Hi Johan,
--------------------------------------------------
From: "Johan Hedberg" <johan.hedberg@xxxxxxxxx>
Sent: Friday, March 09, 2012 5:12 AM
To: "Jaganath Kanakkassery" <jaganath.k@xxxxxxxxxxx>
Cc: <linux-bluetooth@xxxxxxxxxxxxxxx>
Subject: Re: [PATCH obexd 1/2] gobex: flush tx_queue before disconnection
Hi Jaganath,
On Thu, Mar 08, 2012, Jaganath Kanakkassery wrote:
@@ -1335,6 +1335,12 @@ void g_obex_unref(GObex *obex)
g_slist_free_full(obex->req_handlers, g_free);
+ do {
+ ret = write_data(obex->io, G_IO_OUT, obex);
+ if (obex->pending_req && obex->pending_req->cancelled)
+ break;
+ } while(ret);
This is not ok since we should only attempt writing to the transport if
G_IO_OUT is *really* set and not just fake it. Otherwise the call might
block which is not acceptable for the way gobex is designed (to be used
with a single async mainloop).
This actually fixes the issue abort req not sending if user cancels the
transfer because in user cancel scenario, ABORT req will be queued
but before G_IO_OUT happens the transport will be disconnected.
This is actually required to pass a pts test case because pts requires
ABORT before disconnection.
So I have two solutions after considering your and Luiz's opinion. '
1. In g_obex_send_internal() if the packet is ABORT we should send it
immediately using write_data(). Since ABORT packet size is small I think
it is ok to block the write.
2. If user cancels transfer just send ABORT and disconnect transport only
after getting the response. But waiting for ABORT response is not good I
feel.
Please let me know if you have better method.
Thanks,
Jaganath
--
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
