Skip to contentSkip to conversation

Bluetooth: Fix race condition with rfkill handling

PATCH Superseded [email protected] 2 participants 5 messages Sep 12, 2013

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover Bluetooth: Fix race condition with rfkill handling

Hi,

Here's an updated patch set for the rfkill issue. The only change is the added explicit initialization of the HCI_RFKILLED flag by checking the return value of rfkill_blocked() once rfkill_register() has succeeded.

Johan

----------------------------------------------------------------
Johan Hedberg (2):
Bluetooth: Introduce a new HCI_RFKILLED flag
Bluetooth: Fix rfkill functionality during the HCI setup stage

 include/net/bluetooth/hci.h |  1 +
 net/bluetooth/hci_core.c    | 25 ++++++++++++++++++++-----
 2 files changed, 21 insertions(+), 5 deletions(-)
#1 lore raw

PATCH v2 1/2 Bluetooth: Introduce a new HCI_RFKILLED flag

From: Johan Hedberg <[email protected]>

This makes it more convenient to check for rfkill (no need to check for dev->rfkill before calling rfkill_blocked()) and also avoids potential races if the RFKILL state needs to be checked from within the rfkill callback.

---
v2: Explicitly intialize HCI_RFKILLED flag after calling rfkill_register

 include/net/bluetooth/hci.h |  1 +
 net/bluetooth/hci_core.c    | 15 ++++++++++-----
 2 files changed, 11 insertions(+), 5 deletions(-)
include/net/bluetooth/hci.h +1 -0
index 30c88b5..ba008d5 100644
@@ -104,6 +104,7 @@ enum {
 enum {
 	HCI_SETUP,
 	HCI_AUTO_OFF,
+	HCI_RFKILLED,
 	HCI_MGMT,
 	HCI_PAIRABLE,
 	HCI_SERVICE_CACHE,
net/bluetooth/hci_core.c +10 -5
index 4dbb6cb..d0d6cf8 100644
@@ -1148,7 +1148,7 @@ int hci_dev_open(__u16 dev)
 		goto done;
 	}
 
-	if (hdev->rfkill && rfkill_blocked(hdev->rfkill)) {
+	if (test_bit(HCI_RFKILLED, &hdev->dev_flags)) {
 		ret = -ERFKILL;
 		goto done;
 	}
@@ -1597,10 +1597,12 @@ static int hci_rfkill_set_block(void *data, bool blocked)
 	if (test_bit(HCI_USER_CHANNEL, &hdev->dev_flags))
 		return -EBUSY;
 
-	if (!blocked)
-		return 0;
-
-	hci_dev_do_close(hdev);
+	if (blocked) {
+		set_bit(HCI_RFKILLED, &hdev->dev_flags);
+		hci_dev_do_close(hdev);
+	} else {
+		clear_bit(HCI_RFKILLED, &hdev->dev_flags);
+	}
 
 	return 0;
 }
@@ -2244,6 +2246,9 @@ int hci_register_dev(struct hci_dev *hdev)
 		}
 	}
 
+	if (hdev->rfkill && rfkill_blocked(hdev->rfkill))
+		set_bit(HCI_RFKILLED, &hdev->dev_flags);
+
 	set_bit(HCI_SETUP, &hdev->dev_flags);
 
 	if (hdev->dev_type != HCI_AMP)
Signature · 1 line
1.8.4.rc3
maintainer Acked-by on v2 1/2
#2 lore raw

Re: [PATCH v2 1/2] Bluetooth: Introduce a new HCI_RFKILLED flag

Hi Johan,

Quoted 5 lines

This makes it more convenient to check for rfkill (no need to check for dev->rfkill before calling rfkill_blocked()) and also avoids potential races if the RFKILL state needs to be checked from within the rfkill callback.

Quoted 8 lines

---
v2: Explicitly intialize HCI_RFKILLED flag after calling rfkill_register

include/net/bluetooth/hci.h | 1 +
net/bluetooth/hci_core.c | 15 ++++++++++-----

2 files changed, 11 insertions(+), 5 deletions(-)

Regards

Marcel

#3 lore raw

PATCH v2 2/2 Bluetooth: Fix rfkill functionality during the HCI setup stage

In reply to [email protected]

From: Johan Hedberg <[email protected]>

We need to let the setup stage complete cleanly even when the HCI device is rfkilled. Otherwise the HCI device will stay in an undefined state and never get notified to user space through mgmt (even when it gets unblocked through rfkill).

This patch makes sure that hci_dev_open() can be called in the HCI_SETUP stage, that blocking the device doesn't abort the setup stage, and that the device gets proper powered down as soon as the setup stage completes in case it was blocked meanwhile.

The bug that this patch fixed can be very easily reproduced using e.g. the rfkill command line too. By running "rfkill block all" before inserting a Bluetooth dongle the resulting HCI device goes into a state where it is never announced over mgmt, not even when "rfkill unblock all" is run.

---
 net/bluetooth/hci_core.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)
net/bluetooth/hci_core.c +12 -2
index d0d6cf8..433820a 100644
@@ -1148,7 +1148,11 @@ int hci_dev_open(__u16 dev)
 		goto done;
 	}
 
-	if (test_bit(HCI_RFKILLED, &hdev->dev_flags)) {
+	/* Check for rfkill but allow the HCI setup stage to proceed
+	 * (which in itself doesn't cause any RF activity).
+	 */
+	if (test_bit(HCI_RFKILLED, &hdev->dev_flags) &&
+	    !test_bit(HCI_SETUP, &hdev->dev_flags)) {
 		ret = -ERFKILL;
 		goto done;
 	}
@@ -1599,7 +1603,8 @@ static int hci_rfkill_set_block(void *data, bool blocked)
 
 	if (blocked) {
 		set_bit(HCI_RFKILLED, &hdev->dev_flags);
-		hci_dev_do_close(hdev);
+		if (!test_bit(HCI_SETUP, &hdev->dev_flags))
+			hci_dev_do_close(hdev);
 	} else {
 		clear_bit(HCI_RFKILLED, &hdev->dev_flags);
 	}
@@ -1624,6 +1629,11 @@ static void hci_power_on(struct work_struct *work)
 		return;
 	}
 
+	if (test_bit(HCI_RFKILLED, &hdev->dev_flags)) {
+		clear_bit(HCI_AUTO_OFF, &hdev->dev_flags);
+		hci_dev_do_close(hdev);
+	}
+
 	if (test_bit(HCI_AUTO_OFF, &hdev->dev_flags))
 		queue_delayed_work(hdev->req_workqueue, &hdev->power_off,
 				   HCI_AUTO_OFF_TIMEOUT);
Signature · 1 line
1.8.4.rc3
maintainer
#4 lore raw

Re: [PATCH v2 2/2] Bluetooth: Fix rfkill functionality during the HCI setup stage

Hi Johan,

Quoted 50 lines

We need to let the setup stage complete cleanly even when the HCI device is rfkilled. Otherwise the HCI device will stay in an undefined state and never get notified to user space through mgmt (even when it gets unblocked through rfkill).

This patch makes sure that hci_dev_open() can be called in the HCI_SETUP stage, that blocking the device doesn't abort the setup stage, and that the device gets proper powered down as soon as the setup stage completes in case it was blocked meanwhile.

The bug that this patch fixed can be very easily reproduced using e.g. the rfkill command line too. By running "rfkill block all" before inserting a Bluetooth dongle the resulting HCI device goes into a state where it is never announced over mgmt, not even when "rfkill unblock all" is run.

---
net/bluetooth/hci_core.c | 14 ++++++++++++--

1 file changed, 12 insertions(+), 2 deletions(-)
net/bluetooth/hci_core.c +7 -2
index d0d6cf8..433820a 100644
@@ -1148,7 +1148,11 @@ int hci_dev_open(__u16 dev)
 		goto done;
 	}
 
-	if (test_bit(HCI_RFKILLED, &hdev->dev_flags)) {
+	/* Check for rfkill but allow the HCI setup stage to proceed
+	 * (which in itself doesn't cause any RF activity).
+	 */
+	if (test_bit(HCI_RFKILLED, &hdev->dev_flags) &&
+	    !test_bit(HCI_SETUP, &hdev->dev_flags)) {
 		ret = -ERFKILL;
 		goto done;
 	}
@@ -1599,7 +1603,8 @@ static int hci_rfkill_set_block(void *data, bool blocked)
 
 	if (blocked) {
 		set_bit(HCI_RFKILLED, &hdev->dev_flags);
-		hci_dev_do_close(hdev);
+		if (!test_bit(HCI_SETUP, &hdev->dev_flags))
+			hci_dev_do_close(hdev);
 	} else {
 		clear_bit(HCI_RFKILLED, &hdev->dev_flags);
 	}

net/bluetooth/hci_core.c static void hci_power_on(struct work_struct *work) · line 1636

Quoted 8 lines
net/bluetooth/hci_core.c +5 -0
@@ -1624,6 +1629,11 @@ static void hci_power_on(struct work_struct *work)
 		return;
 	}
 
+	if (test_bit(HCI_RFKILLED, &hdev->dev_flags)) {
+		clear_bit(HCI_AUTO_OFF, &hdev->dev_flags);
+		hci_dev_do_close(hdev);
+	}
+

this might be better extend with

	} else if (test_bit(HCI_AUTO_OFF, ...)

net/bluetooth/hci_core.c static void hci_power_on(struct work_struct *work) · line 1639

Quoted 3 lines
net/bluetooth/hci_core.c +0 -0
 	if (test_bit(HCI_AUTO_OFF, &hdev->dev_flags))
 		queue_delayed_work(hdev->req_workqueue, &hdev->power_off,
 				   HCI_AUTO_OFF_TIMEOUT);

Makes the code a little bit more readable for the reason that I want to add a return from the function to the HCI_RFKILLED check every time I read it, but that would be actually wrong.

Regards

Marcel