Return-Path: <cyrus@holtmann.org>
Date: Thu, 20 Nov 2014 13:38:13 +0200
From: Johan Hedberg <johan.hedberg@gmail.com>
To: Steven Walter <stevenrwalter@gmail.com>
Cc: marcel@holtmann.org, gustavo@padovan.org,
	linux-bluetooth@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3] Bluetooth: automatically flushable packets aren't
 allowed on LE links
Message-ID: <20141120113813.GA23288@t440s.lan>
References: <1416419951-27659-1-git-send-email-stevenrwalter@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <1416419951-27659-1-git-send-email-stevenrwalter@gmail.com>
List-ID: <linux-bluetooth.vger.kernel.org>

Hi Steven,

On Wed, Nov 19, 2014, Steven Walter wrote:
> The bluetooth spec states that automatically flushable packets may not
> be sent over a LE-U link.
> 
> Signed-off-by: Steven Walter <stevenrwalter@gmail.com>
> ---
>  net/bluetooth/l2cap_core.c | 14 ++++++++------
>  1 file changed, 8 insertions(+), 6 deletions(-)
> 
> diff --git a/net/bluetooth/l2cap_core.c b/net/bluetooth/l2cap_core.c
> index 4af3821..7c4350f 100644
> --- a/net/bluetooth/l2cap_core.c
> +++ b/net/bluetooth/l2cap_core.c
> @@ -764,7 +764,7 @@ static void l2cap_send_cmd(struct l2cap_conn *conn, u8 ident, u8 code, u16 len,
>  	if (!skb)
>  		return;
>  
> -	if (lmp_no_flush_capable(conn->hcon->hdev))
> +	if (lmp_no_flush_capable(conn->hcon->hdev) || conn->hcon->type == LE_LINK)
>  		flags = ACL_START_NO_FLUSH;
>  	else
>  		flags = ACL_START;
> @@ -784,7 +784,7 @@ static bool __chan_is_moving(struct l2cap_chan *chan)
>  static void l2cap_do_send(struct l2cap_chan *chan, struct sk_buff *skb)
>  {
>  	struct hci_conn *hcon = chan->conn->hcon;
> -	u16 flags;
> +	u16 flags = ACL_START;
>  
>  	BT_DBG("chan %p, skb %p len %d priority %u", chan, skb, skb->len,
>  	       skb->priority);
> @@ -798,11 +798,13 @@ static void l2cap_do_send(struct l2cap_chan *chan, struct sk_buff *skb)
>  		return;
>  	}
>  
> -	if (!test_bit(FLAG_FLUSHABLE, &chan->flags) &&
> -	    lmp_no_flush_capable(hcon->hdev))
> +	if (hcon->type == LE_LINK) {
> +		/* LE-U does not support auto-flushing packets */
>  		flags = ACL_START_NO_FLUSH;
> -	else
> -		flags = ACL_START;
> +	} else if (!test_bit(FLAG_FLUSHABLE, &chan->flags) &&
> +		    lmp_no_flush_capable(hcon->hdev)) {
> +		flags = ACL_START_NO_FLUSH;
> +	}

I think Marcel was after just providing a clarifying code comment in
both places - having two branches of an if-statement doing exactly the
same thing looks a bit weird to me. To make thins completely clear I'd
suggest adding a simple helper function that you can call from both
places to get the needed flags, something like the following:

static u16 get_acl_flags(struct hci_conn *hcon, struct l2cap_chan *chan)
{
	/* LE-U does not support auto-flushing packets */
	if (hcon->type == LE_LINK)
		return ACL_START_NO_FLUSH;

	/* If non-flushable packets are not supported don't request them */
	if (!lmp_no_flush_capable(hcon->hdev))
		 return ACL_START;

	/* If the chan has requested auto-flushing go with that */
	if (chan && test_bit(FLAG_FLUSHABLE, &chan->flags))
		return ACL_START;

	/* Otherwise go with a non-flushable packet */
        return ACL_START_NO_FLUSH;
}

This way we'd avoid complex if-statements and can clearly document each
condition independently.

Johan