Linux kernel -stable discussions
 help / color / mirror / Atom feed
* [PATCH v2] Bluetooth: hci_serdev: stop write_work before closing the port
@ 2026-11-20  6:20 Zhao Li
  2026-11-20 13:39 ` Loic Poulain
  0 siblings, 1 reply; 3+ messages in thread
From: Zhao Li @ 2026-11-20  6:20 UTC (permalink / raw)
  To: Marcel Holtmann, Luiz Augusto von Dentz
  Cc: linux-bluetooth, linux-kernel, Michael McAllister, stable,
	Zhao Li

With HCI_QUIRK_NON_PERSISTENT_SETUP, hci_uart_close() clears
HCI_UART_PROTO_READY without hu->proto_lock and closes the serdev port
without waiting for write_work. The hci_uart_flush() just before it
re-enables write_work and may replay a recorded wakeup, and
hci_uart_tx_wakeup() queues it from any caller that passed the
HCI_UART_PROTO_READY test, such as the serdev write_wakeup callback.
write_work can then call ttyport_write_buf() while ttyport_close()
releases the tty, so it writes through a released tty.

A partial frame that write_work leaves in hu->tx_skb also survives into
the next open, which flushes nothing, and would be sent first to the
controller after it is powered on again.

Quiesce the transmit path as hci_uart_unregister_device() does: clear
HCI_UART_PROTO_READY under percpu_down_write() so that no wakeup is in
flight, cancel write_work, free the leftover frame, then close the port.
A wakeup sets HCI_UART_SENDING and HCI_UART_TX_WAKEUP before it queues
write_work, so clear both after the cancel: a work cancelled while
still pending would leave SENDING set, and the next open, which reuses
hu, would never queue write_work again.

Found by code review. Compile-tested with W=1.

Fixes: de8892df72be ("Bluetooth: hci_serdev: Close UART port if NON_PERSISTENT_SETUP is set")
Cc: stable@vger.kernel.org
Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
---
v2:
- Clear HCI_UART_SENDING and HCI_UART_TX_WAKEUP after cancelling
  write_work. A pending write_work cancelled there left SENDING set, so
  the next power-on never queued write_work (Sashiko's review of v1).
v1: https://lore.kernel.org/linux-bluetooth/20261118124821.3990-1-enderaoelyther@gmail.com/

Before 6.16 the quirk test in hci_uart_close() reads
test_bit(HCI_QUIRK_NON_PERSISTENT_SETUP, &hdev->quirks); with that
context line the patch applies to 6.12.y, 6.6.y, 6.1.y, 5.15.y and
5.10.y. It applies as is to 6.18.y and 7.3.y.

No new W=1, sparse or smatch warnings.

AI assistance was used.

 drivers/bluetooth/hci_serdev.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/bluetooth/hci_serdev.c b/drivers/bluetooth/hci_serdev.c
index 194f11251cc5..a5e4b3d32144 100644
--- a/drivers/bluetooth/hci_serdev.c
+++ b/drivers/bluetooth/hci_serdev.c
@@ -164,7 +164,20 @@ static int hci_uart_close(struct hci_dev *hdev)
 	 * open may drain the battery.
 	 */
 	if (hci_test_quirk(hdev, HCI_QUIRK_NON_PERSISTENT_SETUP)) {
+		percpu_down_write(&hu->proto_lock);
 		clear_bit(HCI_UART_PROTO_READY, &hu->flags);
+		percpu_up_write(&hu->proto_lock);
+
+		cancel_work_sync(&hu->write_work);
+		clear_bit(HCI_UART_SENDING, &hu->tx_state);
+		clear_bit(HCI_UART_TX_WAKEUP, &hu->tx_state);
+
+		/* Do not send a stale partial frame after the next open. */
+		if (hu->tx_skb) {
+			kfree_skb(hu->tx_skb);
+			hu->tx_skb = NULL;
+		}
+
 		serdev_device_close(hu->serdev);
 	}
 

base-commit: 1d79f61152c315e30e758116e525871732181a98
-- 
2.56.1


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH v2] Bluetooth: hci_serdev: stop write_work before closing the port
  2026-11-20  6:20 [PATCH v2] Bluetooth: hci_serdev: stop write_work before closing the port Zhao Li
@ 2026-11-20 13:39 ` Loic Poulain
  2026-11-21 12:28   ` Zhao Li
  0 siblings, 1 reply; 3+ messages in thread
From: Loic Poulain @ 2026-11-20 13:39 UTC (permalink / raw)
  To: Zhao Li
  Cc: Marcel Holtmann, Luiz Augusto von Dentz, linux-bluetooth,
	linux-kernel, Michael McAllister, stable

On Fri, Nov 20, 2026 at 7:21 AM Zhao Li <enderaoelyther@gmail.com> wrote:
>
> With HCI_QUIRK_NON_PERSISTENT_SETUP, hci_uart_close() clears
> HCI_UART_PROTO_READY without hu->proto_lock and closes the serdev port
> without waiting for write_work. The hci_uart_flush() just before it
> re-enables write_work and may replay a recorded wakeup, and
> hci_uart_tx_wakeup() queues it from any caller that passed the
> HCI_UART_PROTO_READY test, such as the serdev write_wakeup callback.
> write_work can then call ttyport_write_buf() while ttyport_close()
> releases the tty, so it writes through a released tty.
>
> A partial frame that write_work leaves in hu->tx_skb also survives into
> the next open, which flushes nothing, and would be sent first to the
> controller after it is powered on again.
>
> Quiesce the transmit path as hci_uart_unregister_device() does: clear
> HCI_UART_PROTO_READY under percpu_down_write() so that no wakeup is in
> flight, cancel write_work, free the leftover frame, then close the port.
> A wakeup sets HCI_UART_SENDING and HCI_UART_TX_WAKEUP before it queues
> write_work, so clear both after the cancel: a work cancelled while
> still pending would leave SENDING set, and the next open, which reuses
> hu, would never queue write_work again.
>
> Found by code review. Compile-tested with W=1.
>
> Fixes: de8892df72be ("Bluetooth: hci_serdev: Close UART port if NON_PERSISTENT_SETUP is set")
> Cc: stable@vger.kernel.org
> Signed-off-by: Zhao Li <enderaoelyther@gmail.com>
> ---
> v2:
> - Clear HCI_UART_SENDING and HCI_UART_TX_WAKEUP after cancelling
>   write_work. A pending write_work cancelled there left SENDING set, so
>   the next power-on never queued write_work (Sashiko's review of v1).
> v1: https://lore.kernel.org/linux-bluetooth/20261118124821.3990-1-enderaoelyther@gmail.com/
>
> Before 6.16 the quirk test in hci_uart_close() reads
> test_bit(HCI_QUIRK_NON_PERSISTENT_SETUP, &hdev->quirks); with that
> context line the patch applies to 6.12.y, 6.6.y, 6.1.y, 5.15.y and
> 5.10.y. It applies as is to 6.18.y and 7.3.y.
>
> No new W=1, sparse or smatch warnings.
>
> AI assistance was used.
>
>  drivers/bluetooth/hci_serdev.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
>
> diff --git a/drivers/bluetooth/hci_serdev.c b/drivers/bluetooth/hci_serdev.c
> index 194f11251cc5..a5e4b3d32144 100644
> --- a/drivers/bluetooth/hci_serdev.c
> +++ b/drivers/bluetooth/hci_serdev.c
> @@ -164,7 +164,20 @@ static int hci_uart_close(struct hci_dev *hdev)
>          * open may drain the battery.
>          */
>         if (hci_test_quirk(hdev, HCI_QUIRK_NON_PERSISTENT_SETUP)) {
> +               percpu_down_write(&hu->proto_lock);
>                 clear_bit(HCI_UART_PROTO_READY, &hu->flags);
> +               percpu_up_write(&hu->proto_lock);
> +
> +               cancel_work_sync(&hu->write_work);
> +               clear_bit(HCI_UART_SENDING, &hu->tx_state);
> +               clear_bit(HCI_UART_TX_WAKEUP, &hu->tx_state);

Looks like you are mostly re-doing what has already been done in
hci_uart_flush()...
Wouldn't it be simpler to have a hci_uart_flush parameter to not
reschedule tx on flush?

> +
> +               /* Do not send a stale partial frame after the next open. */
> +               if (hu->tx_skb) {
> +                       kfree_skb(hu->tx_skb);
> +                       hu->tx_skb = NULL;
> +               }
> +
>                 serdev_device_close(hu->serdev);
>         }
>
>
> base-commit: 1d79f61152c315e30e758116e525871732181a98
> --
> 2.56.1
>
>

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH v2] Bluetooth: hci_serdev: stop write_work before closing the port
  2026-11-20 13:39 ` Loic Poulain
@ 2026-11-21 12:28   ` Zhao Li
  0 siblings, 0 replies; 3+ messages in thread
From: Zhao Li @ 2026-11-21 12:28 UTC (permalink / raw)
  To: Loic Poulain
  Cc: Marcel Holtmann, Luiz Augusto von Dentz, linux-bluetooth,
	linux-kernel, Michael McAllister, stable

On Fri, Nov 20, 2026 at 02:39:41PM +0100, Loic Poulain wrote:
> > +               cancel_work_sync(&hu->write_work);
> > +               clear_bit(HCI_UART_SENDING, &hu->tx_state);
> > +               clear_bit(HCI_UART_TX_WAKEUP, &hu->tx_state);
>
> Looks like you are mostly re-doing what has already been done in
> hci_uart_flush()...
> Wouldn't it be simpler to have a hci_uart_flush parameter to not
> reschedule tx on flush?

Good point. The only reason hci_uart_close() has to stop TX a second
time is that hci_uart_flush() unconditionally re-enables write_work at
the end.

Skipping just the final wakeup replay wouldn't be enough, though. Once
hci_uart_flush() re-enables write_work and clears HCI_UART_SENDING, a
serdev write_wakeup callback can still queue write_work before
serdev_device_close() runs.

Instead, we can split out a __hci_uart_flush(hdev, bool resume_tx)
helper (since hci_uart_flush() itself is the hdev->flush callback):

- Normal flushes pass resume_tx = true and behave as today.
- With HCI_QUIRK_NON_PERSISTENT_SETUP, hci_uart_close() passes
  resume_tx = false, leaving write_work disabled and HCI_UART_SENDING
  set across the close.
- After clearing HCI_UART_PROTO_READY under hu->proto_lock (to drain
  in-flight wakeups) and calling serdev_device_close(), hci_uart_close()
  clears HCI_UART_TX_WAKEUP, re-enables write_work, and clears
  HCI_UART_SENDING for the next open.

That drops the duplicate cancel_work_sync() and tx_skb cleanup while
keeping write_work completely disabled until the port is closed.

I'll send a v3 along these lines.

Thanks,
Zhao

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-11-21 12:28 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-20  6:20 [PATCH v2] Bluetooth: hci_serdev: stop write_work before closing the port Zhao Li
2026-11-20 13:39 ` Loic Poulain
2026-11-21 12:28   ` Zhao Li

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox