Skip to contentSkip to conversation

Fix managing dbus filters depending on BT state

PATCH In review Bartosz Szatkowski 3 participants 4 messages Nov 22, 2010 Bluetooth drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix managing dbus filters depending on BT state

Previously even if BT was disabled, bluetoothd would handle some dbus singals (mostly connected with HFP). Now filters are added when BT is enabled and removed when BT is disabled.

---
 audio/manager.c          |    4 ++--
 audio/telephony-dummy.c  |   10 ++++++++++
 audio/telephony-maemo5.c |   10 ++++++++++
 audio/telephony-maemo6.c |   10 ++++++++++
 audio/telephony-ofono.c  |   10 ++++++++++
 audio/telephony.h        |    1 +
 src/adapter.c            |    5 +++++
 7 files changed, 48 insertions(+), 2 deletions(-)
audio/manager.c +2 -2
index 816c807..2fc7bf1 100644
@@ -1178,7 +1178,7 @@ proceed:
 		btd_register_adapter_driver(&media_server_driver);
 
 	if (enabled.headset) {
-		telephony_init();
+		telephony_set_state(1);
 		btd_register_adapter_driver(&headset_server_driver);
 	}
 
@@ -1220,7 +1220,7 @@ void audio_manager_exit(void)
 
 	if (enabled.headset) {
 		btd_unregister_adapter_driver(&headset_server_driver);
-		telephony_exit();
+		telephony_set_state(0);
 	}
 
 	if (enabled.gateway)
audio/telephony-dummy.c +10 -0
index 06cb798..ef7c5fa 100644
@@ -57,6 +57,8 @@ static gboolean events_enabled = FALSE;
  */
 static int response_and_hold = -1;
 
+static int telephony_state = 0;
+
 static struct indicator dummy_indicators[] =
 {
 	{ "battchg",	"0-5",	5,	TRUE },
@@ -411,8 +413,16 @@ static GDBusSignalTable dummy_signals[] = {
 	{ }
 };
 
+int telephony_set_state(int state)
+{
+	telephony_state = state;
+}
+
 int telephony_init(void)
 {
+	if (telephony_state == 0)
+		return;
+
 	uint32_t features = AG_FEATURE_REJECT_A_CALL |
 				AG_FEATURE_ENHANCED_CALL_STATUS |
 				AG_FEATURE_EXTENDED_ERROR_RESULT_CODES;
audio/telephony-maemo5.c +10 -0
index 4d0134c..1dd5449 100644
@@ -222,6 +222,8 @@ static char *last_dialed_number = NULL;
 /* Timer for tracking call creation requests */
 static guint create_request_timer = 0;
 
+static int telephony_state = 0;
+
 static struct indicator maemo_indicators[] =
 {
 	{ "battchg",	"0-5",	5,	TRUE },
@@ -2044,8 +2046,16 @@ static DBusHandlerResult signal_filter(DBusConnection *conn,
 	return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
 }
 
+int telephony_set_state(int state)
+{
+	telephony_state = state;
+}
+
 int telephony_init(void)
 {
+	if (telephony_state == 0)
+		return;
+
 	const char *battery_cap = "battery";
 	uint32_t features = AG_FEATURE_EC_ANDOR_NR |
 				AG_FEATURE_INBAND_RINGTONE |
audio/telephony-maemo6.c +10 -0
index 72c8e36..916dcb8 100644
@@ -181,6 +181,8 @@ static char *last_dialed_number = NULL;
 /* Timer for tracking call creation requests */
 static guint create_request_timer = 0;
 
+static int telephony_state = 0;
+
 static struct indicator maemo_indicators[] =
 {
 	{ "battchg",	"0-5",	5,	TRUE },
@@ -1937,8 +1939,16 @@ static DBusHandlerResult signal_filter(DBusConnection *conn,
 	return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
 }
 
+int telephony_set_state(int state)
+{
+	telephony_state = state;
+}
+
 int telephony_init(void)
 {
+	if (telephony_state == 0)
+		return;
+
 	const char *battery_cap = "battery";
 	uint32_t features = AG_FEATURE_EC_ANDOR_NR |
 				AG_FEATURE_INBAND_RINGTONE |
audio/telephony-ofono.c +10 -0
index 693207c..fe3ae19 100644
@@ -96,6 +96,8 @@ static gboolean events_enabled = FALSE;
  */
 static int response_and_hold = -1;
 
+static int telephony_state = 0;
+
 static struct indicator ofono_indicators[] =
 {
 	{ "battchg",	"0-5",	5,	TRUE },
@@ -1062,8 +1064,16 @@ done:
 	dbus_message_unref(reply);
 }
 
+int telephony_set_state(int state)
+{
+	telephony_state = state;
+}
+
 int telephony_init(void)
 {
+	if (telephony_state == 0)
+		return;
+
 	const char *battery_cap = "battery";
 	int ret;
 
audio/telephony.h +1 -0
index 5343e8c..9b80e56 100644
@@ -232,5 +232,6 @@ static inline int telephony_get_indicator(const struct indicator *indicators,
 	return -ENOENT;
 }
 
+int telephony_set_active(int state);
 int telephony_init(void);
 void telephony_exit(void);
src/adapter.c +5 -0
index 6b4a354..8ed2040 100644
@@ -57,6 +57,7 @@
 #include "glib-helper.h"
 #include "agent.h"
 #include "storage.h"
+#include "../audio/telephony.h"
 
 #define IO_CAPABILITY_DISPLAYONLY	0x00
 #define IO_CAPABILITY_DISPLAYYESNO	0x01
@@ -2404,6 +2405,8 @@ int adapter_start(struct btd_adapter *adapter)
 
 	err = adapter_up(adapter, mode);
 
+	telephony_init();
+
 	info("Adapter %s has been enabled", adapter->path);
 
 	return err;
@@ -2533,6 +2536,8 @@ int adapter_stop(struct btd_adapter *adapter)
 
 	call_adapter_powered_callbacks(adapter, FALSE);
 
+	telephony_exit();
+
 	info("Adapter %s has been disabled", adapter->path);
 
 	set_mode_complete(adapter);
Signature · 1 line
1.7.0.4
#1 lore raw

Hi Bartosz,

Quoted 12 lines · Bartosz Szatkowski

Previously even if BT was disabled, bluetoothd would handle some dbus singals (mostly connected with HFP). Now filters are added when BT is enabled and removed when BT is disabled.

---
 audio/manager.c          |    4 ++--
 audio/telephony-dummy.c  |   10 ++++++++++
 audio/telephony-maemo5.c |   10 ++++++++++
 audio/telephony-maemo6.c |   10 ++++++++++
 audio/telephony-ofono.c  |   10 ++++++++++
 audio/telephony.h        |    1 +
 src/adapter.c            |    5 +++++
 7 files changed, 48 insertions(+), 2 deletions(-)

Nack on this one. I understand the issue you're trying to fix but it needs to be done differently.

audio/manager.c proceed: · line 1183

Quoted 12 lines
audio/manager.c +1 -1
index 816c807..2fc7bf1 100644
@@ -1178,7 +1178,7 @@ proceed:
 		btd_register_adapter_driver(&media_server_driver);
 
 	if (enabled.headset) {
-		telephony_init();
+		telephony_set_state(1);
 		btd_register_adapter_driver(&headset_server_driver);
 	}

If you're gonna call this "state" you should have proper defines or enums for the values, however in this case it's essentially a boolean so that's not necessary. In fact since it's a boolean you don't even need to have any new function or variable at all for it: just use telephony_init and telephony_exit.

Quoted 11 lines
src/adapter.c +1 -0
@@ -57,6 +57,7 @@
 #include "glib-helper.h"
 #include "agent.h"
 #include "storage.h"
+#include "../audio/telephony.h"
 
 #define IO_CAPABILITY_DISPLAYONLY	0x00
 #define IO_CAPABILITY_DISPLAYYESNO	0x01

src/adapter.c int adapter_start(struct btd_adapter *adapter) · line 2409

Quoted 5 lines
src/adapter.c +2 -0
@@ -2404,6 +2405,8 @@ int adapter_start(struct btd_adapter *adapter)
 
 	err = adapter_up(adapter, mode);
 
+	telephony_init();
+

This is just wrong. The core daemon should never have direct access to the telephony driver. Instead, you should have the audio plugin (probably audio/manager.c or audio/headset.c) register a adapter powered callback and then call telehony_init/exit from that callback.

Johan

#2 lore raw

Hi,

Quoted 52 lines · Johan Hedberg

Hi Bartosz,

Quoted 11 lines · Bartosz Szatkowski

Previously even if BT was disabled, bluetoothd would handle some dbus singals (mostly connected with HFP). Now filters are added when BT is enabled and removed when BT is disabled.
---
 audio/manager.c          |    4 ++-- audio/telephony-dummy.c  |   10 ++++++++++ audio/telephony-maemo5.c |   10 ++++++++++ audio/telephony-maemo6.c |   10 ++++++++++ audio/telephony-ofono.c  |   10 ++++++++++ audio/telephony.h        |    1 +
 src/adapter.c            |    5 +++++

Quoted 1 line
 7 files changed, 48 insertions(+), 2 deletions(-)

Nack on this one. I understand the issue you're trying to fix but it needs to be done differently.

audio/manager.c proceed: · line 1183

Quoted 12 lines
audio/manager.c +1 -1
index 816c807..2fc7bf1 100644
@@ -1178,7 +1178,7 @@ proceed:
               btd_register_adapter_driver(&media_server_driver);
 
       if (enabled.headset) {
-             telephony_init();
+             telephony_set_state(1);
               btd_register_adapter_driver(&headset_server_driver);
       }

If you're gonna call this "state" you should have proper defines or enums for the values, however in this case it's essentially a boolean so that's not necessary. In fact since it's a boolean you don't even need to have any new function or variable at all for it: just use telephony_init and telephony_exit.

Quoted 14 lines
src/adapter.c +1 -0
@@ -57,6 +57,7 @@
  #include "glib-helper.h"
  #include "agent.h"
  #include "storage.h"
+#include "../audio/telephony.h"
 
  #define IO_CAPABILITY_DISPLAYONLY    0x00
  #define IO_CAPABILITY_DISPLAYYESNO   0x01
@@ -2404,6 +2405,8 @@ int adapter_start(struct btd_adapter *adapter)
 
       err = adapter_up(adapter, mode);
 
Quoted 7 lines

src/adapter.c int adapter_start(struct btd_adapter *adapter) · line 2409

Quoted 2 lines
src/adapter.c +2 -0
+     telephony_init();
+

This is just wrong. The core daemon should never have direct access to the telephony driver. Instead, you should have the audio plugin (probably audio/manager.c or audio/headset.c) register a adapter powered callback and then call telehony_init/exit from that callback.

I have an almost working version of this using adapter drivers and powered changes via callback registration, the tricky part here is how to detect when to do headset_init/headset_exit since those should be called only once. I didn't know there somebody working on this but anyway I gonna try to finish this asap.

Signature · 2 lines
Luiz Augusto von Dentz
Computer Engineer
#3 lore raw
Quoted 69 lines · Luiz Augusto von Dentz

Hi,

Quoted 52 lines · Johan Hedberg

Hi Bartosz,

Quoted 11 lines · Bartosz Szatkowski

Previously even if BT was disabled, bluetoothd would handle some dbus singals (mostly connected with HFP). Now filters are added when BT is enabled and removed when BT is disabled.
---
 audio/manager.c          |    4 ++-- audio/telephony-dummy.c  |   10 ++++++++++ audio/telephony-maemo5.c |   10 ++++++++++ audio/telephony-maemo6.c |   10 ++++++++++ audio/telephony-ofono.c  |   10 ++++++++++ audio/telephony.h        |    1 +
 src/adapter.c            |    5 +++++

Quoted 1 line
 7 files changed, 48 insertions(+), 2 deletions(-)

Nack on this one. I understand the issue you're trying to fix but it needs to be done differently.

audio/manager.c proceed: · line 1183

Quoted 12 lines
audio/manager.c +1 -1
index 816c807..2fc7bf1 100644
@@ -1178,7 +1178,7 @@ proceed:
               btd_register_adapter_driver(&media_server_driver);
 
       if (enabled.headset) {
-             telephony_init();
+             telephony_set_state(1);
               btd_register_adapter_driver(&headset_server_driver);
       }

If you're gonna call this "state" you should have proper defines or enums for the values, however in this case it's essentially a boolean so that's not necessary. In fact since it's a boolean you don't even need to have any new function or variable at all for it: just use telephony_init and telephony_exit.

Quoted 14 lines
src/adapter.c +1 -0
@@ -57,6 +57,7 @@
  #include "glib-helper.h"
  #include "agent.h"
  #include "storage.h"
+#include "../audio/telephony.h"
 
  #define IO_CAPABILITY_DISPLAYONLY    0x00
  #define IO_CAPABILITY_DISPLAYYESNO   0x01
@@ -2404,6 +2405,8 @@ int adapter_start(struct btd_adapter *adapter)
 
       err = adapter_up(adapter, mode);
 
Quoted 7 lines

src/adapter.c int adapter_start(struct btd_adapter *adapter) · line 2409

Quoted 2 lines
src/adapter.c +2 -0
+     telephony_init();
+

This is just wrong. The core daemon should never have direct access to the telephony driver. Instead, you should have the audio plugin (probably audio/manager.c or audio/headset.c) register a adapter powered callback and then call telehony_init/exit from that callback.

I have an almost working version of this using adapter drivers and powered changes via callback registration, the tricky part here is how to detect when to do headset_init/headset_exit since those should be called only once. I didn't know there somebody working on this but anyway I gonna try to finish this asap.

Quoted 4 lines

--
Luiz Augusto von Dentz
Computer Engineer

Hi Luiz,
please fell free to send it (no hard fillings :) ) Ive just getting to know bluez so probably it would take me a while to get through whole plugin/callback subsystem.

Signature · 2 lines
Pozdrowienia,
Bartosz Szatkowski