RE: [PATCH v4 2/2] Bluetooth: btusb: Support Bluetooth Reset for Mediatek Chip(MT7921)
From: Mark-YW Chen (陳揚文)
Date: Tue Jul 27 2021 - 11:16:52 EST
> And we need to work towards abstracting the vendor specific position out of btusb.c (and I include the Intel part here as well). The btusb.c is getting overloaded with vendor stuff.
I agree with your comment, The vendor should create the vendor.c for their specific behavior.
First, I add the prefix into these patches.
After these patches merged, we will create the btmtk.c to handle Mediatek stuff.
With best regards,
From: Marcel Holtmann [mailto:marcel@xxxxxxxxxxxx]
Sent: Thursday, July 22, 2021 10:04 PM
To: Mark-YW Chen (陳揚文)
Cc: Johan Hedberg; Chris Lu (陸稚泓); Will-CY Lee (李政穎); Sean Wang; Bluetooth Kernel Mailing List; linux-mediatek@xxxxxxxxxxxxxxxxxxx; open list; michaelfsun@xxxxxxxxxx; shawnku@xxxxxxxxxx; jemele@xxxxxxxxxx; apusaka@xxxxxxxxxx
Subject: Re: [PATCH v4 2/2] Bluetooth: btusb: Support Bluetooth Reset for Mediatek Chip(MT7921)
> When the firmware hang or command no response, driver can reset the
> bluetooth mcu via USB to recovery it. The reset steps as follows.
> 1. Cancel USB transfer requests before reset.
> 2. It use speicific USB HW Register to reset Bluetooth MCU, at the
> same time, the USB Endpoint0 still keep alive.
> 3. Poll the USB HW register until reset is completed by Endpoint0.
> 4. To recovery unexpected USB state and behavior during resetting the
> Bluetooth MCU, the driver need to reset the USB device for MT7921.
> 5. After the reset is completed, the Bluetooth MCU need to re-setup,
> such as download patch, power-on sequence and etc.
> Signed-off-by: mark-yw.chen <mark-yw.chen@xxxxxxxxxxxx>
> drivers/bluetooth/btusb.c | 127 ++++++++++++++++++++++++++++++++++++++
> 1 file changed, 127 insertions(+)
> diff --git a/drivers/bluetooth/btusb.c b/drivers/bluetooth/btusb.c
> index b1249b664981..122a928ebcb5 100644
> --- a/drivers/bluetooth/btusb.c
> +++ b/drivers/bluetooth/btusb.c
> @@ -3076,6 +3076,17 @@ static int btusb_shutdown_intel_new(struct hci_dev *hdev)
> return 0;
> +/* UHW CR mapping */
> +#define BT_MISC 0x70002510
> +#define BT_SUBSYS_RST 0x70002610
> +#define UDMA_INT_STA_BT 0x74000024
> +#define UDMA_INT_STA_BT1 0x74000308
> +#define BT_WDT_STATUS 0x740003A0
> +#define EP_RST_OPT 0x74011890
> +#define EP_RST_IN_OUT_OPT 0x00010001
> +#define BT_RST_DONE 0x00000100
> +#define BT_RESET_WAIT_MS 100
> +#define BT_RESET_NUM_TRIES 10
you need to prefix these with MTK since otherwise we get confused.
And we need to work towards abstracting the vendor specific position out of btusb.c (and I include the Intel part here as well). The btusb.c is getting overloaded with vendor stuff.