The below issues are found in our recent testing.
1. Observed device is not going into off state or not responding.
As wcn3990 require a power pulses to turn on the irrespctive of
igniting regulators, it was observed that power on or power off
pulses are not in sync with respective to chip.
The below patch will help us to wait until byte is pushed on to wires.
* Bluetooth: hci_qca: use wait_until_sent() for power pulses
2. Observed Chip responding when we are in sleep.
This is due to turn on flow control during change baudrate request.
The below patch will only pull the RTS line high instead of turning off
the flow.
* Bluetooth: hci_qca: Pull RTS line high for baudrate change command.
3. Disable IBS state machine and flush Tx buffer
We are disabling IBS and flushing the Tx buffer before turning off the chip.
This is due to IBS state machine is active when we turn off the chip.
This will stop queuing IBS protocol bytes.
* Bluetooth: hci_qca: Disable IBS state machine and flush Tx buffer
Changes in v8:
* dropped inject the command complete event.
* added one second time for the power pules instead of the indefinite time.
Changes in v7:
* dropped frame reassmebly error patch.
* dropped baudrate change wait time patch.
* increased a wait to 5 ms for power pulses.
Changes in v6:
* added serdev_device_write_flush in qca_send_power_pulse().
* added new patch to update the baudrate change wait time.
Changes in V5:
* added serdev_device_write_flush before sending the power off pulse
during shutdown.
Changes in v4:
* used serdev_device_write_buf() instead of serdev_device_write().
* added new patch to stop logging of 0xfc00 timeout on console.
Changes in v3:
* moved IBS & qca_flush to different patch
* updated comments in code fo Deassert RTS patch
Balakrishna Godavarthi (3):
Bluetooth: hci_qca: use wait_until_sent() for power pulses
Bluetooth: hci_qca: Deassert RTS while baudrate change command
Bluetooth: hci_qca: Disable IBS state machine and flush Tx buffer
drivers/bluetooth/hci_qca.c | 76 +++++++++++++++++++------------------
1 file changed, 40 insertions(+), 36 deletions(-)
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
During hci down we observed IBS sleep commands are queued in the Tx
buffer and hci_uart_write_work is sending data to the chip which is
not required as the chip is powered off. This patch will disable IBS
and flush the Tx buffer before we turn off the chip.
Signed-off-by: Balakrishna Godavarthi <[email protected]>
---
drivers/bluetooth/hci_qca.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
index 7e4afcf40da2..7330ba71ada4 100644
--- a/drivers/bluetooth/hci_qca.c
+++ b/drivers/bluetooth/hci_qca.c
@@ -1275,6 +1275,14 @@ static const struct qca_vreg_data qca_soc_data = {
static void qca_power_shutdown(struct hci_uart *hu)
{
+ struct qca_data *qca = hu->priv;
+
+ /* From this point we go into power off state. But serial port is
+ * still open, stop queueing the IBS data and flush all the buffered
+ * data in skb's.
+ */
+ clear_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags);
+ qca_flush(hu);
host_set_baudrate(hu, 2400);
qca_send_power_pulse(hu, QCA_WCN3990_POWEROFF_PULSE);
qca_power_setup(hu, false);
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
This patch will help to stop frame reassembly errors while changing
the baudrate. This is because host send a change baudrate request
command to the chip with 115200 bps, Whereas chip will change their
UART clocks to the enable for new baudrate and sends the response
for the change request command with newer baudrate, On host side
we are still operating in 115200 bps which results of reading garbage
data. Here we are pulling RTS line, so that chip we will wait to send data
to host until host change its baudrate.
Signed-off-by: Balakrishna Godavarthi <[email protected]>
Tested-by: Matthias Kaehlcke <[email protected]>
Reviewed-by: Matthias Kaehlcke <[email protected]>
---
Note: will keep this change for now. If any change required after further
discussions i will update accordingly.
---
drivers/bluetooth/hci_qca.c | 24 +++++++++++++-----------
1 file changed, 13 insertions(+), 11 deletions(-)
diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
index 681bfa30467e..7e4afcf40da2 100644
--- a/drivers/bluetooth/hci_qca.c
+++ b/drivers/bluetooth/hci_qca.c
@@ -964,7 +964,6 @@ static int qca_set_baudrate(struct hci_dev *hdev, uint8_t baudrate)
struct hci_uart *hu = hci_get_drvdata(hdev);
struct qca_data *qca = hu->priv;
struct sk_buff *skb;
- struct qca_serdev *qcadev;
u8 cmd[] = { 0x01, 0x48, 0xFC, 0x01, 0x00 };
if (baudrate > QCA_BAUDRATE_3200000)
@@ -978,13 +977,6 @@ static int qca_set_baudrate(struct hci_dev *hdev, uint8_t baudrate)
return -ENOMEM;
}
- /* Disabling hardware flow control is mandatory while
- * sending change baudrate request to wcn3990 SoC.
- */
- qcadev = serdev_device_get_drvdata(hu->serdev);
- if (qcadev->btsoc_type == QCA_WCN3990)
- hci_uart_set_flow_control(hu, true);
-
/* Assign commands to change baudrate and packet type. */
skb_put_data(skb, cmd, sizeof(cmd));
hci_skb_pkt_type(skb) = HCI_COMMAND_PKT;
@@ -1000,9 +992,6 @@ static int qca_set_baudrate(struct hci_dev *hdev, uint8_t baudrate)
schedule_timeout(msecs_to_jiffies(BAUDRATE_SETTLE_TIMEOUT_MS));
set_current_state(TASK_RUNNING);
- if (qcadev->btsoc_type == QCA_WCN3990)
- hci_uart_set_flow_control(hu, false);
-
return 0;
}
@@ -1091,6 +1080,7 @@ static int qca_check_speeds(struct hci_uart *hu)
static int qca_set_speed(struct hci_uart *hu, enum qca_speed_type speed_type)
{
unsigned int speed, qca_baudrate;
+ struct qca_serdev *qcadev;
int ret;
if (speed_type == QCA_INIT_SPEED) {
@@ -1102,6 +1092,15 @@ static int qca_set_speed(struct hci_uart *hu, enum qca_speed_type speed_type)
if (!speed)
return 0;
+ /* Deassert RTS while changing the baudrate of chip and host.
+ * This will prevent chip from transmitting its response with
+ * the new baudrate while the host port is still operating at
+ * the old speed.
+ */
+ qcadev = serdev_device_get_drvdata(hu->serdev);
+ if (qcadev->btsoc_type == QCA_WCN3990)
+ serdev_device_set_rts(hu->serdev, false);
+
qca_baudrate = qca_get_baudrate_value(speed);
bt_dev_dbg(hu->hdev, "Set UART speed to %d", speed);
ret = qca_set_baudrate(hu->hdev, qca_baudrate);
@@ -1109,6 +1108,9 @@ static int qca_set_speed(struct hci_uart *hu, enum qca_speed_type speed_type)
return ret;
host_set_baudrate(hu, speed);
+
+ if (qcadev->btsoc_type == QCA_WCN3990)
+ serdev_device_set_rts(hu->serdev, true);
}
return 0;
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
On Wed, Jan 16, 2019 at 05:16:01PM +0530, Balakrishna Godavarthi wrote:
> wcn3990 requires a power pulse to turn ON/OFF along with
> regulators. Sometimes we are observing the power pulses are sent
> out with some time delay, due to queuing these commands. This is
> causing synchronization issues with chip, which intern delay the
> chip setup or may end up with communication issues.
>
> Signed-off-by: Balakrishna Godavarthi <[email protected]>
> ---
> Changes in v8:
> * Updated 1 second timeout instead of indefinite wait.
>
> Changes in v7:
> * updated the wait time to 5 ms after sending power pulses.
>
> Changes in v6:
> * added serdev_device_write_flush() in qca_send_power_pulse
> instead during the power off pulse.
>
> Changes in v5:
> * added serdev_device_write_flush() in qca_power_off().
> ---
> drivers/bluetooth/hci_qca.c | 46 ++++++++++++++++---------------------
> 1 file changed, 20 insertions(+), 26 deletions(-)
>
> diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
> index f036c8f98ea3..681bfa30467e 100644
> --- a/drivers/bluetooth/hci_qca.c
> +++ b/drivers/bluetooth/hci_qca.c
> @@ -60,6 +60,7 @@
> #define IBS_WAKE_RETRANS_TIMEOUT_MS 100
> #define IBS_TX_IDLE_TIMEOUT_MS 2000
> #define BAUDRATE_SETTLE_TIMEOUT_MS 300
> +#define POWER_PULSE_TRANS_TIMEOUT_MS 1000
nit: Not that it should make a different in normal operation, but 1s
seems extreme. Is there really any chance that the byte hasn't been
sent after say 100ms (which is still an eternity for a single byte)?
> /* susclk rate */
> #define SUSCLK_RATE_32KHZ 32768
> @@ -1013,11 +1014,10 @@ static inline void host_set_baudrate(struct hci_uart *hu, unsigned int speed)
> hci_uart_set_baudrate(hu, speed);
> }
>
> -static int qca_send_power_pulse(struct hci_dev *hdev, u8 cmd)
> +static int qca_send_power_pulse(struct hci_uart *hu, u8 cmd)
> {
> - struct hci_uart *hu = hci_get_drvdata(hdev);
> - struct qca_data *qca = hu->priv;
> - struct sk_buff *skb;
> + int ret;
> + int timeout = __msecs_to_jiffies(POWER_PULSE_TRANS_TIMEOUT_MS);
use msecs_to_jiffies()
> /* These power pulses are single byte command which are sent
> * at required baudrate to wcn3990. On wcn3990, we have an external
> @@ -1029,22 +1029,22 @@ static int qca_send_power_pulse(struct hci_dev *hdev, u8 cmd)
> * save power. Disabling hardware flow control is mandatory while
> * sending power pulses to SoC.
> */
> - bt_dev_dbg(hdev, "sending power pulse %02x to SoC", cmd);
> -
> - skb = bt_skb_alloc(sizeof(cmd), GFP_KERNEL);
> - if (!skb)
> - return -ENOMEM;
> + bt_dev_dbg(hu->hdev, "sending power pulse %02x to controller", cmd);
>
> + serdev_device_write_flush(hu->serdev);
> hci_uart_set_flow_control(hu, true);
> + ret = serdev_device_write_buf(hu->serdev, &cmd, sizeof(cmd));
> + if (ret < 0) {
> + bt_dev_err(hu->hdev, "failed to send power pulse %02x", cmd);
> + return ret;
> + }
>
> - skb_put_u8(skb, cmd);
> - hci_skb_pkt_type(skb) = HCI_COMMAND_PKT;
> -
> - skb_queue_tail(&qca->txq, skb);
> - hci_uart_tx_wakeup(hu);
> -
> - /* Wait for 100 uS for SoC to settle down */
> - usleep_range(100, 200);
> + serdev_device_wait_until_sent(hu->serdev, timeout);
> + /* Wait of 5ms is required for assuring to send the byte on the Tx
> + * line and also for the controller to settle down for the received
> + * byte.
> + */
> + usleep_range(5000, 6000);
I incorrectly claimed that there might be still bytes sitting in the
UART FIFO when serdev_device_wait_until_sent() returns, Johan
corrected me on that (thanks!). So if it takes the SoC 100us to settle
down we should be good with the original code.
Cheers
Matthias
wcn3990 requires a power pulse to turn ON/OFF along with
regulators. Sometimes we are observing the power pulses are sent
out with some time delay, due to queuing these commands. This is
causing synchronization issues with chip, which intern delay the
chip setup or may end up with communication issues.
Signed-off-by: Balakrishna Godavarthi <[email protected]>
---
Changes in v8:
* Updated 1 second timeout instead of indefinite wait.
Changes in v7:
* updated the wait time to 5 ms after sending power pulses.
Changes in v6:
* added serdev_device_write_flush() in qca_send_power_pulse
instead during the power off pulse.
Changes in v5:
* added serdev_device_write_flush() in qca_power_off().
---
drivers/bluetooth/hci_qca.c | 46 ++++++++++++++++---------------------
1 file changed, 20 insertions(+), 26 deletions(-)
diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
index f036c8f98ea3..681bfa30467e 100644
--- a/drivers/bluetooth/hci_qca.c
+++ b/drivers/bluetooth/hci_qca.c
@@ -60,6 +60,7 @@
#define IBS_WAKE_RETRANS_TIMEOUT_MS 100
#define IBS_TX_IDLE_TIMEOUT_MS 2000
#define BAUDRATE_SETTLE_TIMEOUT_MS 300
+#define POWER_PULSE_TRANS_TIMEOUT_MS 1000
/* susclk rate */
#define SUSCLK_RATE_32KHZ 32768
@@ -1013,11 +1014,10 @@ static inline void host_set_baudrate(struct hci_uart *hu, unsigned int speed)
hci_uart_set_baudrate(hu, speed);
}
-static int qca_send_power_pulse(struct hci_dev *hdev, u8 cmd)
+static int qca_send_power_pulse(struct hci_uart *hu, u8 cmd)
{
- struct hci_uart *hu = hci_get_drvdata(hdev);
- struct qca_data *qca = hu->priv;
- struct sk_buff *skb;
+ int ret;
+ int timeout = __msecs_to_jiffies(POWER_PULSE_TRANS_TIMEOUT_MS);
/* These power pulses are single byte command which are sent
* at required baudrate to wcn3990. On wcn3990, we have an external
@@ -1029,22 +1029,22 @@ static int qca_send_power_pulse(struct hci_dev *hdev, u8 cmd)
* save power. Disabling hardware flow control is mandatory while
* sending power pulses to SoC.
*/
- bt_dev_dbg(hdev, "sending power pulse %02x to SoC", cmd);
-
- skb = bt_skb_alloc(sizeof(cmd), GFP_KERNEL);
- if (!skb)
- return -ENOMEM;
+ bt_dev_dbg(hu->hdev, "sending power pulse %02x to controller", cmd);
+ serdev_device_write_flush(hu->serdev);
hci_uart_set_flow_control(hu, true);
+ ret = serdev_device_write_buf(hu->serdev, &cmd, sizeof(cmd));
+ if (ret < 0) {
+ bt_dev_err(hu->hdev, "failed to send power pulse %02x", cmd);
+ return ret;
+ }
- skb_put_u8(skb, cmd);
- hci_skb_pkt_type(skb) = HCI_COMMAND_PKT;
-
- skb_queue_tail(&qca->txq, skb);
- hci_uart_tx_wakeup(hu);
-
- /* Wait for 100 uS for SoC to settle down */
- usleep_range(100, 200);
+ serdev_device_wait_until_sent(hu->serdev, timeout);
+ /* Wait of 5ms is required for assuring to send the byte on the Tx
+ * line and also for the controller to settle down for the received
+ * byte.
+ */
+ usleep_range(5000, 6000);
hci_uart_set_flow_control(hu, false);
return 0;
@@ -1116,7 +1116,6 @@ static int qca_set_speed(struct hci_uart *hu, enum qca_speed_type speed_type)
static int qca_wcn3990_init(struct hci_uart *hu)
{
- struct hci_dev *hdev = hu->hdev;
struct qca_serdev *qcadev;
int ret;
@@ -1139,12 +1138,12 @@ static int qca_wcn3990_init(struct hci_uart *hu)
/* Forcefully enable wcn3990 to enter in to boot mode. */
host_set_baudrate(hu, 2400);
- ret = qca_send_power_pulse(hdev, QCA_WCN3990_POWEROFF_PULSE);
+ ret = qca_send_power_pulse(hu, QCA_WCN3990_POWEROFF_PULSE);
if (ret)
return ret;
qca_set_speed(hu, QCA_INIT_SPEED);
- ret = qca_send_power_pulse(hdev, QCA_WCN3990_POWERON_PULSE);
+ ret = qca_send_power_pulse(hu, QCA_WCN3990_POWERON_PULSE);
if (ret)
return ret;
@@ -1274,13 +1273,8 @@ static const struct qca_vreg_data qca_soc_data = {
static void qca_power_shutdown(struct hci_uart *hu)
{
- struct serdev_device *serdev = hu->serdev;
- unsigned char cmd = QCA_WCN3990_POWEROFF_PULSE;
-
host_set_baudrate(hu, 2400);
- hci_uart_set_flow_control(hu, true);
- serdev_device_write_buf(serdev, &cmd, sizeof(cmd));
- hci_uart_set_flow_control(hu, false);
+ qca_send_power_pulse(hu, QCA_WCN3990_POWEROFF_PULSE);
qca_power_setup(hu, false);
}
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
On Wed, Jan 16, 2019 at 05:16:03PM +0530, Balakrishna Godavarthi wrote:
> During hci down we observed IBS sleep commands are queued in the Tx
> buffer and hci_uart_write_work is sending data to the chip which is
> not required as the chip is powered off. This patch will disable IBS
> and flush the Tx buffer before we turn off the chip.
>
> Signed-off-by: Balakrishna Godavarthi <[email protected]>
> ---
> drivers/bluetooth/hci_qca.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
> index 7e4afcf40da2..7330ba71ada4 100644
> --- a/drivers/bluetooth/hci_qca.c
> +++ b/drivers/bluetooth/hci_qca.c
> @@ -1275,6 +1275,14 @@ static const struct qca_vreg_data qca_soc_data = {
>
> static void qca_power_shutdown(struct hci_uart *hu)
> {
> + struct qca_data *qca = hu->priv;
> +
> + /* From this point we go into power off state. But serial port is
> + * still open, stop queueing the IBS data and flush all the buffered
> + * data in skb's.
> + */
> + clear_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags);
> + qca_flush(hu);
> host_set_baudrate(hu, 2400);
> qca_send_power_pulse(hu, QCA_WCN3990_POWEROFF_PULSE);
> qca_power_setup(hu, false);
Due to a race-condition there could be an IBS sleep command queued
even after clearing the bit and flushing the queue.
In qca_enqueue() we have this:
static int qca_enqueue(struct hci_uart *hu, struct sk_buff *skb)
{
...
/* Don't go to sleep in middle of patch download or
* Out-Of-Band(GPIOs control) sleep is selected.
*/
if (!test_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags)) {
skb_queue_tail(&qca->txq, skb);
return 0;
}
spin_lock_irqsave(&qca->hci_ibs_lock, flags);
}
With process X executing qca_power_shutdown() and process Y running
qca_enqueue() this could happen:
[X] test_bit(STATE_IN_BAND_SLEEP_ENABLED) => set
[Y] clear_bit(STATE_IN_BAND_SLEEP_ENABLED)
[Y] qca_flush(hu);
[X] skb_queue_tail(&qca->txq, skb);
The following should fix this race:
--- a/drivers/bluetooth/hci_qca.c
+++ b/drivers/bluetooth/hci_qca.c
@@ -770,16 +770,17 @@ static int qca_enqueue(struct hci_uart *hu, struct sk_buff *skb)
/* Prepend skb with frame type */
memcpy(skb_push(skb, 1), &hci_skb_pkt_type(skb), 1);
+ spin_lock_irqsave(&qca->hci_ibs_lock, flags);
+
/* Don't go to sleep in middle of patch download or
* Out-Of-Band(GPIOs control) sleep is selected.
*/
if (!test_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags)) {
skb_queue_tail(&qca->txq, skb);
+ spin_unlock_irqrestore(&qca->hci_ibs_lock, flags);
return 0;
}
- spin_lock_irqsave(&qca->hci_ibs_lock, flags);
-
/* Act according to current state */
switch (qca->tx_ibs_state) {
case HCI_IBS_TX_AWAKE:
@@ -1275,13 +1276,17 @@ static const struct qca_vreg_data qca_soc_data = {
static void qca_power_shutdown(struct hci_uart *hu)
{
struct qca_data *qca = hu->priv;
+ unsigned long flags;
/* From this point we go into power off state. But serial port is
* still open, stop queueing the IBS data and flush all the buffered
* data in skb's.
*/
+ spin_lock_irqsave(&qca->hci_ibs_lock, flags);
clear_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags);
qca_flush(hu);
+ spin_unlock_irqrestore(&qca->hci_ibs_lock, flags);
+
host_set_baudrate(hu, 2400);
qca_send_power_pulse(hu, QCA_WCN3990_POWEROFF_PULSE);
qca_power_setup(hu, false);
Cheers
Matthias
Hi Matthias,
On 2019-01-17 01:52, Matthias Kaehlcke wrote:
> On Wed, Jan 16, 2019 at 05:16:01PM +0530, Balakrishna Godavarthi wrote:
>> wcn3990 requires a power pulse to turn ON/OFF along with
>> regulators. Sometimes we are observing the power pulses are sent
>> out with some time delay, due to queuing these commands. This is
>> causing synchronization issues with chip, which intern delay the
>> chip setup or may end up with communication issues.
>>
>> Signed-off-by: Balakrishna Godavarthi <[email protected]>
>> ---
>> Changes in v8:
>> * Updated 1 second timeout instead of indefinite wait.
>>
>> Changes in v7:
>> * updated the wait time to 5 ms after sending power pulses.
>>
>> Changes in v6:
>> * added serdev_device_write_flush() in qca_send_power_pulse
>> instead during the power off pulse.
>>
>> Changes in v5:
>> * added serdev_device_write_flush() in qca_power_off().
>> ---
>> drivers/bluetooth/hci_qca.c | 46
>> ++++++++++++++++---------------------
>> 1 file changed, 20 insertions(+), 26 deletions(-)
>>
>> diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
>> index f036c8f98ea3..681bfa30467e 100644
>> --- a/drivers/bluetooth/hci_qca.c
>> +++ b/drivers/bluetooth/hci_qca.c
>> @@ -60,6 +60,7 @@
>> #define IBS_WAKE_RETRANS_TIMEOUT_MS 100
>> #define IBS_TX_IDLE_TIMEOUT_MS 2000
>> #define BAUDRATE_SETTLE_TIMEOUT_MS 300
>> +#define POWER_PULSE_TRANS_TIMEOUT_MS 1000
>
> nit: Not that it should make a different in normal operation, but 1s
> seems extreme. Is there really any chance that the byte hasn't been
> sent after say 100ms (which is still an eternity for a single byte)?
>
>> /* susclk rate */
>> #define SUSCLK_RATE_32KHZ 32768
>> @@ -1013,11 +1014,10 @@ static inline void host_set_baudrate(struct
>> hci_uart *hu, unsigned int speed)
>> hci_uart_set_baudrate(hu, speed);
>> }
>>
>> -static int qca_send_power_pulse(struct hci_dev *hdev, u8 cmd)
>> +static int qca_send_power_pulse(struct hci_uart *hu, u8 cmd)
>> {
>> - struct hci_uart *hu = hci_get_drvdata(hdev);
>> - struct qca_data *qca = hu->priv;
>> - struct sk_buff *skb;
>> + int ret;
>> + int timeout = __msecs_to_jiffies(POWER_PULSE_TRANS_TIMEOUT_MS);
>
> use msecs_to_jiffies()
>
[Bala]: will upadte.
>> /* These power pulses are single byte command which are sent
>> * at required baudrate to wcn3990. On wcn3990, we have an external
>> @@ -1029,22 +1029,22 @@ static int qca_send_power_pulse(struct hci_dev
>> *hdev, u8 cmd)
>> * save power. Disabling hardware flow control is mandatory while
>> * sending power pulses to SoC.
>> */
>> - bt_dev_dbg(hdev, "sending power pulse %02x to SoC", cmd);
>> -
>> - skb = bt_skb_alloc(sizeof(cmd), GFP_KERNEL);
>> - if (!skb)
>> - return -ENOMEM;
>> + bt_dev_dbg(hu->hdev, "sending power pulse %02x to controller", cmd);
>>
>> + serdev_device_write_flush(hu->serdev);
>> hci_uart_set_flow_control(hu, true);
>> + ret = serdev_device_write_buf(hu->serdev, &cmd, sizeof(cmd));
>> + if (ret < 0) {
>> + bt_dev_err(hu->hdev, "failed to send power pulse %02x", cmd);
>> + return ret;
>> + }
>>
>> - skb_put_u8(skb, cmd);
>> - hci_skb_pkt_type(skb) = HCI_COMMAND_PKT;
>> -
>> - skb_queue_tail(&qca->txq, skb);
>> - hci_uart_tx_wakeup(hu);
>> -
>> - /* Wait for 100 uS for SoC to settle down */
>> - usleep_range(100, 200);
>> + serdev_device_wait_until_sent(hu->serdev, timeout);
>> + /* Wait of 5ms is required for assuring to send the byte on the Tx
>> + * line and also for the controller to settle down for the received
>> + * byte.
>> + */
>> + usleep_range(5000, 6000);
>
> I incorrectly claimed that there might be still bytes sitting in the
> UART FIFO when serdev_device_wait_until_sent() returns, Johan
> corrected me on that (thanks!). So if it takes the SoC 100us to settle
> down we should be good with the original code.
[Bala]: sure will revert, i think he commented that wait_until_sent()
will only guarantee circular buffer is empty. if
wait_until_sent()
guarantee us that the data was transmitted from the FIFO, then
100us will work.
>
> Cheers
>
> Matthias
--
Regards
Balakrishna.
On 2019-01-17 04:38, Matthias Kaehlcke wrote:
> On Wed, Jan 16, 2019 at 05:16:03PM +0530, Balakrishna Godavarthi wrote:
>> During hci down we observed IBS sleep commands are queued in the Tx
>> buffer and hci_uart_write_work is sending data to the chip which is
>> not required as the chip is powered off. This patch will disable IBS
>> and flush the Tx buffer before we turn off the chip.
>>
>> Signed-off-by: Balakrishna Godavarthi <[email protected]>
>> ---
>> drivers/bluetooth/hci_qca.c | 8 ++++++++
>> 1 file changed, 8 insertions(+)
>>
>> diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
>> index 7e4afcf40da2..7330ba71ada4 100644
>> --- a/drivers/bluetooth/hci_qca.c
>> +++ b/drivers/bluetooth/hci_qca.c
>> @@ -1275,6 +1275,14 @@ static const struct qca_vreg_data qca_soc_data
>> = {
>>
>> static void qca_power_shutdown(struct hci_uart *hu)
>> {
>> + struct qca_data *qca = hu->priv;
>> +
>> + /* From this point we go into power off state. But serial port is
>> + * still open, stop queueing the IBS data and flush all the buffered
>> + * data in skb's.
>> + */
>> + clear_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags);
>> + qca_flush(hu);
>> host_set_baudrate(hu, 2400);
>> qca_send_power_pulse(hu, QCA_WCN3990_POWEROFF_PULSE);
>> qca_power_setup(hu, false);
>
> Due to a race-condition there could be an IBS sleep command queued
> even after clearing the bit and flushing the queue.
>
> In qca_enqueue() we have this:
>
> static int qca_enqueue(struct hci_uart *hu, struct sk_buff *skb)
> {
> ...
>
> /* Don't go to sleep in middle of patch download or
> * Out-Of-Band(GPIOs control) sleep is selected.
> */
> if (!test_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags)) {
> skb_queue_tail(&qca->txq, skb);
> return 0;
> }
>
> spin_lock_irqsave(&qca->hci_ibs_lock, flags);
> }
>
> With process X executing qca_power_shutdown() and process Y running
> qca_enqueue() this could happen:
>
> [X] test_bit(STATE_IN_BAND_SLEEP_ENABLED) => set
> [Y] clear_bit(STATE_IN_BAND_SLEEP_ENABLED)
> [Y] qca_flush(hu);
> [X] skb_queue_tail(&qca->txq, skb);
>
> The following should fix this race:
>
> --- a/drivers/bluetooth/hci_qca.c
> +++ b/drivers/bluetooth/hci_qca.c
> @@ -770,16 +770,17 @@ static int qca_enqueue(struct hci_uart *hu,
> struct sk_buff *skb)
> /* Prepend skb with frame type */
> memcpy(skb_push(skb, 1), &hci_skb_pkt_type(skb), 1);
>
> + spin_lock_irqsave(&qca->hci_ibs_lock, flags);
> +
> /* Don't go to sleep in middle of patch download or
> * Out-Of-Band(GPIOs control) sleep is selected.
> */
> if (!test_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags)) {
> skb_queue_tail(&qca->txq, skb);
> + spin_unlock_irqrestore(&qca->hci_ibs_lock, flags);
> return 0;
> }
>
> - spin_lock_irqsave(&qca->hci_ibs_lock, flags);
> -
> /* Act according to current state */
> switch (qca->tx_ibs_state) {
> case HCI_IBS_TX_AWAKE:
> @@ -1275,13 +1276,17 @@ static const struct qca_vreg_data qca_soc_data
> = {
> static void qca_power_shutdown(struct hci_uart *hu)
> {
> struct qca_data *qca = hu->priv;
> + unsigned long flags;
>
> /* From this point we go into power off state. But serial port is
> * still open, stop queueing the IBS data and flush all the buffered
> * data in skb's.
> */
> + spin_lock_irqsave(&qca->hci_ibs_lock, flags);
> clear_bit(STATE_IN_BAND_SLEEP_ENABLED, &qca->flags);
> qca_flush(hu);
> + spin_unlock_irqrestore(&qca->hci_ibs_lock, flags);
> +
[Bala]: Thanks for catch this. yes you are rite, we will have an byte
queued in the skb.
will update with the lock.
> host_set_baudrate(hu, 2400);
> qca_send_power_pulse(hu, QCA_WCN3990_POWEROFF_PULSE);
> qca_power_setup(hu, false);
>
> Cheers
>
> Matthias
--
Regards
Balakrishna.
On Thu, Jan 17, 2019 at 03:55:17PM +0530, Balakrishna Godavarthi wrote:
> Hi Matthias,
>
> On 2019-01-17 01:52, Matthias Kaehlcke wrote:
> >> - /* Wait for 100 uS for SoC to settle down */
> >> - usleep_range(100, 200);
> >> + serdev_device_wait_until_sent(hu->serdev, timeout);
> >> + /* Wait of 5ms is required for assuring to send the byte on the Tx
> >> + * line and also for the controller to settle down for the received
> >> + * byte.
> >> + */
> >> + usleep_range(5000, 6000);
> >
> > I incorrectly claimed that there might be still bytes sitting in the
> > UART FIFO when serdev_device_wait_until_sent() returns, Johan
> > corrected me on that (thanks!). So if it takes the SoC 100us to settle
> > down we should be good with the original code.
>
> [Bala]: sure will revert, i think he commented that wait_until_sent()
> will only guarantee circular buffer is empty. if wait_until_sent()
> guarantee us that the data was transmitted from the FIFO, then 100us
> will work.
No, Matthias is correct; I claimed that the UART FIFO will be empty (at
least as long as flow control is disabled, otherwise it may never empty
and we therefore also have a time out).
Johan
On 2019-01-17 21:43, Johan Hovold wrote:
> On Thu, Jan 17, 2019 at 03:55:17PM +0530, Balakrishna Godavarthi wrote:
>> Hi Matthias,
>>
>> On 2019-01-17 01:52, Matthias Kaehlcke wrote:
>
>> >> - /* Wait for 100 uS for SoC to settle down */
>> >> - usleep_range(100, 200);
>> >> + serdev_device_wait_until_sent(hu->serdev, timeout);
>> >> + /* Wait of 5ms is required for assuring to send the byte on the Tx
>> >> + * line and also for the controller to settle down for the received
>> >> + * byte.
>> >> + */
>> >> + usleep_range(5000, 6000);
>> >
>> > I incorrectly claimed that there might be still bytes sitting in the
>> > UART FIFO when serdev_device_wait_until_sent() returns, Johan
>> > corrected me on that (thanks!). So if it takes the SoC 100us to settle
>> > down we should be good with the original code.
>>
>> [Bala]: sure will revert, i think he commented that wait_until_sent()
>> will only guarantee circular buffer is empty. if wait_until_sent()
>> guarantee us that the data was transmitted from the FIFO, then 100us
>> will work.
>
> No, Matthias is correct; I claimed that the UART FIFO will be empty (at
> least as long as flow control is disabled, otherwise it may never empty
> and we therefore also have a time out).
>
> Johan
[Bala]: Thanks Johan for clarification.
--
Regards
Balakrishna.
Hi Matthias,
On 2019-01-17 01:52, Matthias Kaehlcke wrote:
> On Wed, Jan 16, 2019 at 05:16:01PM +0530, Balakrishna Godavarthi wrote:
>> wcn3990 requires a power pulse to turn ON/OFF along with
>> regulators. Sometimes we are observing the power pulses are sent
>> out with some time delay, due to queuing these commands. This is
>> causing synchronization issues with chip, which intern delay the
>> chip setup or may end up with communication issues.
>>
>> Signed-off-by: Balakrishna Godavarthi <[email protected]>
>> ---
>> Changes in v8:
>> * Updated 1 second timeout instead of indefinite wait.
>>
>> Changes in v7:
>> * updated the wait time to 5 ms after sending power pulses.
>>
>> Changes in v6:
>> * added serdev_device_write_flush() in qca_send_power_pulse
>> instead during the power off pulse.
>>
>> Changes in v5:
>> * added serdev_device_write_flush() in qca_power_off().
>> ---
>> drivers/bluetooth/hci_qca.c | 46
>> ++++++++++++++++---------------------
>> 1 file changed, 20 insertions(+), 26 deletions(-)
>>
>> diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
>> index f036c8f98ea3..681bfa30467e 100644
>> --- a/drivers/bluetooth/hci_qca.c
>> +++ b/drivers/bluetooth/hci_qca.c
>> @@ -60,6 +60,7 @@
>> #define IBS_WAKE_RETRANS_TIMEOUT_MS 100
>> #define IBS_TX_IDLE_TIMEOUT_MS 2000
>> #define BAUDRATE_SETTLE_TIMEOUT_MS 300
>> +#define POWER_PULSE_TRANS_TIMEOUT_MS 1000
>
> nit: Not that it should make a different in normal operation, but 1s
> seems extreme. Is there really any chance that the byte hasn't been
> sent after say 100ms (which is still an eternity for a single byte)?
>
[Bala]: i missed to address this. for now let us have 1 second delay.
based on stress test or further enchantments we can reduce this
delay.
>> /* susclk rate */
>> #define SUSCLK_RATE_32KHZ 32768
>> @@ -1013,11 +1014,10 @@ static inline void host_set_baudrate(struct
>> hci_uart *hu, unsigned int speed)
>> hci_uart_set_baudrate(hu, speed);
>> }
>>
>> -static int qca_send_power_pulse(struct hci_dev *hdev, u8 cmd)
>> +static int qca_send_power_pulse(struct hci_uart *hu, u8 cmd)
>> {
>> - struct hci_uart *hu = hci_get_drvdata(hdev);
>> - struct qca_data *qca = hu->priv;
>> - struct sk_buff *skb;
>> + int ret;
>> + int timeout = __msecs_to_jiffies(POWER_PULSE_TRANS_TIMEOUT_MS);
>
> use msecs_to_jiffies()
>
>> /* These power pulses are single byte command which are sent
>> * at required baudrate to wcn3990. On wcn3990, we have an external
>> @@ -1029,22 +1029,22 @@ static int qca_send_power_pulse(struct hci_dev
>> *hdev, u8 cmd)
>> * save power. Disabling hardware flow control is mandatory while
>> * sending power pulses to SoC.
>> */
>> - bt_dev_dbg(hdev, "sending power pulse %02x to SoC", cmd);
>> -
>> - skb = bt_skb_alloc(sizeof(cmd), GFP_KERNEL);
>> - if (!skb)
>> - return -ENOMEM;
>> + bt_dev_dbg(hu->hdev, "sending power pulse %02x to controller", cmd);
>>
>> + serdev_device_write_flush(hu->serdev);
>> hci_uart_set_flow_control(hu, true);
>> + ret = serdev_device_write_buf(hu->serdev, &cmd, sizeof(cmd));
>> + if (ret < 0) {
>> + bt_dev_err(hu->hdev, "failed to send power pulse %02x", cmd);
>> + return ret;
>> + }
>>
>> - skb_put_u8(skb, cmd);
>> - hci_skb_pkt_type(skb) = HCI_COMMAND_PKT;
>> -
>> - skb_queue_tail(&qca->txq, skb);
>> - hci_uart_tx_wakeup(hu);
>> -
>> - /* Wait for 100 uS for SoC to settle down */
>> - usleep_range(100, 200);
>> + serdev_device_wait_until_sent(hu->serdev, timeout);
>> + /* Wait of 5ms is required for assuring to send the byte on the Tx
>> + * line and also for the controller to settle down for the received
>> + * byte.
>> + */
>> + usleep_range(5000, 6000);
>
> I incorrectly claimed that there might be still bytes sitting in the
> UART FIFO when serdev_device_wait_until_sent() returns, Johan
> corrected me on that (thanks!). So if it takes the SoC 100us to settle
> down we should be good with the original code.
>
> Cheers
>
> Matthias
--
Regards
Balakrishna.