net/bluetooth/hci_sync.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
From: Yang Li <yang.li@amlogic.com>
PA sync HCI commands incorrectly use conn->conn_timeout, which defaults
to 20s and is too long. If the PA Sync Established event is not
received, the command remains pending and causes userspace timeouts.
Signed-off-by: Yang Li <yang.li@amlogic.com>
---
net/bluetooth/hci_sync.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
index f04a90bce4a9..58292b01f062 100644
--- a/net/bluetooth/hci_sync.c
+++ b/net/bluetooth/hci_sync.c
@@ -7159,7 +7159,7 @@ static int hci_le_pa_create_sync(struct hci_dev *hdev, void *data)
if (conn->sid == HCI_SID_INVALID) {
err = __hci_cmd_sync_status_sk(hdev, HCI_OP_NOP, 0, NULL,
HCI_EV_LE_EXT_ADV_REPORT,
- conn->conn_timeout, NULL);
+ HCI_CMD_TIMEOUT, NULL);
if (err == -ETIMEDOUT)
goto done;
}
@@ -7187,7 +7187,7 @@ static int hci_le_pa_create_sync(struct hci_dev *hdev, void *data)
err = __hci_cmd_sync_status_sk(hdev, HCI_OP_LE_PA_CREATE_SYNC,
sizeof(cp), &cp,
HCI_EV_LE_PA_SYNC_ESTABLISHED,
- conn->conn_timeout, NULL);
+ HCI_CMD_TIMEOUT, NULL);
if (err == -ETIMEDOUT)
__hci_cmd_sync_status(hdev, HCI_OP_LE_PA_CREATE_SYNC_CANCEL,
0, NULL, HCI_CMD_TIMEOUT);
---
base-commit: bdb9aba465a85122bcba01a6154e237b9843e25b
change-id: 20260108-pa_sync_cmd_timeout-ff46f67b5d2a
Best regards,
--
Yang Li <yang.li@amlogic.com>
Hi Yang,
On Thu, Jan 8, 2026 at 1:03 AM Yang Li via B4 Relay
<devnull+yang.li.amlogic.com@kernel.org> wrote:
>
> From: Yang Li <yang.li@amlogic.com>
>
> PA sync HCI commands incorrectly use conn->conn_timeout, which defaults
> to 20s and is too long. If the PA Sync Established event is not
> received, the command remains pending and causes userspace timeouts.
Im pretty sure this is impossible scenario with bluetooth-next, since
1. only one work can be running hci_le_pa_create_sync at time so
HCI_EV_LE_PA_SYNC_ESTABLISHED cannot be outstanding 2. The
conn->timeout shall be aligned with socket timeout so perhaps the
problem is userspace is not setting the socket timeout correctly or
something.
> Signed-off-by: Yang Li <yang.li@amlogic.com>
> ---
> net/bluetooth/hci_sync.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
> index f04a90bce4a9..58292b01f062 100644
> --- a/net/bluetooth/hci_sync.c
> +++ b/net/bluetooth/hci_sync.c
> @@ -7159,7 +7159,7 @@ static int hci_le_pa_create_sync(struct hci_dev *hdev, void *data)
> if (conn->sid == HCI_SID_INVALID) {
> err = __hci_cmd_sync_status_sk(hdev, HCI_OP_NOP, 0, NULL,
> HCI_EV_LE_EXT_ADV_REPORT,
> - conn->conn_timeout, NULL);
> + HCI_CMD_TIMEOUT, NULL);
> if (err == -ETIMEDOUT)
> goto done;
> }
> @@ -7187,7 +7187,7 @@ static int hci_le_pa_create_sync(struct hci_dev *hdev, void *data)
> err = __hci_cmd_sync_status_sk(hdev, HCI_OP_LE_PA_CREATE_SYNC,
> sizeof(cp), &cp,
> HCI_EV_LE_PA_SYNC_ESTABLISHED,
> - conn->conn_timeout, NULL);
> + HCI_CMD_TIMEOUT, NULL);
> if (err == -ETIMEDOUT)
> __hci_cmd_sync_status(hdev, HCI_OP_LE_PA_CREATE_SYNC_CANCEL,
> 0, NULL, HCI_CMD_TIMEOUT);
>
> ---
> base-commit: bdb9aba465a85122bcba01a6154e237b9843e25b
> change-id: 20260108-pa_sync_cmd_timeout-ff46f67b5d2a
>
> Best regards,
> --
> Yang Li <yang.li@amlogic.com>
>
>
--
Luiz Augusto von Dentz
Hi Luiz
> [ EXTERNAL EMAIL ]
>
> Hi Yang,
>
> On Thu, Jan 8, 2026 at 1:03 AM Yang Li via B4 Relay
> <devnull+yang.li.amlogic.com@kernel.org> wrote:
>> From: Yang Li <yang.li@amlogic.com>
>>
>> PA sync HCI commands incorrectly use conn->conn_timeout, which defaults
>> to 20s and is too long. If the PA Sync Established event is not
>> received, the command remains pending and causes userspace timeouts.
> Im pretty sure this is impossible scenario with bluetooth-next, since
> 1. only one work can be running hci_le_pa_create_sync at time so
> HCI_EV_LE_PA_SYNC_ESTABLISHED cannot be outstanding 2. The
> conn->timeout shall be aligned with socket timeout so perhaps the
> problem is userspace is not setting the socket timeout correctly or
> something.
Well, I got it.
I noticed that BlueZ has a default setting for sync_time with the value
BT_ISO_SYNC_TIMEOUT=0x07d0 /* 20 secs */.
I would like to know the rationale behind setting it to 20 seconds
initially. This value seems a bit long to me. Can I directly change this
value as:
BT_ISO_SYNC_TIMEOUT=0x00C8 /* 2 secs */
>> Signed-off-by: Yang Li <yang.li@amlogic.com>
>> ---
>> net/bluetooth/hci_sync.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
>> index f04a90bce4a9..58292b01f062 100644
>> --- a/net/bluetooth/hci_sync.c
>> +++ b/net/bluetooth/hci_sync.c
>> @@ -7159,7 +7159,7 @@ static int hci_le_pa_create_sync(struct hci_dev *hdev, void *data)
>> if (conn->sid == HCI_SID_INVALID) {
>> err = __hci_cmd_sync_status_sk(hdev, HCI_OP_NOP, 0, NULL,
>> HCI_EV_LE_EXT_ADV_REPORT,
>> - conn->conn_timeout, NULL);
>> + HCI_CMD_TIMEOUT, NULL);
>> if (err == -ETIMEDOUT)
>> goto done;
>> }
>> @@ -7187,7 +7187,7 @@ static int hci_le_pa_create_sync(struct hci_dev *hdev, void *data)
>> err = __hci_cmd_sync_status_sk(hdev, HCI_OP_LE_PA_CREATE_SYNC,
>> sizeof(cp), &cp,
>> HCI_EV_LE_PA_SYNC_ESTABLISHED,
>> - conn->conn_timeout, NULL);
>> + HCI_CMD_TIMEOUT, NULL);
>> if (err == -ETIMEDOUT)
>> __hci_cmd_sync_status(hdev, HCI_OP_LE_PA_CREATE_SYNC_CANCEL,
>> 0, NULL, HCI_CMD_TIMEOUT);
>>
>> ---
>> base-commit: bdb9aba465a85122bcba01a6154e237b9843e25b
>> change-id: 20260108-pa_sync_cmd_timeout-ff46f67b5d2a
>>
>> Best regards,
>> --
>> Yang Li <yang.li@amlogic.com>
>>
>>
>
> --
> Luiz Augusto von Dentz
© 2016 - 2026 Red Hat, Inc.