Return-Path: <cyrus@holtmann.org>
Subject: Re: regression introduced on v2.6.30-rc1
From: Marcel Holtmann <marcel@holtmann.org>
To: Luiz Augusto von Dentz <luiz.dentz@gmail.com>
Cc: linux-bluetooth@vger.kernel.org
In-Reply-To: <2d5a2c100906230801q6bd68397y34b5aa138fee5fe2@mail.gmail.com>
References: <2d5a2c100906090617m167d3815pae998d06bdbd6646@mail.gmail.com>
	 <1245595707.15367.66.camel@violet>
	 <2d5a2c100906210930q394fcf77hcd2bf39c26ae74f8@mail.gmail.com>
	 <1245603529.15367.68.camel@violet>
	 <2d5a2c100906211047n31b9d4dajac290601c184380b@mail.gmail.com>
	 <1245611096.15367.76.camel@violet>
	 <2d5a2c100906221449y6e37cb19x1bc27085bac318f4@mail.gmail.com>
	 <2d5a2c100906221608v5e3e3d8aydf66b13826ad88fa@mail.gmail.com>
	 <2d5a2c100906230651m679ccfat5bf4bfb20c130955@mail.gmail.com>
	 <1245768036.15367.78.camel@violet>
	 <2d5a2c100906230801q6bd68397y34b5aa138fee5fe2@mail.gmail.com>
Content-Type: text/plain
Date: Tue, 23 Jun 2009 17:06:12 +0200
Message-Id: <1245769572.15367.80.camel@violet>
Mime-Version: 1.0
Sender: linux-bluetooth-owner@vger.kernel.org
List-ID: <linux-bluetooth.vger.kernel.org>

Hi Luiz,

> > what about just triggering the timer and then sending DISC for DLCI 0. I
> > don't see a big benefit for this reference counting overhead.
> >
> > When we send the DISC, we will receive UA and thus get the required
> > rfcomm_session_put() that then leads to the ACL disconnect.
> 
> Hmm, that seems to work too and is what we are doing in case of
> receiving a DM, but there is a chance that the remote stack doesn't
> respond the DISC in that case we can also use the session timer to
> timeout. Anyway I don't think such broken stack exist, although we do
> set timeout when sending DISC to a specific DLCI, so lets leave this
> for latter when he actually have a real offender which doesn't respond
> with UA.
> 
> Moving ahead, what about the timeout, 20 seconds seems too much
> doesn't it? Are you fine with the places where I clear the timer?

the timeout should be 2 seconds. At most 5 seconds.

I still prefer if we just send DISC and then wait for the remote stack
to do the right thing. If it doesn't then that stack is so broken that
whatever we try to do is wrong anyway at that point.

Regards

Marcel