btbcm issue

btbcm issue

[Ryan]

Hi,

I am using btbcm on an embedded device which has a broadcom bt module.

I understand that the below code suspends the bt device. Can anyone show
me where i need to specify wakeup irq and device_wakeup gpios?

I cant find examples of boards which do these? I understand that device_wakeup
is required for suspending the device.

Below is the code segment for suspend-resume which i suppose will be
called to suspend the device.


+static int bcm_suspend(struct device *pdev)
+{
+ struct bcm_device *dev = platform_get_drvdata(to_platform_device(pdev));
+
+ BT_DBG("suspend (%p): is_suspended %d", dev, dev->is_suspended);
+
+ if (!dev->is_suspended) {
+ hci_uart_set_flow_control(dev->hu, true);
+
+ /* Once this callback returns, driver suspends BT via GPIO */
+ dev->is_suspended = true;
+ }
+
+ /* Suspend the device */
+ if (dev->device_wakeup) {
+ gpiod_set_value(dev->device_wakeup, false);
+ BT_DBG("suspend, delaying 15 ms");
+ mdelay(15);
+ }
+
+ return 0;
+}

Re: btbcm issue

[Marcel Holtmann]

Hi Ryan,

> I am using btbcm on an embedded device which has a broadcom bt module.
> 
> I understand that the below code suspends the bt device. Can anyone show
> me where i need to specify wakeup irq and device_wakeup gpios?
> 
> I cant find examples of boards which do these? I understand that device_wakeup
> is required for suspending the device.
> 
> Below is the code segment for suspend-resume which i suppose will be
> called to suspend the device.
> 
> 
> +static int bcm_suspend(struct device *pdev)
> +{
> + struct bcm_device *dev = platform_get_drvdata(to_platform_device(pdev));
> +
> + BT_DBG("suspend (%p): is_suspended %d", dev, dev->is_suspended);
> +
> + if (!dev->is_suspended) {
> + hci_uart_set_flow_control(dev->hu, true);
> +
> + /* Once this callback returns, driver suspends BT via GPIO */
> + dev->is_suspended = true;
> + }
> +
> + /* Suspend the device */
> + if (dev->device_wakeup) {
> + gpiod_set_value(dev->device_wakeup, false);
> + BT_DBG("suspend, delaying 15 ms");
> + mdelay(15);
> + }
> +
> + return 0;
> +}

I think the current hci_bcm.c (and btbcm.c) only ever enabled ACPI enumeration. Meaning someone might want to update that to support device tree (DT). There were efforts from Broadcom to add support for this, but they never agreed with the DT maintainer on the definitions.

Regards

Marcel

Re: btbcm issue

[Marcel Holtmann]

Hi Ryan,

> I wanted to check whats the best way of going about adding
> suspend-resume support for bcm bt chip
> 
> Should one have a separate driver
> (https://android.googlesource.com/kernel/bcm/+/3d6d35972004824a9111f13926bed2a4a495b668/drivers/misc/bcm-bt-lpm.c)
> 
> or should i use the existing hci_bcm.c driver & add dt support or use
> a platform device to pass the gpiod_desc for wakeup/sleep?

no separate driver please. As soon as btattach links the UART and we enable the device, everything should be done in background by the driver. The Broadcom LPM driver is just a hack for Android since they still haven't switched to using HCI User Channel for Android (like Intel for example).

In general DT and ACPI are the same. They describe the hardware and there are methods to read these descriptions from them and then use them. The only problem with DT was that Broadcom never got the definition acked by the DT maintainer. Otherwise all the functionality should be already in hci_bcm.c right now.

Regards

Marcel

Re: btbcm issue

[Ryan]

Hi Marcel,

Is this what you are referring to?

http://www.spinics.net/lists/linux-bluetooth/msg61742.html

In this patchset, bcm tried to create new functionality/files for uart and =
bcm.

If not, can you point me to the right set of patches?

Regards,
ryan

On Tue, Aug 9, 2016 at 1:07 PM, Marcel Holtmann <marcel@holtmann.org> wrote=
:
> Hi Ryan,
>
>> I wanted to check whats the best way of going about adding
>> suspend-resume support for bcm bt chip
>>
>> Should one have a separate driver
>> (https://android.googlesource.com/kernel/bcm/+/3d6d35972004824a9111f1392=
6bed2a4a495b668/drivers/misc/bcm-bt-lpm.c)
>>
>> or should i use the existing hci_bcm.c driver & add dt support or use
>> a platform device to pass the gpiod_desc for wakeup/sleep?
>
> no separate driver please. As soon as btattach links the UART and we enab=
le the device, everything should be done in background by the driver. The B=
roadcom LPM driver is just a hack for Android since they still haven't swit=
ched to using HCI User Channel for Android (like Intel for example).
>
> In general DT and ACPI are the same. They describe the hardware and there=
 are methods to read these descriptions from them and then use them. The on=
ly problem with DT was that Broadcom never got the definition acked by the =
DT maintainer. Otherwise all the functionality should be already in hci_bcm=
.c right now.
>
> Regards
>
> Marcel
>