Return-Path: <cyrus@holtmann.org>
Date: Thu, 10 Mar 2011 09:32:06 +0200
From: Johan Hedberg <johan.hedberg@gmail.com>
To: Ed Tomlinson <edt@aei.ca>
Cc: "Justin P. Mattock" <justinmattock@gmail.com>,
	Mikko Vinni <mmvinni@yahoo.com>,
	linux-bluetooth@vger.kernel.org,
	"Gustavo F. Padovan" <padovan@profusion.mobi>
Subject: Re: re "bluetooth disabled" and "[BUG] usb problems in .38-rc3+"
Message-ID: <20110310073206.GA1033@jh-x301>
References: <635561.59590.qm@web161820.mail.bf1.yahoo.com>
 <359639.21781.qm@web161807.mail.bf1.yahoo.com>
 <4D77EFEB.1020804@gmail.com>
 <201103092154.22796.edt@aei.ca>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <201103092154.22796.edt@aei.ca>
Sender: linux-bluetooth-owner@vger.kernel.org
List-ID: <linux-bluetooth.vger.kernel.org>

Hi,

On Wed, Mar 09, 2011, Ed Tomlinson wrote:
> > diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
> > index 9c4541b..f690eb5 100644
> > --- a/net/bluetooth/hci_core.c
> > +++ b/net/bluetooth/hci_core.c
> > @@ -95,18 +95,17 @@ void hci_req_complete(struct hci_dev *hdev, __u16 
> > cmd, int result)
> >   {
> >   	BT_DBG("%s command 0x%04x result 0x%2.2x", hdev->name, cmd, result);
> > 
> > +	if (hdev->req_status == HCI_REQ_PEND) {
> > +		hdev->req_result = result;
> > +		hdev->req_status = HCI_REQ_DONE;
> > +		wake_up_interruptible(&hdev->req_wait_q);
> > +	}
> >   	/* If the request has set req_last_cmd (typical for multi-HCI
> >   	 * command requests) check if the completed command matches
> >   	 * this, and if not just return. Single HCI command requests
> >   	 * typically leave req_last_cmd as 0 */
> >   	if (hdev->req_last_cmd && cmd != hdev->req_last_cmd)
> >   		return;
> > -
> > -	if (hdev->req_status == HCI_REQ_PEND) {
> > -		hdev->req_result = result;
> > -		hdev->req_status = HCI_REQ_DONE;
> > -		wake_up_interruptible(&hdev->req_wait_q);
> > -	}
> >   }
> > 
> >   static void hci_req_cancel(struct hci_dev *hdev, int err)
> > 
> 
> Works here too - bluetooth comes up enabled.  On to Linus?

All the patch does is it re-breaks the __hci_request function by
removing proper synchronization from it (again) and thereby hides the
real bug which is the kernel sending commands to the controller before
the HCI_Reset command has completed.

Johan