Implementing the PIN helper support

we need to think about implementing the variable PIN helper support within the D-Bus API. What is the best way to implement such kind of communication?

Quoted 3 lines

we need to think about implementing the variable PIN helper support within the D-Bus API. What is the best way to implement such kind of communication?

I'll first describe a couple of typical usage scenarios when this feature would be needed so others not involved in previous discussions know what we are talking about: Currently the D-BUS pin helper works so that when enabled hcid will call a PinRequest D-BUS method at a hardcoded service and object path (org.bluez.PinAgent and /org/bluez/PinAgent).

This works fine as long as there is only need for one generic pin handler, e.g. a popup dialog. However, a piece of SW might want to implement e.g. a "pairing wizard" and take over handling of pin requests for a certain remote device for a short period of time. A SIM Access Profile implementation might also want to override the default handler with a GUI that automatically generates a 16 character PIN (which SIM Access Profile requires).

A (somewhat simple) way to allow this kind of flexibility in PIN requests could be e.g. to have two methods for registering PIN handlers:

1. For registering the "default" handle you could have RegisterDefault(string object_path)
This would tell hcid (or bluetoothd) that the "object" at the path object_path implements the org.bluez.PinAgent
interface and is ready to handle PIN requests. I havent listed the service (or dbus name as it is called nowdays) in the arguments list since that information could be picked directly from the method call (using dbus_message_get_sender()). If we want to allow a service to register a handler on behalf of another service, then this extra parameter can of course be added.

2. For registering a "special" pin handler you could have e.g. RegisterSpecial(string object_path, string bt_addr)
Same applies as for RegisterDefault, except that only PIN requests for the device identified by bt_addr would be delegated to this handler. A "unregister" method for this type of handler might not be necessary as we can have e.g. a timer for removing it automatically if nothing hapens for a while. The handler would naturally also be removed once pairing has succeeded with the device that the handler was registered for or if the service which registered the handler disappears from the bus.

my current idea is to have a default handler that the applications have to register over org.bluez.Manager and that they can have multiple other PIN handlers registered via the org.bluez.Device interface for specific devices. While the default handler is permanent, all other PIN handlers may expire and get removed automatically.

Do we really only need the object path? I really like to get the default handler thing implemented very soon.

Quoted 5 lines

my current idea is to have a default handler that the applications have to register over org.bluez.Manager and that they can have multiple other PIN handlers registered via the org.bluez.Device interface for specific devices. While the default handler is permanent, all other PIN handlers may expire and get removed automatically.

I think you'll still need a way to unregister the default handler, when e.g. a user logs out from his session (so the next user that logs in can register his own default handler).

Using org.bluez.Manager interface for registering the default handler seems to me extending too much its purpose compared to hald where we copied it from. The only purpose of the manager interface there is to allow local device discovery. Placing the default handler registration method at the interface org.bluez.Device (or org.bluez.Device.PinAgent) and object path /org/bluez/Device should be enough in my opinion. If you want to register a handler only for a specific local device then the exact path for that device (e.g. /org/bluez/Device/hci0) could be used.

Quoted 2 lines

Do we really only need the object path? I really like to get the default handler thing implemented very soon.

For the handler registration the object path should be enough. However we should probably rethink the org.bluez.PinAgent interface which currently consists of only one method: PinRequest.

The PinRequest method currently takes two arguments: remote device address and a boolean indicating whether the connection is incomming or outgoing. A third parameter should probably be added for object path of the local device, since that information could be used for e.g. querying the name of the remote device.

Quoted 8 lines

Using org.bluez.Manager interface for registering the default handler seems to me extending too much its purpose compared to hald where we copied it from. The only purpose of the manager interface there is to allow local device discovery. Placing the default handler registration method at the interface org.bluez.Device (or org.bluez.Device.PinAgent) and object path /org/bluez/Device should be enough in my opinion. If you want to register a handler only for a specific local device then the exact path for that device (e.g. /org/bluez/Device/hci0) could be used.

For the default handler the Manager interface makes perfect sense to me. It has nothing to do with any device and it can be present even if no Bluetooth device is attached at all. From me perspective we put every non device specific things into the Manager interface. However I might be missing something.

Quoted 7 lines

Quoted 2 lines

Do we really only need the object path? I really like to get the default handler thing implemented very soon.

For the handler registration the object path should be enough. However we should probably rethink the org.bluez.PinAgent interface which currently consists of only one method: PinRequest.

Quoted 5 lines

The PinRequest method currently takes two arguments: remote device address and a boolean indicating whether the connection is incomming or outgoing. A third parameter should probably be added for object path of the local device, since that information could be used for e.g. querying the name of the remote device.

The interface must change. We use strings for the BD_ADDR now a clean break is a good thing. Can you propose something?

Quoted 9 lines

Quoted 8 lines

Using org.bluez.Manager interface for registering the default handler seems to me extending too much its purpose compared to hald where we copied it from. The only purpose of the manager interface there is to allow local device discovery. Placing the default handler registration method at the interface org.bluez.Device (or org.bluez.Device.PinAgent) and object path /org/bluez/Device should be enough in my opinion. If you want to register a handler only for a specific local device then the exact path for that device (e.g. /org/bluez/Device/hci0) could be used.

Quoted 5 lines

For the default handler the Manager interface makes perfect sense to me. It has nothing to do with any device and it can be present even if no Bluetooth device is attached at all. From me perspective we put every non device specific things into the Manager interface. However I might be missing something.

Well, it depends how we define the purpose of the Manager interface :) If we say it's for "all stuff which is not device specific" then what you propose is ok. If we say it's only for local device discovery, as in hald, then the default handler registration shouldn't be in Manager.

In my opinion non-device specific things could just use the root device path (/org/bluez/Device), while device specific stuff would use the exact device paths (e.g. /org/bluez/Device/hci0). But, I don't have any really strong opinion about this, so if you still insist on using Manager I wont complain anymore ;)

Quoted 7 lines

Quoted 7 lines

Quoted 2 lines

Do we really only need the object path? I really like to get the default handler thing implemented very soon.

For the handler registration the object path should be enough. However we should probably rethink the org.bluez.PinAgent interface which currently consists of only one method: PinRequest.

Quoted 8 lines

Quoted 5 lines

The PinRequest method currently takes two arguments: remote device address and a boolean indicating whether the connection is incomming or outgoing. A third parameter should probably be added for object path of the local device, since that information could be used for e.g. querying the name of the remote device.

The interface must change. We use strings for the BD_ADDR now a clean break is a good thing. Can you propose something?

I'm not so convinced about the necessity of the "outgoing" boolean parameter, so depending on whether you want to keep it or not (we could add another device method for getting this information later) I'd propose one of the following:
PinRequest(string bt_addr, string device_path) PinRequest(string bt_addr, string device_path, boolean outgoing)

So a call could look like:
PinRequest("11:22:33:44:55:66", "/org/bluez/Device/hci0")

Hi Johan,

Quoted 1 line

Quoted 1 line

my current idea is to have a default handler that the applications ha=

Quoted 1 line

Quoted 1 line

to register over org.bluez.Manager and that they can have multiple ot=

Quoted 1 line

Quoted 1 line

PIN handlers registered via the org.bluez.Device interface for specif=

Quoted 1 line

Quoted 1 line

devices. While the default handler is permanent, all other PIN handle=

Quoted 4 lines

Quoted 1 line

may expire and get removed automatically.

I think you'll still need a way to unregister the default handler, when e.g. a user logs out from his session (so the next user that logs in ca=

Quoted 1 line

register his own default handler).

I agree. In the first place I thought we can automatically unregister it when the program ends, but in case of desktop switching we need the Bluetooth applet playing nice.

Quoted 6 lines

Using org.bluez.Manager interface for registering the default handler seems to me extending too much its purpose compared to hald where we copied it from. The only purpose of the manager interface there is to allow local device discovery. Placing the default handler registration method at the interface org.bluez.Device (or org.bluez.Device.PinAgent) and object path /org/bluez/Device should be enough in my opinion. If yo=

Quoted 2 lines

want to register a handler only for a specific local device then the exact path for that device (e.g. /org/bluez/Device/hci0) could be used.

For the default handler the Manager interface makes perfect sense to me. It has nothing to do with any device and it can be present even if no Bluetooth device is attached at all. From me perspective we put every non device specific things into the Manager interface. However I might be missing something.

Quoted 1 line

Quoted 1 line

Do we really only need the object path? I really like to get the defa=

Quoted 10 lines

Quoted 1 line

handler thing implemented very soon.

For the handler registration the object path should be enough. However we should probably rethink the org.bluez.PinAgent interface which currently consists of only one method: PinRequest.

The PinRequest method currently takes two arguments: remote device address and a boolean indicating whether the connection is incomming or outgoing. A third parameter should probably be added for object path of the local device, since that information could be used for e.g. queryin=

Quoted 1 line

the name of the remote device.

The interface must change. We use strings for the BD_ADDR now a clean break is a good thing. Can you propose something?

Regards

Marcel

-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log fi=

for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D103432&bid=3D230486&dat=

_______________________________________________
Bluez-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bluez-devel

Let me try expose some D-Bus features that can help us design the PIN Agent better...

Currently PinAgent uses the system bus connection. D-Bus is able to start applications automatically. Applications with user interfaces can't be started by system dbus-daemon, only the session can run this kind of applications. Using the session bus connection, it is not necessary has a applet/or a application always active. The D-Bus starts the application when the first message is sent to the path. For more information read the section "Message Bus Starting Services":
http://dbus.freedesktop.org/doc/dbus-specification.html

Quoted 6 lines

Quoted 6 lines

Quoted 5 lines

my current idea is to have a default handler that the applications have to register over org.bluez.Manager and that they can have multiple other PIN handlers registered via the org.bluez.Device interface for specific devices. While the default handler is permanent, all other PIN handlers may expire and get removed automatically.

Quoted 3 lines

Quoted 3 lines

I think you'll still need a way to unregister the default handler, when e.g. a user logs out from his session (so the next user that logs in can register his own default handler).

Remember that we can register fallback path(hierarchical path), where the root level can handle messages not handled by the child paths. Therefore, the default PIN Agent can be the base and the user defined can child paths.

Let me try expose some D-Bus features that can help us design the PIN Agent better...

Currently PinAgent uses the system bus connection. D-Bus is able to start applications automatically. Applications with user interfaces can't be started by system dbus-daemon, only the session can run this kind of applications. Using the session bus connection, it is not necessary has a applet/or a application always active. The D-Bus starts the application when the first message is sent to the path. For more information read the section "Message Bus Starting Services":
http://dbus.freedesktop.org/doc/dbus-specification.html

Quoted 11 lines

Let me try expose some D-Bus features that can help us design the PIN Agent better...

Currently PinAgent uses the system bus connection. D-Bus is able to start applications automatically. Applications with user interfaces can't be started by system dbus-daemon, only the session can run this kind of applications. Using the session bus connection, it is not necessary has a applet/or a application always active. The D-Bus starts the application when the first message is sent to the path. For more information read the section "Message Bus Starting Services":
http://dbus.freedesktop.org/doc/dbus-specification.html

I'm aware of the activation feature of D-BUS, and if you look through the bluez-devel archives you'll even find mails from me suggesting to use it a year ago or so. However, it's most likely not the right solution nor even feasible currently.

The main problem is that we can't use the session bus for hcid/bluetoothd since the daemon is a system level service which doesn't know about any specific session (it might become possible in the future to discover existing session busses through the system bus, but currently this is not possible).

Then there's the possibility of creating a bluetoothd specific interface for the user to tell it about the existence of his session bus. However, I have a feeling that this would over-complicate the D-BUS code, create all sorts of extra (e.g. security related) issues, and would conflict with the "discover session busses through system bus" feature which has been hinted as a possible addition to post-1.0 versions of dbus (you'll need to dig the dbus mailing list archives for the exact reference).