Hi,
On Wed, Oct 29, 2014 at 12:03 PM, Bharat Panda <bharat.panda@xxxxxxxxxxx> wrote:
> Checks for NULL session before calling avdtp_get_device, as
> setup->session will be NULL if configuraion was aborted.
Could you please add the backtrace of the crash?
> ---
> profiles/audio/a2dp.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/profiles/audio/a2dp.c b/profiles/audio/a2dp.c
> index c9dac9a..810d1e6 100644
> --- a/profiles/audio/a2dp.c
> +++ b/profiles/audio/a2dp.c
> @@ -379,7 +379,7 @@ static void stream_state_changed(struct avdtp_stream *stream,
> static gboolean auto_config(gpointer data)
> {
> struct a2dp_setup *setup = data;
> - struct btd_device *dev = avdtp_get_device(setup->session);
> + struct btd_device *dev = NULL;
> struct btd_service *service;
>
> /* Check if configuration was aborted */
> @@ -389,6 +389,10 @@ static gboolean auto_config(gpointer data)
> if (setup->err != NULL)
> goto done;
>
> + /* session will be NULL if configuration was aborted */
> + if (setup->session)
> + dev = avdtp_get_device(setup->session);
> +
> avdtp_stream_add_cb(setup->session, setup->stream,
> stream_state_changed, setup->sep);
Perhaps this is caused by g_idle_add in endpoint_setconf_ind, but by
looking at it seems that we have some better options as either call it
auto_config directly which may cause the session to be freed before
returning to avdtp_setconf_cmd, take a extra reference which might
still cause some problems since the session would still exists while
we may be disconnected already, or the one I prefer is to take the id
of g_idle_add and do g_source_remove if the setup is aborted.
--
Luiz Augusto von Dentz
--
To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
