Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Claudio Takahasi]

The following changes since commit e7b1d878d15c0eb1bc0109c19cbdf5f84a29394d:

  Fix coding style issues in attribute client code (2010-09-22 16:33:15 -0400)

are available in the git repository at:
  git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

Claudio Takahasi (23):
      Add LE start and stop scanning
      Remove RSSI field from the advertising report event
      Decoding the RSSI parameter from the advertising report event
      Send Discovering property "FALSE" when the interleave finishes
      Add length argument on hciops start discovery function
      Stop inquiry using the length parameter
      Fix remote name resolution for interleave discovery
      Add Write LE host supported function
      Set the LE host supported and disable simultaneous LE and BR/EDR flags
      Add extended feature mask constants definition
      Read the local extended features
      Stop LE scanning when discovery is suspended
      Rename hciops {start, stop}_discovery to {start, stop}_inquiry
      Don't enter on interleave mode if there isn't active sessions
      Code cleanup: improving inquiry logic
      Clear the remote device found list in the state transition
      Fix periodic inquiry signals
      Fixing DeviceDisappeared signal
      Postpone discovery if still resolving names
      Add adapter discovery type function
      Do not send another Discovering TRUE signal if still resolving names
      Forcing periodic inquiry exit
      Fix interleave scanning

Vinicius Costa Gomes (1):
      Add BR/EDR LE interleaved discovery

 lib/hci.c        |   29 +++++
 lib/hci.h        |   17 +++-
 lib/hci_lib.h    |    1 +
 plugins/hciops.c |   53 +++++++++-
 src/adapter.c    |  302 +++++++++++++++++++++++++++++++++++++-----------------
 src/adapter.h    |   37 ++++---
 src/dbus-hci.c   |   45 +++++++--
 src/dbus-hci.h   |    1 +
 src/security.c   |  133 ++++++++++++++----------
 9 files changed, 436 insertions(+), 182 deletions(-)


Implements interleave discovery for dual mode adapters(LE and BR/EDR).
Now, the behavior is:
1) for single mode: LE scanning for 5.12 seconds every DiscoverSchedulerInterval
2) for BR/EDR only: still the same
3) for dual mode: inquiry(5.12) sec, LE scanning(5.12) sec, resolve
names. The next discovery interval is controlled by
DiscoverSchedulerInterval.

Signals still the same, the idea is try to hide from the users/apps
the discovery internals. No mater the adapter type(LE only, BR/EDR or
Dual Mode)
PropertyChanged("Discovering", TRUE/FALSE) is sent to notify when a
new discovery "session" starts/finishes. Found LE devices are reported
using DeviceFound() signals.

Regards,
Claudio
-- 
--
Claudio Takahasi
Instituto Nokia de Tecnologia
Recife - Pernambuco - Brasil
+55 81 30879999

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Luiz Augusto von Dentz]

Hi Claudio,

On Thu, Sep 23, 2010 at 7:19 PM, Claudio Takahasi
<claudio.takahasi@openbossa.org> wrote:
> Implements interleave discovery for dual mode adapters(LE and BR/EDR).
> Now, the behavior is:
> 1) for single mode: LE scanning for 5.12 seconds every DiscoverSchedulerInterval
> 2) for BR/EDR only: still the same
> 3) for dual mode: inquiry(5.12) sec, LE scanning(5.12) sec, resolve
> names. The next discovery interval is controlled by
> DiscoverSchedulerInterval.

Is that enough time to discover devices around? I remember tweaking
the interval so that it doesn't cause too much noise/too many new
devices found or disappearing in each round, so at least to BR/EDR
5.12 sec. sounds too little time, also does LE needs name resolving
too?

-- 
Luiz Augusto von Dentz
Computer Engineer

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Claudio Takahasi]

On Thu, Sep 23, 2010 at 2:55 PM, Luiz Augusto von Dentz
<luiz.dentz@gmail.com> wrote:
> Hi Claudio,
>
> On Thu, Sep 23, 2010 at 7:19 PM, Claudio Takahasi
> <claudio.takahasi@openbossa.org> wrote:
>> Implements interleave discovery for dual mode adapters(LE and BR/EDR).
>> Now, the behavior is:
>> 1) for single mode: LE scanning for 5.12 seconds every DiscoverSchedulerInterval
>> 2) for BR/EDR only: still the same
>> 3) for dual mode: inquiry(5.12) sec, LE scanning(5.12) sec, resolve
>> names. The next discovery interval is controlled by
>> DiscoverSchedulerInterval.
>
> Is that enough time to discover devices around? I remember tweaking
> the interval so that it doesn't cause too much noise/too many new
> devices found or disappearing in each round, so at least to BR/EDR
> 5.12 sec. sounds too little time, also does LE needs name resolving
> too?
>
> --
> Luiz Augusto von Dentz
> Computer Engineer
>

Hi Luiz,

5.12sec is based on the SPEC recommendation(core page 1746), we can
change it to 10.24s or add a config option.
For LE, there is a characteristic called "Device Name" UUID 0x2A00,
this piece is missing.

Regards,
Claudio
-- 
--
Claudio Takahasi
Instituto Nokia de Tecnologia
Recife - Pernambuco - Brasil
+55 81 30879999

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Johan Hedberg]

Hi Claudio,

On Thu, Sep 23, 2010, Claudio Takahasi wrote:
>       Rename hciops {start, stop}_discovery to {start, stop}_inquiry

I'm not sure about this patch. Wasn't the idea to hide the device
discovery details from the higher layers? I.e. the method could be
called discovery and it'd take care of both BR/EDR and LE discovery
methods.

Johan

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Claudio Takahasi]

On Wed, Sep 29, 2010 at 10:45 AM, Johan Hedberg <johan.hedberg@gmail.com> wrote:
> Hi Claudio,
>
> On Thu, Sep 23, 2010, Claudio Takahasi wrote:
>>       Rename hciops {start, stop}_discovery to {start, stop}_inquiry
>
> I'm not sure about this patch. Wasn't the idea to hide the device
> discovery details from the higher layers? I.e. the method could be
> called discovery and it'd take care of both BR/EDR and LE discovery
> methods.
>
> Johan
>

Hi Johan,

StartDiscovery() triggers all the process. The adapter type: BR/EDR
only/LE only or dual mode is hidden from the higher layers.
So, your suggestion it try to hide all discovery details inside the
hciops as much as possible? I am not sure at the moment if it is
possible, but we can try. We need to clean the code anyway in order to
have a "smooth" migration to the new Bluetooth Management Interface.

Regards,
Claudio

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Johan Hedberg]

Hi Claudio,

On Wed, Sep 29, 2010, Claudio Takahasi wrote:
> StartDiscovery() triggers all the process. The adapter type: BR/EDR
> only/LE only or dual mode is hidden from the higher layers.
> So, your suggestion it try to hide all discovery details inside the
> hciops as much as possible?

That was the idea yes, but I'm not 100% sure that it's the best
approach. Another question is also how do we want to represent this with
the new management interface. Should the kernel do it "all" through a
single request from userspace or should userspace control the separate
BR/EDR and LE discoveries. If it's the former, then having a single
adapter_ops callback makes sense. If it's the later then we should
probably have two separate ones. Either way I now realize that to
implement this with a single callback we'd need to have the HCI event
processing moved over to hciops and that's not something that will
happen very quickly. So maybe it's better to have these as two separate
adapter_ops callbacks for now.

Johan

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Claudio Takahasi]

On Thu, Sep 30, 2010 at 4:39 AM, Johan Hedberg <johan.hedberg@gmail.com> wrote:
> Hi Claudio,
>
> On Wed, Sep 29, 2010, Claudio Takahasi wrote:
>> StartDiscovery() triggers all the process. The adapter type: BR/EDR
>> only/LE only or dual mode is hidden from the higher layers.
>> So, your suggestion it try to hide all discovery details inside the
>> hciops as much as possible?
>
> That was the idea yes, but I'm not 100% sure that it's the best
> approach. Another question is also how do we want to represent this with
> the new management interface. Should the kernel do it "all" through a
> single request from userspace or should userspace control the separate
> BR/EDR and LE discoveries. If it's the former, then having a single
> adapter_ops callback makes sense. If it's the later then we should
> probably have two separate ones. Either way I now realize that to
> implement this with a single callback we'd need to have the HCI event
> processing moved over to hciops and that's not something that will
> happen very quickly. So maybe it's better to have these as two separate
> adapter_ops callbacks for now.
>
> Johan
>

Another point is: scanning needs to be disabled "actively", it is not
possible to inform a scanning length like inquiry, it is necessary to
create a timeout to disable the scanning. Joining inquiry and scanning
inside a single adapter_ops callback will be hard to control the
discovery procedure "states". So for now, I prefer two separate
callbacks due the scanning timeout and HCI events processing.

However, for the new management interface, one single request seems to
be more appropriate but extra requests will be necessary to setup the
scan and inquiry parameters.

Claudio

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Claudio Takahasi]

On Thu, Sep 30, 2010 at 11:28 AM, Claudio Takahasi
<claudio.takahasi@openbossa.org> wrote:
> On Thu, Sep 30, 2010 at 4:39 AM, Johan Hedberg <johan.hedberg@gmail.com> wrote:
>> Hi Claudio,
>>
>> On Wed, Sep 29, 2010, Claudio Takahasi wrote:
>>> StartDiscovery() triggers all the process. The adapter type: BR/EDR
>>> only/LE only or dual mode is hidden from the higher layers.
>>> So, your suggestion it try to hide all discovery details inside the
>>> hciops as much as possible?
>>
>> That was the idea yes, but I'm not 100% sure that it's the best
>> approach. Another question is also how do we want to represent this with
>> the new management interface. Should the kernel do it "all" through a
>> single request from userspace or should userspace control the separate
>> BR/EDR and LE discoveries. If it's the former, then having a single
>> adapter_ops callback makes sense. If it's the later then we should
>> probably have two separate ones. Either way I now realize that to
>> implement this with a single callback we'd need to have the HCI event
>> processing moved over to hciops and that's not something that will
>> happen very quickly. So maybe it's better to have these as two separate
>> adapter_ops callbacks for now.
>>
>> Johan
>>
>
> Another point is: scanning needs to be disabled "actively", it is not
> possible to inform a scanning length like inquiry, it is necessary to
> create a timeout to disable the scanning. Joining inquiry and scanning
> inside a single adapter_ops callback will be hard to control the
> discovery procedure "states". So for now, I prefer two separate
> callbacks due the scanning timeout and HCI events processing.
>
> However, for the new management interface, one single request seems to
> be more appropriate but extra requests will be necessary to setup the
> scan and inquiry parameters.
>
> Claudio
>

New pull request! Patches rebased and a new hciops callback function
has been added to export the write LE host supported command.

The following changes since commit 3a8386d428e385d97b95a890623362321a7e3a19:

  Fix left open socket warning for attribute server (2010-09-30 17:26:55 -0300)

are available in the git repository at:
  git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

Claudio Takahasi (23):
      Add LE start and stop scanning
      Remove RSSI field from the advertising report event
      Decoding the RSSI parameter from the advertising report event
      Send Discovering property "FALSE" when the interleave finishes
      Add length argument on hciops start discovery function
      Stop inquiry using the length parameter
      Fix remote name resolution for interleave discovery
      Add Write LE host supported function
      Set the LE host supported and disable simultaneous LE and BR/EDR flags
      Add extended feature mask constants definition
      Read the local extended features
      Stop LE scanning when discovery is suspended
      Rename hciops {start, stop}_discovery to {start, stop}_inquiry
      Don't enter on interleave mode if there isn't active sessions
      Code cleanup: improving inquiry logic
      Clear the remote device found list in the state transition
      Fix periodic inquiry signals
      Fixing DeviceDisappeared signal
      Postpone discovery if still resolving names
      Add adapter discovery type function
      Do not send another Discovering TRUE signal if still resolving names
      Forcing periodic inquiry exit
      Fix interleave scanning

Vinicius Costa Gomes (1):
      Add BR/EDR LE interleaved discovery

 lib/hci.c        |   29 +++++
 lib/hci.h        |   17 +++-
 lib/hci_lib.h    |    1 +
 plugins/hciops.c |   69 +++++++++++-
 src/adapter.c    |  301 +++++++++++++++++++++++++++++++++++++-----------------
 src/adapter.h    |   38 ++++---
 src/dbus-hci.c   |   45 +++++++--
 src/dbus-hci.h   |    1 +
 src/security.c   |  131 ++++++++++++++----------
 9 files changed, 449 insertions(+), 183 deletions(-)

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Claudio Takahasi]

On Thu, Sep 30, 2010 at 6:51 PM, Claudio Takahasi
<claudio.takahasi@openbossa.org> wrote:
> On Thu, Sep 30, 2010 at 11:28 AM, Claudio Takahasi
> <claudio.takahasi@openbossa.org> wrote:
>> On Thu, Sep 30, 2010 at 4:39 AM, Johan Hedberg <johan.hedberg@gmail.com> wrote:
>>> Hi Claudio,
>>>
>>> On Wed, Sep 29, 2010, Claudio Takahasi wrote:
>>>> StartDiscovery() triggers all the process. The adapter type: BR/EDR
>>>> only/LE only or dual mode is hidden from the higher layers.
>>>> So, your suggestion it try to hide all discovery details inside the
>>>> hciops as much as possible?
>>>
>>> That was the idea yes, but I'm not 100% sure that it's the best
>>> approach. Another question is also how do we want to represent this with
>>> the new management interface. Should the kernel do it "all" through a
>>> single request from userspace or should userspace control the separate
>>> BR/EDR and LE discoveries. If it's the former, then having a single
>>> adapter_ops callback makes sense. If it's the later then we should
>>> probably have two separate ones. Either way I now realize that to
>>> implement this with a single callback we'd need to have the HCI event
>>> processing moved over to hciops and that's not something that will
>>> happen very quickly. So maybe it's better to have these as two separate
>>> adapter_ops callbacks for now.
>>>
>>> Johan
>>>
>>
>> Another point is: scanning needs to be disabled "actively", it is not
>> possible to inform a scanning length like inquiry, it is necessary to
>> create a timeout to disable the scanning. Joining inquiry and scanning
>> inside a single adapter_ops callback will be hard to control the
>> discovery procedure "states". So for now, I prefer two separate
>> callbacks due the scanning timeout and HCI events processing.
>>
>> However, for the new management interface, one single request seems to
>> be more appropriate but extra requests will be necessary to setup the
>> scan and inquiry parameters.
>>
>> Claudio
>>
>
> New pull request! Patches rebased and a new hciops callback function
> has been added to export the write LE host supported command.
>
> The following changes since commit 3a8386d428e385d97b95a890623362321a7e3a19:
>
>  Fix left open socket warning for attribute server (2010-09-30 17:26:55 -0300)
>
> are available in the git repository at:
>  git://git.infradead.org/users/cktakahasi/bluez.git for-upstream
>
> Claudio Takahasi (23):
>      Add LE start and stop scanning
>      Remove RSSI field from the advertising report event
>      Decoding the RSSI parameter from the advertising report event
>      Send Discovering property "FALSE" when the interleave finishes
>      Add length argument on hciops start discovery function
>      Stop inquiry using the length parameter
>      Fix remote name resolution for interleave discovery
>      Add Write LE host supported function
>      Set the LE host supported and disable simultaneous LE and BR/EDR flags
>      Add extended feature mask constants definition
>      Read the local extended features
>      Stop LE scanning when discovery is suspended
>      Rename hciops {start, stop}_discovery to {start, stop}_inquiry
>      Don't enter on interleave mode if there isn't active sessions
>      Code cleanup: improving inquiry logic
>      Clear the remote device found list in the state transition
>      Fix periodic inquiry signals
>      Fixing DeviceDisappeared signal
>      Postpone discovery if still resolving names
>      Add adapter discovery type function
>      Do not send another Discovering TRUE signal if still resolving names
>      Forcing periodic inquiry exit
>      Fix interleave scanning
>
> Vinicius Costa Gomes (1):
>      Add BR/EDR LE interleaved discovery
>
>  lib/hci.c        |   29 +++++
>  lib/hci.h        |   17 +++-
>  lib/hci_lib.h    |    1 +
>  plugins/hciops.c |   69 +++++++++++-
>  src/adapter.c    |  301 +++++++++++++++++++++++++++++++++++++-----------------
>  src/adapter.h    |   38 ++++---
>  src/dbus-hci.c   |   45 +++++++--
>  src/dbus-hci.h   |    1 +
>  src/security.c   |  131 ++++++++++++++----------
>  9 files changed, 449 insertions(+), 183 deletions(-)
>


The following changes since commit 19c8e310329bb6b369f11501afec9ff4223c2971:

  Allow errors to propagate from connection callback (2010-10-04 17:06:49 +0200)

are available in the git repository at:
  git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

Claudio Takahasi (23):
      Add LE start and stop scanning
      Remove RSSI field from the advertising report event
      Decoding the RSSI parameter from the advertising report event
      Send Discovering property "FALSE" when the interleave finishes
      Add length argument on hciops start discovery function
      Stop inquiry using the length parameter
      Fix remote name resolution for interleave discovery
      Add Write LE host supported function
      Set the LE host supported and disable simultaneous LE and BR/EDR flags
      Add extended feature mask constants definition
      Read the local extended features
      Stop LE scanning when discovery is suspended
      Rename hciops {start, stop}_discovery to {start, stop}_inquiry
      Don't enter on interleave mode if there isn't active sessions
      Code cleanup: improving inquiry logic
      Clear the remote device found list in the state transition
      Fix periodic inquiry signals
      Fixing DeviceDisappeared signal
      Postpone discovery if still resolving names
      Add adapter discovery type function
      Do not send another Discovering TRUE signal if still resolving names
      Forcing periodic inquiry exit
      Fix interleave scanning

Vinicius Costa Gomes (1):
      Add BR/EDR LE interleaved discovery

 lib/hci.c        |   29 +++++
 lib/hci.h        |   17 +++-
 lib/hci_lib.h    |    1 +
 plugins/hciops.c |   69 +++++++++++-
 src/adapter.c    |  301 +++++++++++++++++++++++++++++++++++++-----------------
 src/adapter.h    |   38 ++++---
 src/dbus-hci.c   |   45 +++++++--
 src/dbus-hci.h   |    1 +
 src/security.c   |  131 ++++++++++++++----------
 9 files changed, 449 insertions(+), 183 deletions(-)

Re: Pull request: git://git.infradead.org/users/cktakahasi/bluez.git for-upstream

[Johan Hedberg]

Hi Claudio,

On Mon, Oct 04, 2010, Claudio Takahasi wrote:
> The following changes since commit 19c8e310329bb6b369f11501afec9ff4223c2971:
> 
>   Allow errors to propagate from connection callback (2010-10-04 17:06:49 +0200)
> 
> are available in the git repository at:
>   git://git.infradead.org/users/cktakahasi/bluez.git for-upstream
> 
> Claudio Takahasi (23):
>       Add LE start and stop scanning
>       Remove RSSI field from the advertising report event
>       Decoding the RSSI parameter from the advertising report event
>       Send Discovering property "FALSE" when the interleave finishes
>       Add length argument on hciops start discovery function
>       Stop inquiry using the length parameter
>       Fix remote name resolution for interleave discovery
>       Add Write LE host supported function
>       Set the LE host supported and disable simultaneous LE and BR/EDR flags
>       Add extended feature mask constants definition
>       Read the local extended features
>       Stop LE scanning when discovery is suspended
>       Rename hciops {start, stop}_discovery to {start, stop}_inquiry
>       Don't enter on interleave mode if there isn't active sessions
>       Code cleanup: improving inquiry logic
>       Clear the remote device found list in the state transition
>       Fix periodic inquiry signals
>       Fixing DeviceDisappeared signal
>       Postpone discovery if still resolving names
>       Add adapter discovery type function
>       Do not send another Discovering TRUE signal if still resolving names
>       Forcing periodic inquiry exit
>       Fix interleave scanning
> 
> Vinicius Costa Gomes (1):
>       Add BR/EDR LE interleaved discovery
> 
>  lib/hci.c        |   29 +++++
>  lib/hci.h        |   17 +++-
>  lib/hci_lib.h    |    1 +
>  plugins/hciops.c |   69 +++++++++++-
>  src/adapter.c    |  301 +++++++++++++++++++++++++++++++++++++-----------------
>  src/adapter.h    |   38 ++++---
>  src/dbus-hci.c   |   45 +++++++--
>  src/dbus-hci.h   |    1 +
>  src/security.c   |  131 ++++++++++++++----------
>  9 files changed, 449 insertions(+), 183 deletions(-)

Thanks. Pushed upstream.

Johan