Thread (14 messages) flat view 14 messages, 4 authors, 2022-01-06

Re: [RFC PATCH v2] Bluetooth: btintel: Fix broken LED quirk for legacy ROM devices

From: Luiz Augusto von Dentz <luiz.dentz@gmail.com>
Date: 2021-12-17 19:31:40

Hi Tedd,

On Thu, Dec 16, 2021 at 6:47 PM An, Tedd [off-list ref] wrote:
Hi Luiz,

On Thu, 2021-12-16 at 16:47 -0800, Luiz Augusto von Dentz wrote:
quoted
Hi Tedd,

On Thu, Dec 16, 2021 at 4:04 PM Tedd Ho-Jeong An [off-list ref] wrote:
quoted
From: Tedd Ho-Jeong An <redacted>

This patch fixes the broken LED quirk for Intel legacy ROM devices.
To fix the LED issue that doesn't turn off immediately, the host sends
the SW RFKILL command while shutting down the interface and it puts the
devices in an asserted state.

Once the device is in SW RFKILL state, it can only accept HCI_Reset to
exit from the SW RFKILL state. This patch checks the quirk and sends the
HCI_Reset before sending the HCI_Intel_Read_Version command.

The affected legacy ROM devices are
 - 8087:0a2a
 - 8087:0aa7

fixes: ffcba827c0a1d ("Bluetooth: btintel: Fix the LED is not turning off immediately")

Signed-off-by: Tedd Ho-Jeong An <redacted>
---
 drivers/bluetooth/btintel.c | 13 ++++++-------
 drivers/bluetooth/btusb.c   | 10 ++++++++--
 2 files changed, 14 insertions(+), 9 deletions(-)
diff --git a/drivers/bluetooth/btintel.c b/drivers/bluetooth/btintel.c
index e1f96df847b8..75f8d7aceb35 100644
--- a/drivers/bluetooth/btintel.c
+++ b/drivers/bluetooth/btintel.c
@@ -2355,8 +2355,13 @@ static int btintel_setup_combined(struct hci_dev *hdev)
         * As a workaround, send HCI Reset command first which will reset the
         * number of completed commands and allow normal command processing
         * from now on.
+        *
+        * For INTEL_BROKEN_LED, these devices have an issue with LED which
+        * doesn't go off immediately during shutdown. Set the flag here to send
+        * the LED OFF command during shutdown.
         */
-       if (btintel_test_flag(hdev, INTEL_BROKEN_INITIAL_NCMD)) {
+       if (btintel_test_flag(hdev, INTEL_BROKEN_INITIAL_NCMD) ||
+                               btintel_test_flag(hdev, INTEL_BROKEN_LED)) {
                skb = __hci_cmd_sync(hdev, HCI_OP_RESET, 0, NULL,
                                     HCI_INIT_TIMEOUT);
                if (IS_ERR(skb)) {
@@ -2428,12 +2433,6 @@ static int btintel_setup_combined(struct hci_dev *hdev)
                                set_bit(HCI_QUIRK_WIDEBAND_SPEECH_SUPPORTED,
                                        &hdev->quirks);

-                       /* These devices have an issue with LED which doesn't
-                        * go off immediately during shutdown. Set the flag
-                        * here to send the LED OFF command during shutdown.
-                        */
-                       btintel_set_flag(hdev, INTEL_BROKEN_LED);
-
                        err = btintel_legacy_rom_setup(hdev, &ver);
                        break;
                case 0x0b:      /* SfP */
diff --git a/drivers/bluetooth/btusb.c b/drivers/bluetooth/btusb.c
index d1bd9ee0a6ab..c6a070d5284f 100644
--- a/drivers/bluetooth/btusb.c
+++ b/drivers/bluetooth/btusb.c
@@ -60,6 +60,7 @@ static struct usb_driver btusb_driver;
 #define BTUSB_WIDEBAND_SPEECH  0x400000
 #define BTUSB_VALID_LE_STATES   0x800000
 #define BTUSB_QCA_WCN6855      0x1000000
+#define BTUSB_INTEL_BROKEN_LED 0x2000000
I wonder why we haven't been using BIT macro here and did we make a
mistake and leave one bit behind? Or something else was at this bit
position?
There used to be a flag BTUSB_INTEL_NEWGEN, and it was removed while
refactoring the btintel, and the NCMD was added before removing the
NEWGEN flag.
Ok that explains it then, luckily this is just used internally by the
driver so it shouldn't break anything if we reuse, but I'd still think
using BIT macro is clear here.
quoted
quoted
 #define BTUSB_INTEL_BROKEN_INITIAL_NCMD 0x4000000

 static const struct usb_device_id btusb_table[] = {
@@ -382,9 +383,11 @@ static const struct usb_device_id blacklist_table[] = {
        { USB_DEVICE(0x8087, 0x07da), .driver_info = BTUSB_CSR },
        { USB_DEVICE(0x8087, 0x07dc), .driver_info = BTUSB_INTEL_COMBINED |
                                                     BTUSB_INTEL_BROKEN_INITIAL_NCMD },
-       { USB_DEVICE(0x8087, 0x0a2a), .driver_info = BTUSB_INTEL_COMBINED },
+       { USB_DEVICE(0x8087, 0x0a2a), .driver_info = BTUSB_INTEL_COMBINED |
+                                                    BTUSB_INTEL_BROKEN_LED },
        { USB_DEVICE(0x8087, 0x0a2b), .driver_info = BTUSB_INTEL_COMBINED },
-       { USB_DEVICE(0x8087, 0x0aa7), .driver_info = BTUSB_INTEL_COMBINED },
+       { USB_DEVICE(0x8087, 0x0aa7), .driver_info = BTUSB_INTEL_COMBINED |
+                                                    BTUSB_INTEL_BROKEN_LED },
        { USB_DEVICE(0x8087, 0x0aaa), .driver_info = BTUSB_INTEL_COMBINED },

        /* Other Intel Bluetooth devices */
@@ -3724,6 +3727,9 @@ static int btusb_probe(struct usb_interface *intf,

                if (id->driver_info & BTUSB_INTEL_BROKEN_INITIAL_NCMD)
                        btintel_set_flag(hdev, INTEL_BROKEN_INITIAL_NCMD);
+
+               if (id->driver_info & BTUSB_INTEL_BROKEN_LED)
+                       btintel_set_flag(hdev, INTEL_BROKEN_LED);
I wonder if wouldn't be simples to have a flag e.g. INTEL_NEED_RESET
instead of replicating the quirks as flags since in either case the
actual outcome is to send a reset.
The result is same though, I think it is better to be more specific.
I am fine with it also.
Fair enough.
quoted
quoted
        }

        if (id->driver_info & BTUSB_MARVELL)
--
2.25.1

-- 
Luiz Augusto von Dentz
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help