Hi Jonas,

On 04/14/2017 04:36 PM, Jonas Bonn wrote:
> The release_active_contexts method ask the driver to deactive all
> the active contexts it knows about; however, after doing so, the
> context state needs to be released, as well, so that the contexts
> do not continue to appear to be active.
> ---
>  src/gprs.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/src/gprs.c b/src/gprs.c
> index 6ed1c89..b6e11e8 100644
> --- a/src/gprs.c
> +++ b/src/gprs.c
> @@ -1614,6 +1614,9 @@ static void release_active_contexts(struct ofono_gprs *gprs)
>
>  		if (gc->driver->detach_shutdown != NULL)
>  			gc->driver->detach_shutdown(gc, ctx->context.cid);
> +
> +		pri_reset_context_settings(ctx);
> +		release_context(ctx);

Hmm, this seems wrong.  The original commit (according to git blame, git 
show 05b8fe47) was aimed at PPP contexts.  The gprs_context driver will 
call g_at_ppp_shutdown which in turn will result in eventual 
ppp_disconnect.  The context driver will then call 
ofono_gprs_cid_deactivated.

For the other context drivers, it was assumed that the modem will be 
sane enough to issue a +CGEV with a context deactivation.

I'm assuming you need this for QMI as well, so you might want to use 
similar semantics.  E.g. have detach_shutdown call STOP_NET and then 
call ofono_gprs_cid_deactivated.

>  	}
>  }
>
>

Regards,
-Denis