
>  }
>
>  static void write_buffers_queue_free(Ring *write_queue)
> @@ -117,9 +119,11 @@ static void spice_char_device_write_buffer_pool_add(SpiceCharDeviceState *dev,
>          buf->origin = WRITE_BUFFER_ORIGIN_NONE;
>          buf->client = NULL;
>          ring_add(&dev->write_bufs_pool, &buf->link);
> -    } else {
> -        --buf->refs;
> +        return;
>      }
> +
> +    /* Buffer still being used - just unref for the caller */
> +    spice_char_device_write_buffer_unref(buf);
>  }
>
>  static void spice_char_device_client_send_queue_free(SpiceCharDeviceState *dev,
> @@ -593,6 +597,15 @@ static SpiceCharDeviceWriteBuffer *spice_char_device_write_buffer_ref(SpiceCharD
>      return write_buf;
>  }
>
> +static void spice_char_device_write_buffer_unref(SpiceCharDeviceWriteBuffer *write_buf)
> +{
> +    spice_assert(write_buf);
> +
> +    write_buf->refs--;
> +    if (write_buf->refs == 0)
> +        spice_char_device_write_buffer_free(write_buf);

Not important, but you could keep the styling:
if (--write_buf->refs != 0)
    returnl
spice_char_device_write_buffer_free(write_buf);

> +}
> +
>  void spice_char_device_write_buffer_add(SpiceCharDeviceState *dev,
>                                          SpiceCharDeviceWriteBuffer *write_buf)
>  {
> @@ -619,8 +632,7 @@ void spice_char_device_write_buffer_release(SpiceCharDeviceState *dev,
>      spice_assert(!ring_item_is_linked(&write_buf->link));
>      if (!dev) {
>          spice_printerr("no device. write buffer is freed");
> -        free(write_buf->buf);
> -        free(write_buf);
> +        spice_char_device_write_buffer_free(write_buf);
>          return;
>      }
>
> @@ -727,9 +739,7 @@ void spice_char_device_state_destroy(SpiceCharDeviceState *char_dev)
>      }
>      write_buffers_queue_free(&char_dev->write_queue);
>      write_buffers_queue_free(&char_dev->write_bufs_pool);
> -    if (char_dev->cur_write_buf) {
> -        spice_char_device_write_buffer_free(char_dev->cur_write_buf);
> -    }
> +    spice_char_device_write_buffer_free(char_dev->cur_write_buf);
>
>      while (!ring_is_empty(&char_dev->clients)) {
>          RingItem *item = ring_get_tail(&char_dev->clients);
> @@ -895,7 +905,7 @@ static void migrate_data_marshaller_write_buffer_free(uint8_t *data, void *opaqu
>  {
>      SpiceCharDeviceWriteBuffer *write_buf = (SpiceCharDeviceWriteBuffer *)opaque;
>
> -    spice_char_device_write_buffer_free(write_buf);
> +    spice_char_device_write_buffer_unref(write_buf);
>  }
>
>  void spice_char_device_state_migrate_data_marshall(SpiceCharDeviceState *dev,
> --
> 2.5.0
>
> _______________________________________________
> Spice-devel mailing list
> Spice-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/spice-devel

Looks good, ACK!