Skip to contentSkip to conversation

Bluetooth: hci_event: negotiate LE Data Length after features

RFC PATCH In review Kyle Pepper 3 participants 4 messages Oct 13 – Oct 15

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 Bluetooth: hci_event: negotiate LE Data Length after features

Request the controller transmit limits once both peers support DLE. Queue the connection with a reference so pending requests coalesce and connection deletion cancels them. Check the connection under the device lock before constructing the command.

Assisted-by OpenCode:gpt-6.1-sol

---
Submission review notes (not part of the commit message):
The user reports that retaining the explicit DLE request enables the N8 to disconnect automatically after audio stops, whereas removing it leaves the device connected. This idle-transition behavior is the reported reason for retaining the topic, not an in-stream timeout. October 9 no-DLE trials already negotiate TX 251 octets / 2120 us and stream successfully; startup throughput alone does not test idle behavior. The supplied captures do not establish a controlled post-STOP comparison or the causal mechanism. This patch requests data length during setup; it does not itself implement STOP or issue a disconnect command. The claimed ASHA idle-disconnect requirement needs a protocol citation. The corrected ownership implementation is included in the working kernel. Native disconnect/handle reuse, LL ordering and post-snapshot races remain unproved. Submit as an independent RFC with these limitations disclosed.

 include/net/bluetooth/hci_sync.h |  1 +
 net/bluetooth/hci_event.c        | 10 +++++++
 net/bluetooth/hci_sync.c         | 47 ++++++++++++++++++++++++++++++++
 3 files changed, 58 insertions(+)
include/net/bluetooth/hci_sync.h +1 -0
index 59cb918d1..5a37d3e50 100644
@@ -194,3 +194,4 @@ int hci_le_read_remote_features(struct hci_conn *conn);
 
 int hci_acl_change_pkt_type(struct hci_conn *conn, u16 pkt_type);
 int hci_le_set_phy(struct hci_conn *conn, u8 tx_phys, u8 rx_phys);
+int hci_le_set_data_len(struct hci_conn *conn);
net/bluetooth/hci_event.c +10 -0
index c055d16cf..4035f2659 100644
@@ -6846,6 +6846,16 @@ static void hci_le_remote_feat_complete_evt(struct hci_dev *hdev, void *data,
 				conn->le_tx_def_phys &= ~HCI_LE_SET_PHY_CODED;
 				conn->le_rx_def_phys &= ~HCI_LE_SET_PHY_CODED;
 			}
+
+			/* Request the controller maximum when both sides support
+			 * LL Data Length Extension. The Link Layer negotiates
+			 * the remote limits (Core v5.4, Vol 6, Part B,
+			 * Sec. 4.5.10).
+			 */
+			if ((hdev->le_features[0] & HCI_LE_DATA_LEN_EXT) &&
+			    (conn->le_features[0] & HCI_LE_DATA_LEN_EXT) &&
+			    hdev->le_max_tx_len > 0x001b)
+				hci_le_set_data_len(conn);
 		}
 
 		if (conn->state == BT_CONFIG) {
net/bluetooth/hci_sync.c +47 -0

50 lines not on this page · raw message

base-commit: 964430c038967fd06643e866f68115664e1a1bd1

#1 lore raw

This is automated email and please do not reply to this email!

Dear submitter,

Thank you for submitting the patches to the linux bluetooth mailing list. This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/series/1184346/

---Test result---

Test Summary:

CheckPatch                    PASS      0.89 seconds
VerifyFixes                   PASS      0.08 seconds
VerifySignedoff               PASS      0.08 seconds
GitLint                       PASS      0.23 seconds
SubjectPrefix                 PASS      0.07 seconds
BuildKernel                   PASS      22.62 seconds
CheckAllWarning               PASS      26.01 seconds
CheckSparse                   WARNING   28.12 seconds
BuildKernel32                 PASS      23.94 seconds
CheckKernelLLVM               PASS      26.98 seconds
TestRunnerSetup               PASS      528.87 seconds
TestRunner_l2cap-tester       PASS      13.34 seconds
TestRunner_iso-tester         PASS      9.71 seconds
TestRunner_bnep-tester        PASS      2.84 seconds
TestRunner_mgmt-tester        PASS      87.94 seconds
TestRunner_rfcomm-tester      PASS      4.08 seconds
TestRunner_sco-tester         PASS      5.06 seconds
TestRunner_ioctl-tester       PASS      4.38 seconds
TestRunner_mesh-tester        PASS      5.43 seconds
TestRunner_smp-tester         PASS      3.96 seconds
TestRunner_userchan-tester    PASS      2.82 seconds
TestRunner_6lowpan-tester     PASS      3.97 seconds
IncrementalBuild              PASS      21.61 seconds

Details
##############################
Test: CheckSparse - WARNING
Desc: Run sparse tool with linux kernel
Output:
net/bluetooth/hci_event.c: note: in included file (through include/net/bluetooth/hci_core.h):

https://github.com/bluez/bluetooth-next/pull/910

---
Regards,
Linux Bluetooth

#2 lore raw
In reply to Kyle Pepper

Hi Kyle,

Quoted 46 lines · Kyle Pepper

Request the controller transmit limits once both peers support DLE. Queue the connection with a reference so pending requests coalesce and connection deletion cancels them. Check the connection under the device lock before constructing the command.

Assisted-by OpenCode:gpt-6.1-sol

---
Submission review notes (not part of the commit message):
The user reports that retaining the explicit DLE request enables the N8 to disconnect automatically after audio stops, whereas removing it leaves the device connected. This idle-transition behavior is the reported reason for retaining the topic, not an in-stream timeout. October 9 no-DLE trials already negotiate TX 251 octets / 2120 us and stream successfully; startup throughput alone does not test idle behavior. The supplied captures do not establish a controlled post-STOP comparison or the causal mechanism. This patch requests data length during setup; it does not itself implement STOP or issue a disconnect command. The claimed ASHA idle-disconnect requirement needs a protocol citation. The corrected ownership implementation is included in the working kernel. Native disconnect/handle reuse, LL ordering and post-snapshot races remain unproved. Submit as an independent RFC with these limitations disclosed.

 include/net/bluetooth/hci_sync.h |  1 +
 net/bluetooth/hci_event.c        | 10 +++++++
 net/bluetooth/hci_sync.c         | 47 ++++++++++++++++++++++++++++++++
 3 files changed, 58 insertions(+)
include/net/bluetooth/hci_sync.h +1 -0
index 59cb918d1..5a37d3e50 100644
@@ -194,3 +194,4 @@ int hci_le_read_remote_features(struct hci_conn *conn);
 
 int hci_acl_change_pkt_type(struct hci_conn *conn, u16 pkt_type);
 int hci_le_set_phy(struct hci_conn *conn, u8 tx_phys, u8 rx_phys);
+int hci_le_set_data_len(struct hci_conn *conn);
net/bluetooth/hci_event.c +0 -0
index c055d16cf..4035f2659 100644
@@ -6846,6 +6846,16 @@ static void hci_le_remote_feat_complete_evt(struct hci_dev *hdev, void *data,
                                conn->le_tx_def_phys &= ~HCI_LE_SET_PHY_CODED;
                                conn->le_rx_def_phys &= ~HCI_LE_SET_PHY_CODED;
                        }

net/bluetooth/hci_event.c static void hci_le_remote_feat_complete_evt(struct hci_dev *hdev, void *data, · line 6858

Quoted 10 lines
net/bluetooth/hci_event.c +10 -0
+
+                       /* Request the controller maximum when both sides support
+                        * LL Data Length Extension. The Link Layer negotiates
+                        * the remote limits (Core v5.4, Vol 6, Part B,
+                        * Sec. 4.5.10).
+                        */
+                       if ((hdev->le_features[0] & HCI_LE_DATA_LEN_EXT) &&
+                           (conn->le_features[0] & HCI_LE_DATA_LEN_EXT) &&
+                           hdev->le_max_tx_len > 0x001b)
+                               hci_le_set_data_len(conn);

We use HCI_OP_LE_WRITE_DEF_DATA_LEN to setup this for every connection, so either the controller has a bug and doesn't trigger trigger DLE or something else is off here. In case it is really a specific controller then we might need a quirk which once set by the driver indicates the host has to perform the DLE procedure manually like above.

Quoted 61 lines
net/bluetooth/hci_event.c +0 -0
                }
 
                if (conn->state == BT_CONFIG) {
net/bluetooth/hci_sync.c +47 -0
index efbd2e538..20c690007 100644
@@ -8123,3 +8123,50 @@ int hci_le_set_phy(struct hci_conn *conn, u8 tx_phys, u8 rx_phys)
 
        return (err == -EEXIST) ? 0 : err;
 }
+
+static void le_set_data_len_complete(struct hci_dev *hdev, void *data,
+                                    int err)
+{
+       struct hci_conn *conn = data;
+
+       bt_dev_dbg(hdev, "err %d", err);
+
+       hci_conn_put(conn);
+}
+
+static int hci_le_set_data_len_sync(struct hci_dev *hdev, void *data)
+{
+       struct hci_conn *conn = data;
+       struct hci_cp_le_set_data_len cp;
+
+       hci_dev_lock(hdev);
+
+       if (!hci_conn_valid(hdev, conn) || conn->type != LE_LINK ||
+           (conn->state != BT_CONNECTED && conn->state != BT_CONFIG)) {
+               hci_dev_unlock(hdev);
+               return -ECANCELED;
+       }
+
+       cp.handle = cpu_to_le16(conn->handle);
+       cp.tx_len = cpu_to_le16(hdev->le_max_tx_len);
+       cp.tx_time = cpu_to_le16(hdev->le_max_tx_time);
+
+       hci_dev_unlock(hdev);
+
+       return __hci_cmd_sync_status(hdev, HCI_OP_LE_SET_DATA_LEN,
+                                    sizeof(cp), &cp, HCI_CMD_TIMEOUT);
+}
+
+int hci_le_set_data_len(struct hci_conn *conn)
+{
+       struct hci_dev *hdev = conn->hdev;
+       int err;
+
+       err = hci_cmd_sync_queue_once(hdev, hci_le_set_data_len_sync,
+                                     hci_conn_get(conn),
+                                     le_set_data_len_complete);
+       if (err)
+               hci_conn_put(conn);
+
+       return (err == -EEXIST) ? 0 : err;
+}

base-commit: 964430c038967fd06643e866f68115664e1a1bd1

Signature · 1 line
Luiz Augusto von Dentz
author
#3 lore raw

Thanks for pointing that out! Initialization already programs the suggested default data length. I also need to correct my previous submission notes, the N8 returns to environmental sound while Bluetooth remains connected. The longer idle transition showed continued host audio before STOP. Instead of a demonstrated DLE negotiation failure.

I’ll withdraw this RFC. I don’t have evidence of an affected controller that justifies a manual-DLE quirk.

Quoted 133 lines · Luiz Augusto von Dentz

Hi Kyle,

Quoted 46 lines · Kyle Pepper

Request the controller transmit limits once both peers support DLE. Queue the connection with a reference so pending requests coalesce and connection deletion cancels them. Check the connection under the device lock before constructing the command.

Assisted-by OpenCode:gpt-6.1-sol

---
Submission review notes (not part of the commit message):
The user reports that retaining the explicit DLE request enables the N8 to disconnect automatically after audio stops, whereas removing it leaves the device connected. This idle-transition behavior is the reported reason for retaining the topic, not an in-stream timeout. October 9 no-DLE trials already negotiate TX 251 octets / 2120 us and stream successfully; startup throughput alone does not test idle behavior. The supplied captures do not establish a controlled post-STOP comparison or the causal mechanism. This patch requests data length during setup; it does not itself implement STOP or issue a disconnect command. The claimed ASHA idle-disconnect requirement needs a protocol citation. The corrected ownership implementation is included in the working kernel. Native disconnect/handle reuse, LL ordering and post-snapshot races remain unproved. Submit as an independent RFC with these limitations disclosed.

 include/net/bluetooth/hci_sync.h |  1 +
 net/bluetooth/hci_event.c        | 10 +++++++
 net/bluetooth/hci_sync.c         | 47 ++++++++++++++++++++++++++++++++
 3 files changed, 58 insertions(+)
include/net/bluetooth/hci_sync.h +1 -0
index 59cb918d1..5a37d3e50 100644
@@ -194,3 +194,4 @@ int hci_le_read_remote_features(struct hci_conn *conn);
 
 int hci_acl_change_pkt_type(struct hci_conn *conn, u16 pkt_type);
 int hci_le_set_phy(struct hci_conn *conn, u8 tx_phys, u8 rx_phys);
+int hci_le_set_data_len(struct hci_conn *conn);
net/bluetooth/hci_event.c +0 -0
index c055d16cf..4035f2659 100644
@@ -6846,6 +6846,16 @@ static void hci_le_remote_feat_complete_evt(struct hci_dev *hdev, void *data,
                                conn->le_tx_def_phys &= ~HCI_LE_SET_PHY_CODED;
                                conn->le_rx_def_phys &= ~HCI_LE_SET_PHY_CODED;
                        }

net/bluetooth/hci_event.c static void hci_le_remote_feat_complete_evt(struct hci_dev *hdev, void *data, · line 6858

Quoted 10 lines
net/bluetooth/hci_event.c +10 -0
+
+                       /* Request the controller maximum when both sides support
+                        * LL Data Length Extension. The Link Layer negotiates
+                        * the remote limits (Core v5.4, Vol 6, Part B,
+                        * Sec. 4.5.10).
+                        */
+                       if ((hdev->le_features[0] & HCI_LE_DATA_LEN_EXT) &&
+                           (conn->le_features[0] & HCI_LE_DATA_LEN_EXT) &&
+                           hdev->le_max_tx_len > 0x001b)
+                               hci_le_set_data_len(conn);

We use HCI_OP_LE_WRITE_DEF_DATA_LEN to setup this for every connection, so either the controller has a bug and doesn't trigger trigger DLE or something else is off here. In case it is really a specific controller then we might need a quirk which once set by the driver indicates the host has to perform the DLE procedure manually like above.

Quoted 59 lines
net/bluetooth/hci_event.c +0 -0
                }
 
                if (conn->state == BT_CONFIG) {
net/bluetooth/hci_sync.c +47 -0
index efbd2e538..20c690007 100644
@@ -8123,3 +8123,50 @@ int hci_le_set_phy(struct hci_conn *conn, u8 tx_phys, u8 rx_phys)
 
        return (err == -EEXIST) ? 0 : err;
 }
+
+static void le_set_data_len_complete(struct hci_dev *hdev, void *data,
+                                    int err)
+{
+       struct hci_conn *conn = data;
+
+       bt_dev_dbg(hdev, "err %d", err);
+
+       hci_conn_put(conn);
+}
+
+static int hci_le_set_data_len_sync(struct hci_dev *hdev, void *data)
+{
+       struct hci_conn *conn = data;
+       struct hci_cp_le_set_data_len cp;
+
+       hci_dev_lock(hdev);
+
+       if (!hci_conn_valid(hdev, conn) || conn->type != LE_LINK ||
+           (conn->state != BT_CONNECTED && conn->state != BT_CONFIG)) {
+               hci_dev_unlock(hdev);
+               return -ECANCELED;
+       }
+
+       cp.handle = cpu_to_le16(conn->handle);
+       cp.tx_len = cpu_to_le16(hdev->le_max_tx_len);
+       cp.tx_time = cpu_to_le16(hdev->le_max_tx_time);
+
+       hci_dev_unlock(hdev);
+
+       return __hci_cmd_sync_status(hdev, HCI_OP_LE_SET_DATA_LEN,
+                                    sizeof(cp), &cp, HCI_CMD_TIMEOUT);
+}
+
+int hci_le_set_data_len(struct hci_conn *conn)
+{
+       struct hci_dev *hdev = conn->hdev;
+       int err;
+
+       err = hci_cmd_sync_queue_once(hdev, hci_le_set_data_len_sync,
+                                     hci_conn_get(conn),
+                                     le_set_data_len_complete);
+       if (err)
+               hci_conn_put(conn);
+
+       return (err == -EEXIST) ? 0 : err;
+}
Quoted 2 lines

base-commit: 964430c038967fd06643e866f68115664e1a1bd1

--
Luiz Augusto von Dentz