[PATCH] Bluetooth: btqcomsmd: Fix command timeout after setting BD address

[PATCH] Bluetooth: btqcomsmd: Fix command timeout after setting BD address

[Stephan Gerhold]

The Bluetooth firmware seems to become unresponsive for a while after
setting the BD address. At least on recent kernel versions this often
causes timeouts for subsequent commands, e.g. the HCI reset sent by
the Bluetooth core during initialization:

	Bluetooth: hci0: Opcode 0x c03 failed: -110

Unfortunately this behavior does not seem to be documented anywhere.
Experimentation suggests that the minimum necessary delay to avoid
the problem is ~150us. However, to be sure add a sleep for > 1ms
in case it is a bit longer on other firmware versions.

Fixes: 1511cc750c3d ("Bluetooth: Introduce Qualcomm WCNSS SMD based HCI driver")
Signed-off-by: Stephan Gerhold <stephan.gerhold@kernkonzept.com>
---
I tested this using a script that reboots repeatedly and checks for the
error. With this patch, BT shows up successfully for 100+ consecutive
boots. Without this patch it usually fails after 1-5 boots (or even
always on some boards).
---
 drivers/bluetooth/btqcomsmd.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

diff --git a/drivers/bluetooth/btqcomsmd.c b/drivers/bluetooth/btqcomsmd.c
index 2acb719e596f..11c7e04bf394 100644
--- a/drivers/bluetooth/btqcomsmd.c
+++ b/drivers/bluetooth/btqcomsmd.c
@@ -122,6 +122,21 @@ static int btqcomsmd_setup(struct hci_dev *hdev)
 	return 0;
 }
 
+static int btqcomsmd_set_bdaddr(struct hci_dev *hdev, const bdaddr_t *bdaddr)
+{
+	int ret;
+
+	ret = qca_set_bdaddr_rome(hdev, bdaddr);
+	if (ret)
+		return ret;
+
+	/* The firmware stops responding for a while after setting the bdaddr,
+	 * causing timeouts for subsequent commands. Sleep a bit to avoid this.
+	 */
+	usleep_range(1000, 10000);
+	return 0;
+}
+
 static int btqcomsmd_probe(struct platform_device *pdev)
 {
 	struct btqcomsmd *btq;
@@ -162,7 +177,7 @@ static int btqcomsmd_probe(struct platform_device *pdev)
 	hdev->close = btqcomsmd_close;
 	hdev->send = btqcomsmd_send;
 	hdev->setup = btqcomsmd_setup;
-	hdev->set_bdaddr = qca_set_bdaddr_rome;
+	hdev->set_bdaddr = btqcomsmd_set_bdaddr;
 
 	ret = hci_register_dev(hdev);
 	if (ret < 0)
-- 
2.30.2

Re: [PATCH] Bluetooth: btqcomsmd: Fix command timeout after setting BD address

[Paul Menzel]

Dear Stephan,


Thank you for the patch.

Am 13.06.22 um 13:07 schrieb Stephan Gerhold:
> The Bluetooth firmware seems to become unresponsive for a while after

Please mention the device with the problematic firmware.

> setting the BD address. At least on recent kernel versions this often

The commit referenced in the Fixes tag is in Linux 4.9. Can you please 
name the oldest Linux kernel version you experienced the problem with.

> causes timeouts for subsequent commands, e.g. the HCI reset sent by
> the Bluetooth core during initialization:
> 
> 	Bluetooth: hci0: Opcode 0x c03 failed: -110
> 
> Unfortunately this behavior does not seem to be documented anywhere.
> Experimentation suggests that the minimum necessary delay to avoid
> the problem is ~150us. However, to be sure add a sleep for > 1ms
> in case it is a bit longer on other firmware versions.
> 
> Fixes: 1511cc750c3d ("Bluetooth: Introduce Qualcomm WCNSS SMD based HCI driver")
> Signed-off-by: Stephan Gerhold <stephan.gerhold@kernkonzept.com>
> ---
> I tested this using a script that reboots repeatedly and checks for the
> error. With this patch, BT shows up successfully for 100+ consecutive
> boots. Without this patch it usually fails after 1-5 boots (or even
> always on some boards).
> ---
>   drivers/bluetooth/btqcomsmd.c | 17 ++++++++++++++++-
>   1 file changed, 16 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/bluetooth/btqcomsmd.c b/drivers/bluetooth/btqcomsmd.c
> index 2acb719e596f..11c7e04bf394 100644
> --- a/drivers/bluetooth/btqcomsmd.c
> +++ b/drivers/bluetooth/btqcomsmd.c
> @@ -122,6 +122,21 @@ static int btqcomsmd_setup(struct hci_dev *hdev)
>   	return 0;
>   }
>   
> +static int btqcomsmd_set_bdaddr(struct hci_dev *hdev, const bdaddr_t *bdaddr)
> +{
> +	int ret;
> +
> +	ret = qca_set_bdaddr_rome(hdev, bdaddr);
> +	if (ret)
> +		return ret;
> +
> +	/* The firmware stops responding for a while after setting the bdaddr,
> +	 * causing timeouts for subsequent commands. Sleep a bit to avoid this.
> +	 */
> +	usleep_range(1000, 10000);
> +	return 0;
> +}
> +
>   static int btqcomsmd_probe(struct platform_device *pdev)
>   {
>   	struct btqcomsmd *btq;
> @@ -162,7 +177,7 @@ static int btqcomsmd_probe(struct platform_device *pdev)
>   	hdev->close = btqcomsmd_close;
>   	hdev->send = btqcomsmd_send;
>   	hdev->setup = btqcomsmd_setup;
> -	hdev->set_bdaddr = qca_set_bdaddr_rome;
> +	hdev->set_bdaddr = btqcomsmd_set_bdaddr;
>   
>   	ret = hci_register_dev(hdev);
>   	if (ret < 0)


Kind regards,

Paul

Re: [PATCH] Bluetooth: btqcomsmd: Fix command timeout after setting BD address

[Stephan Gerhold]

Hi Paul,

On Mon, Jun 13, 2022 at 06:09:33PM +0200, Paul Menzel wrote:
> Am 13.06.22 um 13:07 schrieb Stephan Gerhold:
> > The Bluetooth firmware seems to become unresponsive for a while after
> 
> Please mention the device with the problematic firmware.
> 

Will add a comment in v2. It seems to affect pretty much all devices
that make use of the "btqcomsmd" driver.

> > setting the BD address. At least on recent kernel versions this often
> 
> The commit referenced in the Fixes tag is in Linux 4.9. Can you please name
> the oldest Linux kernel version you experienced the problem with.
> 

This is quite hard to say. It definitely became more apparent in the
last few months (so at least 5.17+). But since it's a timing problem
it's influenced by various side effects. For example, it is less likely
to happen the more hardware functionality you enable on a board (because
other interrupts come in and might cause the necessary delay).

I believe that the problem also exists in some form on Linux 4.9,
even though it might be harder to trigger it there.

Thanks,
Stephan