Skip to content

Instantly share code, notes, and snippets.

What would you like to do?
[ARCHIVED] Fix for CSR Dongle 0a12:0001 ID 0a12:0001 Cambridge Silicon Radio, Ltd Bluetooth Dongle (HCI mode)

This gist is currenctly archived.

Please refer to previous revisions if you know what to do.

The patch proposed was merged into kernel in 5.8 release, but no longer working as of linux 5.11

Copy link

I`m running Fedora 37 with kernel 6.0.7-301, unmodified.

Also getting errors, it used to work just fine on Fedora 35 and on older kernels in other distros.

[ 15.034386] Bluetooth: Core ver 2.22
[ 15.034400] Bluetooth: HCI device and connection manager initialized
[ 15.034402] Bluetooth: HCI socket layer initialized
[ 15.034404] Bluetooth: L2CAP socket layer initialized
[ 15.034407] Bluetooth: SCO socket layer initialized
[ 15.825231] Bluetooth: hci0: CSR: Unbranded CSR clone detected; adding workarounds and force-suspending once...
[ 15.825235] Bluetooth: hci0: CSR: Couldn't suspend the device for our Barrot 8041a02 receive-issue workaround
[ 15.825238] Bluetooth: hci0: HCI Delete Stored Link Key command is advertised, but not supported.
[ 15.825239] Bluetooth: hci0: HCI Set Event Filter command not supported.
[ 17.948972] Bluetooth: hci0: command 0x0c5a tx timeout
[ 17.949005] Bluetooth: hci0: Opcode 0x c5a failed: -110
[ 20.711035] Bluetooth: BNEP (Ethernet Emulation) ver 1.3
[ 20.711037] Bluetooth: BNEP filters: protocol multicast
[ 20.711041] Bluetooth: BNEP socket layer initialized

Copy link

Swyter commented Nov 22, 2022

See here:

Submitted a patch series here:

There was another simultaneous conversation in the mailing list here:,_ltd_bluetooth_dongle_unusable_again_with_kernel_6.0#post4683596

In the end, Luiz merged 2 out of the 3 patches upstream. There's radio silence about the third one.

Subscribe to the Bugzilla ticket if you want to stay up to date.

Copy link

The patch series talked above is currently empty:

I'd patch my own kernel if possible. This problem has kept me out of Linux entirely.

Copy link

dukercs commented Dec 21, 2022

Anyone who needs a patch for linux 5.15.x LTS series, can use this patch. Credit goes to @Swyter for creating the original patches. I just combined them and made it compatible with linux 5.15.x LTS. It has been tested on linux 5.15.27.

Just copy and save it as a .diff file and apply it on the kernel source.

diff --git a/drivers/bluetooth/btusb.c b/drivers/bluetooth/btusb.c
index ac90392cc..39b189b16 100644
--- a/drivers/bluetooth/btusb.c
+++ b/drivers/bluetooth/btusb.c
@@ -1942,6 +1942,8 @@ static int btusb_setup_csr(struct hci_dev *hdev)
 		set_bit(HCI_QUIRK_BROKEN_STORED_LINK_KEY, &hdev->quirks);
 		set_bit(HCI_QUIRK_BROKEN_ERR_DATA_REPORTING, &hdev->quirks);
+		set_bit(HCI_QUIRK_BROKEN_FILTER_CLEAR_ALL, &hdev->quirks);
+		set_bit(HCI_QUIRK_NO_SUSPEND_NOTIFIER, &hdev->quirks);
 		/* Clear the reset quirk since this is not an actual
 		 * early Bluetooth 1.1 device from CSR.
@@ -1952,16 +1954,16 @@ static int btusb_setup_csr(struct hci_dev *hdev)
 		 * Special workaround for these BT 4.0 chip clones, and potentially more:
-		 * - 0x0134: a Barrot 8041a02                 (HCI rev: 0x1012 sub: 0x0810)
+		 * - 0x0134: a Barrot 8041a02                 (HCI rev: 0x0810 sub: 0x1012)
 		 * - 0x7558: IC markings FR3191AHAL 749H15143 (HCI rev/sub-version: 0x0709)
 		 * These controllers are really messed-up.
 		 * 1. Their bulk RX endpoint will never report any data unless
-		 * the device was suspended at least once (yes, really).
+		 *    the device was suspended at least once (yes, really).
 		 * 2. They will not wakeup when autosuspended and receiving data
-		 * on their bulk RX endpoint from e.g. a keyboard or mouse
-		 * (IOW remote-wakeup support is broken for the bulk endpoint).
+		 *    on their bulk RX endpoint from e.g. a keyboard or mouse
+		 *    (IOW remote-wakeup support is broken for the bulk endpoint).
 		 * To fix 1. enable runtime-suspend, force-suspend the
 		 * HCI and then wake-it up by disabling runtime-suspend.
@@ -1981,7 +1983,7 @@ static int btusb_setup_csr(struct hci_dev *hdev)
 		if (ret >= 0)
-			bt_dev_err(hdev, "CSR: Failed to suspend the device for our Barrot 8041a02 receive-issue workaround");
+			bt_dev_warn(hdev, "CSR: Couldn't suspend the device for our Barrot 8041a02 receive-issue workaround");
diff --git a/include/net/bluetooth/hci.h b/include/net/bluetooth/hci.h
index 9ce46cb85..41f0026b6 100644
--- a/include/net/bluetooth/hci.h
+++ b/include/net/bluetooth/hci.h
@@ -255,6 +255,16 @@ enum {
 	 * during the hdev->setup vendor callback.
+	/* When this quirk is set, HCI_OP_SET_EVENT_FLT requests with
+	 * HCI_FLT_CLEAR_ALL are ignored and event filtering is
+	 * completely avoided. A subset of the CSR controller
+	 * clones struggle with this and instantly lock up.
+	 *
+	 * Note that devices using this must (separately) disable
+	 * runtime suspend, because event filtering takes place there.
+	 */
 /* HCI device flags */
diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
index 53f1b0801..4c7f748b8 100644
--- a/net/bluetooth/hci_core.c
+++ b/net/bluetooth/hci_core.c
@@ -271,6 +271,7 @@ static int hci_init1_req(struct hci_request *req, unsigned long opt)
 static void bredr_setup(struct hci_request *req)
+	struct hci_dev *hdev = req->hdev;
 	__le16 param;
 	__u8 flt_type;
@@ -292,9 +293,14 @@ static void bredr_setup(struct hci_request *req)
 	/* Read Current IAC LAP */
 	hci_req_add(req, HCI_OP_READ_CURRENT_IAC_LAP, 0, NULL);
-	/* Clear Event Filters */
-	flt_type = HCI_FLT_CLEAR_ALL;
-	hci_req_add(req, HCI_OP_SET_EVENT_FLT, 1, &flt_type);
+	/* Clear Event Filters; some fake CSR controllers lock up after setting
+	 * this type of filter, so avoid sending the request altogether.
+	 */
+	if (!test_bit(HCI_QUIRK_BROKEN_FILTER_CLEAR_ALL, &hdev->quirks))
+	{
+		flt_type = HCI_FLT_CLEAR_ALL;
+		hci_req_add(req, HCI_OP_SET_EVENT_FLT, 1, &flt_type);
+	}
 	/* Connection accept timeout ~20 secs */
 	param = cpu_to_le16(0x7d00);
diff --git a/net/bluetooth/hci_request.c b/net/bluetooth/hci_request.c
index 1d34d330a..35308de87 100644
--- a/net/bluetooth/hci_request.c
+++ b/net/bluetooth/hci_request.c
@@ -1156,11 +1156,15 @@ static bool adv_instance_is_scannable(struct hci_dev *hdev, u8 instance)
 static void hci_req_clear_event_filter(struct hci_request *req)
 	struct hci_cp_set_event_filter f;
+	struct hci_dev *hdev = req->hdev;
+	if (!hci_dev_test_flag(hdev, HCI_BREDR_ENABLED))
+		return;
-	if (!hci_dev_test_flag(req->hdev, HCI_BREDR_ENABLED))
+	if (test_bit(HCI_QUIRK_BROKEN_FILTER_CLEAR_ALL, &hdev->quirks))
-	if (hci_dev_test_flag(req->hdev, HCI_EVENT_FILTER_CONFIGURED)) {
+	if (hci_dev_test_flag(hdev, HCI_EVENT_FILTER_CONFIGURED)) {
 		memset(&f, 0, sizeof(f));
 		f.flt_type = HCI_FLT_CLEAR_ALL;
 		hci_req_add(req, HCI_OP_SET_EVENT_FLT, 1, &f);
@@ -1169,15 +1173,18 @@ static void hci_req_clear_event_filter(struct hci_request *req)
 static void hci_req_set_event_filter(struct hci_request *req)
+	struct hci_dev *hdev = req->hdev;
 	struct bdaddr_list_with_flags *b;
 	struct hci_cp_set_event_filter f;
-	struct hci_dev *hdev = req->hdev;
 	u8 scan = SCAN_DISABLED;
 	bool scanning = test_bit(HCI_PSCAN, &hdev->flags);
 	if (!hci_dev_test_flag(hdev, HCI_BREDR_ENABLED))
+	if (test_bit(HCI_QUIRK_BROKEN_FILTER_CLEAR_ALL, &hdev->quirks))
+		return;
 	/* Always clear event filter when starting */

Thank you! Worked on Ubuntu 22.04!
uname -r

Copy link

superian commented Jan 5, 2023

How fascinating - on an earlier standard Ubuntu kernel, the one of these I have Just Worked (and worked better than the 1Mii B10 dongle that's supposed to be the gold standard for doing this on Linux!)

But on the 5.15 kernel, it doesn't. I am not sure I have the patience to patch it.

Copy link

Swyter commented Jan 5, 2023

Try a more updated or newer kernel instead of patching manually. Ubuntu and Debian are notorious for being extremely slow when moving from one version to another. The fix exists, it just hasn't arrived yet to your door. Arch Linux got it in early December, for example.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment