[PATCH v2] Bluetooth: Fix l2cap_tx_window_full

[PATCH v2] Bluetooth: Fix l2cap_tx_window_full

[Luiz Augusto von Dentz]

From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

l2cap_tx_window_full is not checking the window limit properly, first it
computes based on sequence numbers which doesn't take into account the
ReqSeq and always assume 64 not the real window size.

To fix this now it just checks if the number of unacked frames is >= of
tx window which is much simpler.

Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
---
Add debug

 include/net/bluetooth/l2cap.h |   10 +++-------
 1 files changed, 3 insertions(+), 7 deletions(-)

diff --git a/include/net/bluetooth/l2cap.h b/include/net/bluetooth/l2cap.h
index 42fdbb8..bcfddb2 100644
--- a/include/net/bluetooth/l2cap.h
+++ b/include/net/bluetooth/l2cap.h
@@ -661,14 +661,10 @@ static inline __u16 __next_seq(struct l2cap_chan *chan, __u16 seq)
 
 static inline int l2cap_tx_window_full(struct l2cap_chan *ch)
 {
-	int sub;
+	BT_DBG("chan %p unacked %d tx_win %d", ch, ch->unacked_frames,
+							ch->remote_tx_win);
 
-	sub = (ch->next_tx_seq - ch->expected_ack_seq) % 64;
-
-	if (sub < 0)
-		sub += 64;
-
-	return sub == ch->remote_tx_win;
+	return ch->unacked_frames >= ch->remote_tx_win;
 }
 
 static inline __u16 __get_reqseq(struct l2cap_chan *chan, __u32 ctrl)
-- 
1.7.7.6

Re: [PATCH v2] Bluetooth: Fix l2cap_tx_window_full

[Ulisses Furquim]

Hi Luiz,

On Wed, Feb 8, 2012 at 11:47 AM, Luiz Augusto von Dentz
<luiz.dentz@gmail.com> wrote:
> From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
>
> l2cap_tx_window_full is not checking the window limit properly, first it
> computes based on sequence numbers which doesn't take into account the
> ReqSeq and always assume 64 not the real window size.
>
> To fix this now it just checks if the number of unacked frames is >= of
> tx window which is much simpler.
>
> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> ---
> Add debug
>
>  include/net/bluetooth/l2cap.h |   10 +++-------
>  1 files changed, 3 insertions(+), 7 deletions(-)
>
> diff --git a/include/net/bluetooth/l2cap.h b/include/net/bluetooth/l2cap.h
> index 42fdbb8..bcfddb2 100644
> --- a/include/net/bluetooth/l2cap.h
> +++ b/include/net/bluetooth/l2cap.h
> @@ -661,14 +661,10 @@ static inline __u16 __next_seq(struct l2cap_chan *chan, __u16 seq)
>
>  static inline int l2cap_tx_window_full(struct l2cap_chan *ch)
>  {
> -       int sub;
> +       BT_DBG("chan %p unacked %d tx_win %d", ch, ch->unacked_frames,
> +                                                       ch->remote_tx_win);
>
> -       sub = (ch->next_tx_seq - ch->expected_ack_seq) % 64;
> -
> -       if (sub < 0)
> -               sub += 64;
> -
> -       return sub == ch->remote_tx_win;
> +       return ch->unacked_frames >= ch->remote_tx_win;
>  }
>
>  static inline __u16 __get_reqseq(struct l2cap_chan *chan, __u32 ctrl)
> --
> 1.7.7.6

Looks good to me.

Regards,

-- 
Ulisses Furquim
ProFUSION embedded systems
http://profusion.mobi
Mobile: +55 19 9250 0942
Skype: ulissesffs

Re: [PATCH v2] Bluetooth: Fix l2cap_tx_window_full

[Andrei Emeltchenko]

Hi all,

On Wed, Feb 08, 2012 at 11:54:54AM -0200, Ulisses Furquim wrote:
> Hi Luiz,
> 
> On Wed, Feb 8, 2012 at 11:47 AM, Luiz Augusto von Dentz
> <luiz.dentz@gmail.com> wrote:
> > From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> >
> > l2cap_tx_window_full is not checking the window limit properly, first it
> > computes based on sequence numbers which doesn't take into account the
> > ReqSeq and always assume 64 not the real window size.
> >
> > To fix this now it just checks if the number of unacked frames is >= of
> > tx window which is much simpler.
> >
> > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > ---
> > Add debug
> >
> >  include/net/bluetooth/l2cap.h |   10 +++-------
> >  1 files changed, 3 insertions(+), 7 deletions(-)
> >
> > diff --git a/include/net/bluetooth/l2cap.h b/include/net/bluetooth/l2cap.h
> > index 42fdbb8..bcfddb2 100644
> > --- a/include/net/bluetooth/l2cap.h
> > +++ b/include/net/bluetooth/l2cap.h
> > @@ -661,14 +661,10 @@ static inline __u16 __next_seq(struct l2cap_chan *chan, __u16 seq)
> >
> >  static inline int l2cap_tx_window_full(struct l2cap_chan *ch)
> >  {
> > -       int sub;
> > +       BT_DBG("chan %p unacked %d tx_win %d", ch, ch->unacked_frames,
> > +                                                       ch->remote_tx_win);
> >
> > -       sub = (ch->next_tx_seq - ch->expected_ack_seq) % 64;
> > -
> > -       if (sub < 0)
> > -               sub += 64;
> > -
> > -       return sub == ch->remote_tx_win;
> > +       return ch->unacked_frames >= ch->remote_tx_win;
> >  }
> >
> >  static inline __u16 __get_reqseq(struct l2cap_chan *chan, __u32 ctrl)
> > --
> > 1.7.7.6
> 
> Looks good to me.

Same here, I am wondering why it was done in a such complex way? Maybe
there is some logic behind?

Best regards 
Andrei Emeltchenko 

Re: [PATCH v2] Bluetooth: Fix l2cap_tx_window_full

[Gustavo Padovan]

Hi Andrei,

* Andrei Emeltchenko <andrei.emeltchenko.news@gmail.com> [2012-02-08 15:59:11 +0200]:

> Hi all,
> 
> On Wed, Feb 08, 2012 at 11:54:54AM -0200, Ulisses Furquim wrote:
> > Hi Luiz,
> > 
> > On Wed, Feb 8, 2012 at 11:47 AM, Luiz Augusto von Dentz
> > <luiz.dentz@gmail.com> wrote:
> > > From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > >
> > > l2cap_tx_window_full is not checking the window limit properly, first it
> > > computes based on sequence numbers which doesn't take into account the
> > > ReqSeq and always assume 64 not the real window size.
> > >
> > > To fix this now it just checks if the number of unacked frames is >= of
> > > tx window which is much simpler.
> > >
> > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > > ---
> > > Add debug
> > >
> > >  include/net/bluetooth/l2cap.h |   10 +++-------
> > >  1 files changed, 3 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/include/net/bluetooth/l2cap.h b/include/net/bluetooth/l2cap.h
> > > index 42fdbb8..bcfddb2 100644
> > > --- a/include/net/bluetooth/l2cap.h
> > > +++ b/include/net/bluetooth/l2cap.h
> > > @@ -661,14 +661,10 @@ static inline __u16 __next_seq(struct l2cap_chan *chan, __u16 seq)
> > >
> > >  static inline int l2cap_tx_window_full(struct l2cap_chan *ch)
> > >  {
> > > -       int sub;
> > > +       BT_DBG("chan %p unacked %d tx_win %d", ch, ch->unacked_frames,
> > > +                                                       ch->remote_tx_win);
> > >
> > > -       sub = (ch->next_tx_seq - ch->expected_ack_seq) % 64;
> > > -
> > > -       if (sub < 0)
> > > -               sub += 64;
> > > -
> > > -       return sub == ch->remote_tx_win;
> > > +       return ch->unacked_frames >= ch->remote_tx_win;
> > >  }
> > >
> > >  static inline __u16 __get_reqseq(struct l2cap_chan *chan, __u32 ctrl)
> > > --
> > > 1.7.7.6
> > 
> > Looks good to me.
> 
> Same here, I am wondering why it was done in a such complex way? Maybe
> there is some logic behind?

I followed the specification, next_tx_seq minus expecpted_ack_seq is the
definition of the current size of the tx_window.

Anyway,

Acked-by: Gustavo F. Padovan <padovan@profusion.mobi>

	Gustavo

Re: [PATCH v2] Bluetooth: Fix l2cap_tx_window_full

[Luiz Augusto von Dentz]

Hi Gustavo,

On Wed, Feb 8, 2012 at 5:38 PM, Gustavo Padovan <padovan@profusion.mobi> wrote:
>
> I followed the specification, next_tx_seq minus expecpted_ack_seq is the
> definition of the current size of the tx_window.
>
> Anyway,
>
> Acked-by: Gustavo F. Padovan <padovan@profusion.mobi>

I notice it, It was actually possible to fix this by using
__seq_offset but then I notice in Mat's patch he just use
unacked_frames which is way simpler to read/use. Btw, there is still
something else preventing me to use bigger MTU with OBEX, I can use up
to 32k in kvm but only 16k if I boot it directly, at some point it
just stall in the transmitter even though the receiver acks and the
window is not full.

-- 
Luiz Augusto von Dentz

Re: [PATCH v2] Bluetooth: Fix l2cap_tx_window_full

[Andrei Emeltchenko]

On Wed, Feb 08, 2012 at 03:47:35PM +0200, Luiz Augusto von Dentz wrote:
> From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> 
> l2cap_tx_window_full is not checking the window limit properly, first it
> computes based on sequence numbers which doesn't take into account the
> ReqSeq and always assume 64 not the real window size.
> 
> To fix this now it just checks if the number of unacked frames is >= of
> tx window which is much simpler.

So what is the status with this patch?

Best regards 
Andrei Emeltchenko 
> 
> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> ---
> Add debug
> 
>  include/net/bluetooth/l2cap.h |   10 +++-------
>  1 files changed, 3 insertions(+), 7 deletions(-)
> 
> diff --git a/include/net/bluetooth/l2cap.h b/include/net/bluetooth/l2cap.h
> index 42fdbb8..bcfddb2 100644
> --- a/include/net/bluetooth/l2cap.h
> +++ b/include/net/bluetooth/l2cap.h
> @@ -661,14 +661,10 @@ static inline __u16 __next_seq(struct l2cap_chan *chan, __u16 seq)
>  
>  static inline int l2cap_tx_window_full(struct l2cap_chan *ch)
>  {
> -	int sub;
> +	BT_DBG("chan %p unacked %d tx_win %d", ch, ch->unacked_frames,
> +							ch->remote_tx_win);
>  
> -	sub = (ch->next_tx_seq - ch->expected_ack_seq) % 64;
> -
> -	if (sub < 0)
> -		sub += 64;
> -
> -	return sub == ch->remote_tx_win;
> +	return ch->unacked_frames >= ch->remote_tx_win;
>  }
>  
>  static inline __u16 __get_reqseq(struct l2cap_chan *chan, __u32 ctrl)
> -- 
> 1.7.7.6
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html