Skip to contentSkip to conversation

android/pan: Fix wrong struct parameter in disconnect function call

PATCH In review Ravi kumar Veeramally 2 participants 14 messages Nov 19, 2013 Bluetooth drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/5 android/pan: Fix wrong struct parameter in disconnect function call

---
 android/pan.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
android/pan.c +2 -1
index 2a11f46..fba86b8 100644
@@ -58,7 +58,8 @@ static uint8_t bt_pan_connect(struct hal_cmd_pan_connect *cmd, uint16_t len)
 	return HAL_STATUS_FAILED;
 }
 
-static uint8_t bt_pan_disconnect(struct hal_cmd_pan_connect *cmd, uint16_t len)
+static uint8_t bt_pan_disconnect(struct hal_cmd_pan_disconnect *cmd,
+								uint16_t len)
 {
 	DBG("Not Implemented");
 
Signature · 1 line
1.8.3.2
#1 lore raw

PATCH 2/5 android/hal-pan: Return error in case of unsupported PAN roles

---
 android/hal-pan.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
android/hal-pan.c +11 -0
index 2bc560e..30facd4 100644
@@ -77,6 +77,9 @@ static bt_status_t pan_enable(int local_role)
 	if (!interface_ready())
 		return BT_STATUS_NOT_READY;
 
+	if (!(local_role == BTPAN_ROLE_PANU || local_role == BTPAN_ROLE_PANNAP))
+		return BT_STATUS_UNSUPPORTED;
+
 	cmd.local_role = local_role;
 
 	return hal_ipc_cmd(HAL_SERVICE_ID_PAN, HAL_OP_PAN_ENABLE,
@@ -112,6 +115,14 @@ static bt_status_t pan_connect(const bt_bdaddr_t *bd_addr, int local_role,
 	if (!interface_ready())
 		return BT_STATUS_NOT_READY;
 
+	if (!((local_role == BTPAN_ROLE_PANNAP &&
+			remote_role == BTPAN_ROLE_PANU) ||
+		(local_role == BTPAN_ROLE_PANU &&
+			remote_role == BTPAN_ROLE_PANNAP) ||
+		(local_role == BTPAN_ROLE_PANU &&
+			remote_role == BTPAN_ROLE_PANU)))
+		return BT_STATUS_UNSUPPORTED;
+
 	memcpy(cmd.bdaddr, bd_addr, sizeof(cmd.bdaddr));
 	cmd.local_role = local_role;
 	cmd.remote_role = remote_role;
Signature · 1 line
1.8.3.2
#2 lore raw

Re: [PATCH 2/5] android/hal-pan: Return error in case of unsupported PAN roles

Hi Ravi,

Quoted 19 lines · Ravi kumar Veeramally
---
 android/hal-pan.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
android/hal-pan.c +3 -0
index 2bc560e..30facd4 100644
@@ -77,6 +77,9 @@ static bt_status_t pan_enable(int local_role)
 	if (!interface_ready())
 		return BT_STATUS_NOT_READY;
 
+	if (!(local_role == BTPAN_ROLE_PANU || local_role == BTPAN_ROLE_PANNAP))
+		return BT_STATUS_UNSUPPORTED;
+
 	cmd.local_role = local_role;
 
 	return hal_ipc_cmd(HAL_SERVICE_ID_PAN, HAL_OP_PAN_ENABLE,

android/hal-pan.c static bt_status_t pan_connect(const bt_bdaddr_t *bd_addr, int local_role, · line 124

Quoted 10 lines
android/hal-pan.c +7 -0
@@ -112,6 +115,14 @@ static bt_status_t pan_connect(const bt_bdaddr_t *bd_addr, int local_role,
 	if (!interface_ready())
 		return BT_STATUS_NOT_READY;
 
+	if (!((local_role == BTPAN_ROLE_PANNAP &&
+			remote_role == BTPAN_ROLE_PANU) ||
+		(local_role == BTPAN_ROLE_PANU &&
+			remote_role == BTPAN_ROLE_PANNAP) ||
+		(local_role == BTPAN_ROLE_PANU &&
+			remote_role == BTPAN_ROLE_PANU)))
+		return BT_STATUS_UNSUPPORTED;

First of all you've got incorrect indentation here which make this hard to read (the return statement being on the same column as parts of the if-statement. When you break lines indent at least by two tabs to make a clear separation from the actual branch code.

Secondly, shouldn't we be checking that the given local role has been enabled (through pan_enable)? Or does the daemon do that checking? In fact is there a clear reason not to let the daemon do all the verification checks and return an error status over IPC?

Thirdly, this if-statement takes a while to parse, so I'm wondering whether it'd be clearer to format it in a bit more readable way, e.g.:

        switch (local_role) {
        case BTPAN_ROLE_PANNAP:
                if (remote_role != BTPAN_ROLE_PANU)
                        return BT_STATUS_UNSUPPORTED;
                break;
        case BTPAN_ROLE_PANU:
                if (remote_role != BTPAN_ROLE_PANNAP &&
                                                remote_role != BTPAN_ROLE_PANU)
                        return BT_STATUS_UNSUPPORTED;
                break;
        default:
                return BT_STATUS_UNSUPPORTED;
        }

Thoughts?

Johan

#3 lore raw

Re: [PATCH 2/5] android/hal-pan: Return error in case of unsupported PAN roles

Quoted 25 lines · Johan Hedberg

Hi Ravi,

Quoted 22 lines · Ravi kumar Veeramally
---
  android/hal-pan.c | 11 +++++++++++
  1 file changed, 11 insertions(+)
android/hal-pan.c +3 -0
index 2bc560e..30facd4 100644
@@ -77,6 +77,9 @@ static bt_status_t pan_enable(int local_role)
  	if (!interface_ready())
  		return BT_STATUS_NOT_READY;
 
+	if (!(local_role == BTPAN_ROLE_PANU || local_role == BTPAN_ROLE_PANNAP))
+		return BT_STATUS_UNSUPPORTED;
+
  	cmd.local_role = local_role;
 
  	return hal_ipc_cmd(HAL_SERVICE_ID_PAN, HAL_OP_PAN_ENABLE,
@@ -112,6 +115,14 @@ static bt_status_t pan_connect(const bt_bdaddr_t *bd_addr, int local_role,
  	if (!interface_ready())
  		return BT_STATUS_NOT_READY;
 
Quoted 11 lines

android/hal-pan.c static bt_status_t pan_connect(const bt_bdaddr_t *bd_addr, int local_role, · line 124

Quoted 7 lines
android/hal-pan.c +7 -0
+	if (!((local_role == BTPAN_ROLE_PANNAP &&
+			remote_role == BTPAN_ROLE_PANU) ||
+		(local_role == BTPAN_ROLE_PANU &&
+			remote_role == BTPAN_ROLE_PANNAP) ||
+		(local_role == BTPAN_ROLE_PANU &&
+			remote_role == BTPAN_ROLE_PANU)))
+		return BT_STATUS_UNSUPPORTED;

First of all you've got incorrect indentation here which make this hard to read (the return statement being on the same column as parts of the if-statement. When you break lines indent at least by two tabs to make a clear separation from the actual branch code.

Ok.

Quoted 5 lines

Secondly, shouldn't we be checking that the given local role has been enabled (through pan_enable)? Or does the daemon do that checking? In fact is there a clear reason not to let the daemon do all the verification checks and return an error status over IPC?

I thought it would be easier to do basic checks on supported combinations
at HAL level to reduce the ipc traffic.
I don't know exactly whether UI can send these combinations, but we
can at least
try with haltest tool.

Quoted 18 lines

Thirdly, this if-statement takes a while to parse, so I'm wondering whether it'd be clearer to format it in a bit more readable way, e.g.:

         switch (local_role) {
         case BTPAN_ROLE_PANNAP:
                 if (remote_role != BTPAN_ROLE_PANU)
                         return BT_STATUS_UNSUPPORTED;
                 break;
         case BTPAN_ROLE_PANU:
                 if (remote_role != BTPAN_ROLE_PANNAP &&
                                                 remote_role != BTPAN_ROLE_PANU)
                         return BT_STATUS_UNSUPPORTED;
                 break;
         default:
                 return BT_STATUS_UNSUPPORTED;
         }
Quoted 1 line

Thoughts?

Yes, make sense to use 'switch' for more readability. Is it ok to send _v2 with switch or better to handle in daemon?

Thanks,
Ravi.

#4 lore raw

PATCH 3/5 android: Handle multiple init(register) and cleanup(unregister) calls properly

In reply to Ravi kumar Veeramally

This can be tested with haltest.

---
 android/a2dp.c      | 6 ++++++
 android/bluetooth.c | 6 ++++++
 android/hidhost.c   | 6 ++++++
 android/pan.c       | 6 ++++++
 4 files changed, 24 insertions(+)
android/a2dp.c +6 -0
index 74d0082..28a1250 100644
@@ -332,6 +332,9 @@ bool bt_a2dp_register(int sk, const bdaddr_t *addr)
 
 	DBG("");
 
+	if (notification_sk > 0)
+		return false;
+
 	bacpy(&adapter_addr, addr);
 
 	server = bt_io_listen(connect_cb, NULL, NULL, NULL, &err,
@@ -365,6 +368,9 @@ void bt_a2dp_unregister(void)
 {
 	DBG("");
 
+	if (notification_sk < 0)
+		return;
+
 	notification_sk = -1;
 
 	bt_adapter_remove_record(record_id);
android/bluetooth.c +6 -0
index 7dc2ec3..7805d42 100644
@@ -2275,6 +2275,9 @@ bool bt_bluetooth_register(int sk)
 {
 	DBG("");
 
+	if (notification_sk > 0)
+		return false;
+
 	notification_sk = sk;
 
 	return true;
@@ -2284,5 +2287,8 @@ void bt_bluetooth_unregister(void)
 {
 	DBG("");
 
+	if (notification_sk < 0)
+		return;
+
 	notification_sk = -1;
 }
android/hidhost.c +6 -0
index 842b8ad..a929a94 100644
@@ -1190,6 +1190,9 @@ bool bt_hid_register(int sk, const bdaddr_t *addr)
 
 	DBG("");
 
+	if (notification_sk > 0)
+		return false;
+
 	bacpy(&adapter_addr, addr);
 
 	ctrl_io = bt_io_listen(connect_cb, NULL, NULL, NULL, &err,
@@ -1224,6 +1227,9 @@ void bt_hid_unregister(void)
 {
 	DBG("");
 
+	if (notification_sk < 0)
+		return;
+
 	notification_sk = -1;
 
 	if (ctrl_io) {
android/pan.c +6 -0
index fba86b8..970a8a7 100644
@@ -95,6 +95,9 @@ bool bt_pan_register(int sk, const bdaddr_t *addr)
 {
 	DBG("");
 
+	if (notification_sk > 0)
+		return false;
+
 	notification_sk = sk;
 
 	return true;
@@ -104,5 +107,8 @@ void bt_pan_unregister(void)
 {
 	DBG("");
 
+	if (notification_sk < 0)
+		return;
+
 	notification_sk = -1;
 }
Signature · 1 line
1.8.3.2
#5 lore raw

Re: [PATCH 3/5] android: Handle multiple init(register) and cleanup(unregister) calls properly

Hi Ravi,

android/bluetooth.c bool bt_bluetooth_register(int sk) · line 2278

Quoted 5 lines · Ravi kumar Veeramally
android/bluetooth.c +1 -0
@@ -2275,6 +2275,9 @@ bool bt_bluetooth_register(int sk)
 {
 	DBG("");
 
+	if (notification_sk > 0)

0 is a valid file descriptor value so the check should be >= 0

android/a2dp.c bool bt_hid_register(int sk, const bdaddr_t *addr) · line 1194

Quoted 6 lines
android/a2dp.c +2 -0
@@ -1190,6 +1190,9 @@ bool bt_hid_register(int sk, const bdaddr_t *addr)
 
 	DBG("");
 
+	if (notification_sk > 0)
+		return false;

Same here.

android/bluetooth.c bool bt_pan_register(int sk, const bdaddr_t *addr) · line 99

Quoted 6 lines
android/bluetooth.c +2 -0
@@ -95,6 +95,9 @@ bool bt_pan_register(int sk, const bdaddr_t *addr)
 {
 	DBG("");
 
+	if (notification_sk > 0)
+		return false;

And here.

Johan

#6 lore raw

Re: [PATCH 3/5] android: Handle multiple init(register) and cleanup(unregister) calls properly

Hi,

Quoted 22 lines · Johan Hedberg

Hi Ravi,

android/bluetooth.c bool bt_bluetooth_register(int sk) · line 2278

Quoted 5 lines · Ravi kumar Veeramally
android/bluetooth.c +1 -0
@@ -2275,6 +2275,9 @@ bool bt_bluetooth_register(int sk)
  {
  	DBG("");
 
+	if (notification_sk > 0)

0 is a valid file descriptor value so the check should be >= 0

android/a2dp.c bool bt_hid_register(int sk, const bdaddr_t *addr) · line 1194

Quoted 6 lines
android/a2dp.c +2 -0
@@ -1190,6 +1190,9 @@ bool bt_hid_register(int sk, const bdaddr_t *addr)
 
  	DBG("");
 
+	if (notification_sk > 0)
+		return false;

Same here.

Quoted 4 lines
android/a2dp.c +0 -0
@@ -95,6 +95,9 @@ bool bt_pan_register(int sk, const bdaddr_t *addr)
  {
  	DBG("");
 
Quoted 6 lines

Review bool bt_pan_register(int sk, const bdaddr_t *addr) · line 99

Quoted 2 lines
(quoted diff) +2 -0
+	if (notification_sk > 0)
+		return false;

And here.

Johan

Ok.

Thanks,
Ravi.

#7 lore raw

PATCH 4/5 android/hidhost: Handle error case properly in interrupt_connect_cb

In reply to Ravi kumar Veeramally
---
 android/hidhost.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
android/hidhost.c +6 -3
index a929a94..05a3fe4 100644
@@ -522,7 +522,6 @@ static int uhid_create(struct hid_device *dev)
 	dev->uhid_fd = open(UHID_DEVICE_FILE, O_RDWR | O_CLOEXEC);
 	if (dev->uhid_fd < 0) {
 		error("Failed to open uHID device: %s", strerror(errno));
-		bt_hid_notify_state(dev, HAL_HIDHOST_STATE_NO_HID);
 		return -errno;
 	}
 
@@ -541,7 +540,6 @@ static int uhid_create(struct hid_device *dev)
 		error("Failed to create uHID device: %s", strerror(errno));
 		close(dev->uhid_fd);
 		dev->uhid_fd = -1;
-		bt_hid_notify_state(dev, HAL_HIDHOST_STATE_NO_HID);
 		return -errno;
 	}
 
@@ -559,16 +557,20 @@ static void interrupt_connect_cb(GIOChannel *chan, GError *conn_err,
 							gpointer user_data)
 {
 	struct hid_device *dev = user_data;
+	uint8_t state;
 
 	DBG("");
 
 	if (conn_err) {
 		error("%s", conn_err->message);
+		state = HAL_HIDHOST_STATE_FAILED;
 		goto failed;
 	}
 
-	if (uhid_create(dev) < 0)
+	if (uhid_create(dev) < 0) {
+		state = HAL_HIDHOST_STATE_NO_HID;
 		goto failed;
+	}
 
 	dev->intr_watch = g_io_add_watch(dev->intr_io,
 				G_IO_IN | G_IO_HUP | G_IO_ERR | G_IO_NVAL,
@@ -579,6 +581,7 @@ static void interrupt_connect_cb(GIOChannel *chan, GError *conn_err,
 	return;
 
 failed:
+	bt_hid_notify_state(dev, state);
 	hid_device_free(dev);
 }
 
Signature · 1 line
1.8.3.2
#8 lore raw

Re: [PATCH 4/5] android/hidhost: Handle error case properly in interrupt_connect_cb

Hi Ravi,

Quoted 3 lines · Ravi kumar Veeramally
---
 android/hidhost.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

Could you add a bit more descriptive commit message to this one which explains what was broken (and how) and how your patch fixes it. When I get a patch with an empty commit message I expect it to be very trivial and instantly understandable, but that's not the case here.

Johan

#9 lore raw

Re: [PATCH 4/5] android/hidhost: Handle error case properly in interrupt_connect_cb

Hi Johan,

Quoted 10 lines · Johan Hedberg

Hi Ravi,

Quoted 3 lines · Ravi kumar Veeramally
---
  android/hidhost.c | 9 ++++++---
  1 file changed, 6 insertions(+), 3 deletions(-)

Could you add a bit more descriptive commit message to this one which explains what was broken (and how) and how your patch fixes it. When I get a patch with an empty commit message I expect it to be very trivial and instantly understandable, but that's not the case here.

I will write more details.

Regards,
Ravi.

#10 lore raw

PATCH 5/5 android/hidhost: Free all connected devices in profile cleanup call

In reply to Ravi kumar Veeramally
---
 android/hidhost.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
android/hidhost.c +10 -0
index 05a3fe4..20dedc4 100644
@@ -1226,6 +1226,14 @@ bool bt_hid_register(int sk, const bdaddr_t *addr)
 	return true;
 }
 
+static void free_hid_devices(gpointer data, gpointer user_data)
+{
+	struct hid_device *dev = data;
+
+	bt_hid_notify_state(dev, HAL_HIDHOST_STATE_DISCONNECTED);
+	hid_device_free(dev);
+}
+
 void bt_hid_unregister(void)
 {
 	DBG("");
@@ -1233,6 +1241,8 @@ void bt_hid_unregister(void)
 	if (notification_sk < 0)
 		return;
 
+	g_slist_foreach(devices, free_hid_devices, NULL);
+
 	notification_sk = -1;
 
 	if (ctrl_io) {
Signature · 1 line
1.8.3.2
#11 lore raw

Re: [PATCH 5/5] android/hidhost: Free all connected devices in profile cleanup call

Hi Ravi,

Quoted 24 lines · Ravi kumar Veeramally
---
 android/hidhost.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
android/hidhost.c +8 -0
index 05a3fe4..20dedc4 100644
@@ -1226,6 +1226,14 @@ bool bt_hid_register(int sk, const bdaddr_t *addr)
 	return true;
 }
 
+static void free_hid_devices(gpointer data, gpointer user_data)
+{
+	struct hid_device *dev = data;
+
+	bt_hid_notify_state(dev, HAL_HIDHOST_STATE_DISCONNECTED);
+	hid_device_free(dev);
+}
+
 void bt_hid_unregister(void)
 {
 	DBG("");

android/hidhost.c void bt_hid_unregister(void) · line 1244

Quoted 4 lines
android/hidhost.c +1 -0
@@ -1233,6 +1241,8 @@ void bt_hid_unregister(void)
 	if (notification_sk < 0)
 		return;
 
+	g_slist_foreach(devices, free_hid_devices, NULL);

I suppose you also need to set devices = NULL; after this call?

Johan

#12 lore raw

Re: [PATCH 5/5] android/hidhost: Free all connected devices in profile cleanup call

Hi Johan,

Quoted 30 lines · Johan Hedberg

Hi Ravi,

Quoted 27 lines · Ravi kumar Veeramally
---
  android/hidhost.c | 10 ++++++++++
  1 file changed, 10 insertions(+)
android/hidhost.c +8 -0
index 05a3fe4..20dedc4 100644
@@ -1226,6 +1226,14 @@ bool bt_hid_register(int sk, const bdaddr_t *addr)
  	return true;
  }
 
+static void free_hid_devices(gpointer data, gpointer user_data)
+{
+	struct hid_device *dev = data;
+
+	bt_hid_notify_state(dev, HAL_HIDHOST_STATE_DISCONNECTED);
+	hid_device_free(dev);
+}
+
  void bt_hid_unregister(void)
  {
  	DBG("");
@@ -1233,6 +1241,8 @@ void bt_hid_unregister(void)
  	if (notification_sk < 0)
  		return;
 
Quoted 2 lines

android/hidhost.c void bt_hid_unregister(void) · line 1244

Quoted 1 line
android/hidhost.c +1 -0
+	g_slist_foreach(devices, free_hid_devices, NULL);

I suppose you also need to set devices = NULL; after this call?

Sure.

Thanks,
Ravi.