Re: [PATCH] Bluetooth: btusb: clear remote wake on idle Intel ACPI paths

From: Paul Menzel

Date: Thu Mar 19 2026 - 03:37:44 EST


Dear Sean,


Thank you for your patch.

Am 18.03.26 um 20:51 schrieb Sean Rhodes:
Commit 6953217d9f44 ("usb: core: allow ACPI-managed hard-wired ports
to power off") allows internal USB devices on ACPI-managed hard-wired

I am unable to find the commit [1] besides on the Linux kernel list [2].

ports to use the existing runtime power-off path. For Intel combined USB
Bluetooth controllers, btusb still keeps runtime remote wake enabled for
as long as the adapter is open, which prevents the USB PM core from
dropping the last child reference and powering the port off while idle.

Only keep runtime remote wake enabled while the controller is expected
to signal activity on its own: established links, discovery, LE scan,
LE advertising, or BR/EDR page/inquiry scan. When those are idle, clear
needs_remote_wakeup so autosuspend can power-manage the port and a later
host-initiated command can resume the device again.

Could you add a test case and the device you tested with, and possible measurement numbers?

Signed-off-by: Sean Rhodes <sean@starlabs.systems>
---
drivers/bluetooth/btusb.c | 50 +++++++++++++++++++++++++++++++++++++--
1 file changed, 48 insertions(+), 2 deletions(-)

diff --git a/drivers/bluetooth/btusb.c b/drivers/bluetooth/btusb.c
index 5e9ebf0c5312..9df8a7c3254b 100644
--- a/drivers/bluetooth/btusb.c
+++ b/drivers/bluetooth/btusb.c
@@ -901,6 +901,7 @@ struct btusb_data {
unsigned long flags;
bool poll_sync;
+ bool intel_combined;
int intr_interval;
struct work_struct work;
struct work_struct waker;
@@ -958,6 +959,9 @@ struct btusb_data {
struct qca_dump_info qca_dump;
};
+static bool btusb_intel_idle_power_manageable(struct btusb_data *data);
+static bool btusb_needs_runtime_remote_wakeup(struct btusb_data *data);
+
static void btusb_reset(struct hci_dev *hdev)
{
struct btusb_data *data;
@@ -1151,12 +1155,19 @@ static inline void btusb_free_frags(struct btusb_data *data)
static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
{
+ int err;
+
if (data->intr_interval) {
/* Trigger dequeue immediately if an event is received */
schedule_delayed_work(&data->rx_work, 0);
}
- return data->recv_event(data->hdev, skb);
+ err = data->recv_event(data->hdev, skb);
+ if (!err)
+ data->intf->needs_remote_wakeup =
+ btusb_needs_runtime_remote_wakeup(data);
+
+ return err;
}
static int btusb_recv_intr(struct btusb_data *data, void *buffer, int count)
@@ -1924,7 +1935,8 @@ static int btusb_open(struct hci_dev *hdev)
goto setup_fail;
}
- data->intf->needs_remote_wakeup = 1;
+ data->intf->needs_remote_wakeup =
+ btusb_needs_runtime_remote_wakeup(data);
if (test_and_set_bit(BTUSB_INTR_RUNNING, &data->flags))
goto done;
@@ -1967,6 +1979,38 @@ static void btusb_stop_traffic(struct btusb_data *data)
usb_kill_anchored_urbs(&data->ctrl_anchor);
}
+static bool btusb_intel_idle_power_manageable(struct btusb_data *data)
+{
+ struct usb_device *udev = data->udev;
+
+ return data->intel_combined && udev->parent &&
+ usb_acpi_power_manageable(udev->parent, udev->portnum - 1);
+}
+
+static bool btusb_needs_runtime_remote_wakeup(struct btusb_data *data)
+{
+ struct hci_dev *hdev = data->hdev;
+
+ if (!btusb_intel_idle_power_manageable(data))
+ return true;
+
+ if (hci_conn_count(hdev))
+ return true;
+
+ if (hdev->discovery.state == DISCOVERY_FINDING ||
+ hdev->discovery.state == DISCOVERY_RESOLVING)
+ return true;
+
+ if (hci_dev_test_flag(hdev, HCI_LE_SCAN) ||
+ hci_dev_test_flag(hdev, HCI_LE_ADV) ||
+ hci_dev_test_flag(hdev, HCI_ADVERTISING) ||
+ test_bit(HCI_PSCAN, &hdev->flags) ||
+ test_bit(HCI_ISCAN, &hdev->flags))
+ return true;
+
+ return false;
+}
+
static int btusb_close(struct hci_dev *hdev)
{
struct btusb_data *data = hci_get_drvdata(hdev);
@@ -4041,6 +4085,8 @@ static int btusb_probe(struct usb_interface *intf,
data->recv_bulk = btusb_recv_bulk;
if (id->driver_info & BTUSB_INTEL_COMBINED) {
+ data->intel_combined = true;
+

I am not clear about this hunk. Could you please elaborate?

/* Allocate extra space for Intel device */
priv_size += sizeof(struct btintel_data);


Kind regards,

Paul


PS: Thank you so much for working on such details! Do you know, if this is also a problem on Google Chromebooks/-boxes?


[1]: https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git/log/?h=usb-next&qt=author&q=rhodes
[2]: https://lore.kernel.org/all/20260315223433.23452-1-sean@starlabs.systems/