Re: [PATCH v4] Bluetooth: l2cap: defer conn param update to avoid conn->lock/hdev->lock inversion

From: Luiz Augusto von Dentz

Date: Tue Apr 14 2026 - 16:28:17 EST


Hi Mikhail,

On Tue, Apr 14, 2026 at 2:44 PM Mikhail Gavrilov
<mikhail.v.gavrilov@xxxxxxxxx> wrote:
>
> When a BLE peripheral sends an L2CAP Connection Parameter Update Request
> the processing path is:
>
> process_pending_rx() [takes conn->lock]
> l2cap_le_sig_channel()
> l2cap_conn_param_update_req()
> hci_le_conn_update() [takes hdev->lock]
>
> Meanwhile other code paths take the locks in the opposite order:
>
> l2cap_chan_connect() [takes hdev->lock]
> ...
> mutex_lock(&conn->lock)
>
> l2cap_conn_ready() [hdev->lock via hci_cb_list_lock]
> ...
> mutex_lock(&conn->lock)
>
> This is a classic AB/BA deadlock which lockdep reports as a circular
> locking dependency when connecting a BLE MIDI keyboard (Carry-On FC-49).
>
> Fix this by making hci_le_conn_update() defer the HCI command through
> hci_cmd_sync_queue() so it no longer needs to take hdev->lock in the
> caller context. The sync callback uses __hci_cmd_sync_status_sk() to
> wait for the HCI_EV_LE_CONN_UPDATE_COMPLETE event, then updates the
> stored connection parameters (hci_conn_params) and notifies userspace
> (mgmt_new_conn_param) only after the controller has confirmed the update.
>
> A reference on hci_conn is held via hci_conn_get()/hci_conn_put() for
> the lifetime of the queued work to prevent use-after-free, and
> hci_conn_valid() is checked before proceeding in case the connection was
> removed while the work was pending.
>
> Fixes: f044eb0524a0 ("Bluetooth: Store latency and supervision timeout in connection params")
> Signed-off-by: Mikhail Gavrilov <mikhail.v.gavrilov@xxxxxxxxx>
> Reviewed-by: Paul Menzel <pmenzel@xxxxxxxxxxxxx>
> ---
>
> Changes in v4 (Luiz Augusto von Dentz, Sashiko/Gemini AI review):
> - Use hci_conn_get()/hci_conn_put() to hold a reference while work is queued
> - Use __hci_cmd_sync_status_sk() to wait for HCI_EV_LE_CONN_UPDATE_COMPLETE,
> then do params update + mgmt notification in the sync callback — removes the
> need for hci_event.c changes and hci_sent_cmd_data()
> - Use kzalloc_obj() per checkpatch recommendation
> - Fix checkpatch alignment issues
>
> Changes in v3 (Luiz Augusto von Dentz):
> - Move hci_cmd_sync_queue into hci_le_conn_update itself instead of open-coding
> the deferral in l2cap_core.c
> - Move conn_params update and mgmt_new_conn_param into
> hci_le_conn_update_complete_evt, using hci_sent_cmd_data to retrieve
> the originally requested parameters
>
> Changes in v2 (Paul Menzel, Sashiko/Gemini AI review):
> - Allocate before sending ACCEPTED response to avoid state mismatch on OOM
> - Verify connection handle and address in sync callback against reuse race
> - Expand commit message with implementation details
>
> include/net/bluetooth/hci_core.h | 2 +-
> net/bluetooth/hci_conn.c | 98 +++++++++++++++++++++++++-------
> net/bluetooth/l2cap_core.c | 12 +---
> 3 files changed, 82 insertions(+), 30 deletions(-)
>
> diff --git a/include/net/bluetooth/hci_core.h b/include/net/bluetooth/hci_core.h
> index a7bffb908c1e..aa600fbf9a53 100644
> --- a/include/net/bluetooth/hci_core.h
> +++ b/include/net/bluetooth/hci_core.h
> @@ -2495,7 +2495,7 @@ void mgmt_adv_monitor_device_lost(struct hci_dev *hdev, u16 handle,
> bdaddr_t *bdaddr, u8 addr_type);
>
> int hci_abort_conn(struct hci_conn *conn, u8 reason);
> -u8 hci_le_conn_update(struct hci_conn *conn, u16 min, u16 max, u16 latency,
> +void hci_le_conn_update(struct hci_conn *conn, u16 min, u16 max, u16 latency,
> u16 to_multiplier);
> void hci_le_start_enc(struct hci_conn *conn, __le16 ediv, __le64 rand,
> __u8 ltk[16], __u8 key_size);
> diff --git a/net/bluetooth/hci_conn.c b/net/bluetooth/hci_conn.c
> index 11d3ad8d2551..c9cc18fab1f2 100644
> --- a/net/bluetooth/hci_conn.c
> +++ b/net/bluetooth/hci_conn.c
> @@ -480,40 +480,100 @@ bool hci_setup_sync(struct hci_conn *conn, __u16 handle)
> return hci_setup_sync_conn(conn, handle);
> }
>
> -u8 hci_le_conn_update(struct hci_conn *conn, u16 min, u16 max, u16 latency,
> - u16 to_multiplier)
> +struct le_conn_update_data {
> + struct hci_conn *conn;
> + u16 min;
> + u16 max;
> + u16 latency;
> + u16 to_multiplier;
> +};
> +
> +static int le_conn_update_sync(struct hci_dev *hdev, void *data)
> {
> - struct hci_dev *hdev = conn->hdev;
> + struct le_conn_update_data *d = data;
> + struct hci_conn *conn = d->conn;
> struct hci_conn_params *params;
> struct hci_cp_le_conn_update cp;
> + u8 store_hint;
> + int err;
>
> + /* Verify connection is still alive. */
> hci_dev_lock(hdev);
> -
> - params = hci_conn_params_lookup(hdev, &conn->dst, conn->dst_type);
> - if (params) {
> - params->conn_min_interval = min;
> - params->conn_max_interval = max;
> - params->conn_latency = latency;
> - params->supervision_timeout = to_multiplier;
> + if (!hci_conn_valid(hdev, conn)) {
> + hci_dev_unlock(hdev);
> + return -ECANCELED;
> }
> -
> hci_dev_unlock(hdev);
>
> memset(&cp, 0, sizeof(cp));
> cp.handle = cpu_to_le16(conn->handle);
> - cp.conn_interval_min = cpu_to_le16(min);
> - cp.conn_interval_max = cpu_to_le16(max);
> - cp.conn_latency = cpu_to_le16(latency);
> - cp.supervision_timeout = cpu_to_le16(to_multiplier);
> + cp.conn_interval_min = cpu_to_le16(d->min);
> + cp.conn_interval_max = cpu_to_le16(d->max);
> + cp.conn_latency = cpu_to_le16(d->latency);
> + cp.supervision_timeout = cpu_to_le16(d->to_multiplier);
> cp.min_ce_len = cpu_to_le16(0x0000);
> cp.max_ce_len = cpu_to_le16(0x0000);
>
> - hci_send_cmd(hdev, HCI_OP_LE_CONN_UPDATE, sizeof(cp), &cp);
> + err = __hci_cmd_sync_status_sk(hdev, HCI_OP_LE_CONN_UPDATE,
> + sizeof(cp), &cp,
> + HCI_EV_LE_CONN_UPDATE_COMPLETE,
> + HCI_CMD_TIMEOUT, NULL);
> + if (err)
> + return err;
> +
> + /* Update stored connection parameters after the controller has
> + * confirmed the update via the LE Connection Update Complete event.
> + */
> + hci_dev_lock(hdev);
>
> - if (params)
> - return 0x01;
> + params = hci_conn_params_lookup(hdev, &conn->dst, conn->dst_type);
> + if (params) {
> + params->conn_min_interval = d->min;
> + params->conn_max_interval = d->max;
> + params->conn_latency = d->latency;
> + params->supervision_timeout = d->to_multiplier;
> + store_hint = 0x01;
> + } else {
> + store_hint = 0x00;
> + }
>
> - return 0x00;
> + hci_dev_unlock(hdev);
> +
> + mgmt_new_conn_param(hdev, &conn->dst, conn->dst_type, store_hint,
> + d->min, d->max, d->latency, d->to_multiplier);
> +
> + return 0;
> +}
> +
> +static void le_conn_update_complete(struct hci_dev *hdev, void *data, int err)
> +{
> + struct le_conn_update_data *d = data;
> +
> + hci_conn_put(d->conn);
> + kfree(d);
> +}
> +
> +void hci_le_conn_update(struct hci_conn *conn, u16 min, u16 max, u16 latency,
> + u16 to_multiplier)
> +{
> + struct le_conn_update_data *d;
> +
> + d = kzalloc_obj(*d);
> + if (!d)
> + return;
> +
> + hci_conn_get(conn);
> + d->conn = conn;
> + d->min = min;
> + d->max = max;
> + d->latency = latency;
> + d->to_multiplier = to_multiplier;
> +
> + if (hci_cmd_sync_queue(conn->hdev, le_conn_update_sync, d,
> + le_conn_update_complete) < 0) {
> + hci_conn_put(conn);
> + kfree(d);
> + }
> }
>
> void hci_le_start_enc(struct hci_conn *conn, __le16 ediv, __le64 rand,
> diff --git a/net/bluetooth/l2cap_core.c b/net/bluetooth/l2cap_core.c
> index 95c65fece39b..aac2db1d6fbb 100644
> --- a/net/bluetooth/l2cap_core.c
> +++ b/net/bluetooth/l2cap_core.c
> @@ -4706,16 +4706,8 @@ static inline int l2cap_conn_param_update_req(struct l2cap_conn *conn,
> l2cap_send_cmd(conn, cmd->ident, L2CAP_CONN_PARAM_UPDATE_RSP,
> sizeof(rsp), &rsp);
>
> - if (!err) {
> - u8 store_hint;
> -
> - store_hint = hci_le_conn_update(hcon, min, max, latency,
> - to_multiplier);
> - mgmt_new_conn_param(hcon->hdev, &hcon->dst, hcon->dst_type,
> - store_hint, min, max, latency,
> - to_multiplier);
> -
> - }
> + if (!err)
> + hci_le_conn_update(hcon, min, max, latency, to_multiplier);
>
> return 0;
> }
> --
> 2.53.0

https://sashiko.dev/#/patchset/20260414184415.17114-1-mikhail.v.gavrilov%40gmail.com

1. I don't think it really matters because connection intervals are
typically 60ms at most, so 6*60 = 360ms. That said, perhaps we should
use 6 * max_interval as the timeout since that is how long the
controller has to update the connection interval, well if the 6
consecutive intervals are actually a thing since I don't recall
anything like that being mentioned in the command, but perhaps it is
in the link layer. But if it times out I don't think there is a way to
cancel the change either, so I'm not sure how we'll be able to
synchronize if the controller misbehaves.
2. Well I guess that is would need to defer the response and wait the
HCI proceedure to complete to really respond properly, but that would
mean the remote could get a LL procedure to change the connection
interval before the L2CAP response, so I can see this going wrong
either way.


--
Luiz Augusto von Dentz