It turns out that received CEC messages can arrive faster than
can be processed by the CEC framework, resulting in lost messages.
Instead of storing only one CEC message, store up to 8.
Also fix a bug where the EOM bit wasn't checked for a received
message of length 1, so POLL messages weren't properly reported.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xxxxxxxxx>
---
drivers/media/usb/pulse8-cec/pulse8-cec.c | 80 +++++++++++++++++------
1 file changed, 61 insertions(+), 19 deletions(-)
diff --git a/drivers/media/usb/pulse8-cec/pulse8-cec.c b/drivers/media/usb/pulse8-cec/pulse8-cec.c
index 34dbc567dbe0..961b34dfaf6d 100644
--- a/drivers/media/usb/pulse8-cec/pulse8-cec.c
+++ b/drivers/media/usb/pulse8-cec/pulse8-cec.c
@@ -164,6 +164,8 @@ static const char *pulse8_msgname(u8 cmd)
#define PING_PERIOD (15 * HZ)
+#define NUM_MSGS 8
+
struct pulse8 {
struct device *dev;
struct serio *serio;
@@ -173,8 +175,12 @@ struct pulse8 {
struct delayed_work ping_eeprom_work;
struct work_struct irq_work;
- u8 irq_work_result;
- struct cec_msg rx_msg;
+ struct cec_msg rx_msg[NUM_MSGS];
+ unsigned int rx_msg_cur_idx, rx_msg_num;
+ /* protect rx_msg_cur_idx and rx_msg_num */
+ spinlock_t msg_lock;
+ u8 new_rx_msg[CEC_MAX_MSG_SIZE];
+ u8 new_rx_msg_len;
struct work_struct tx_work;
u32 tx_done_status;
@@ -321,15 +327,22 @@ static void pulse8_irq_work_handler(struct work_struct *work)
{
struct pulse8 *pulse8 =
container_of(work, struct pulse8, irq_work);
- u8 result = pulse8->irq_work_result;
+ unsigned long flags;
u32 status;
- pulse8->irq_work_result = 0;
- switch (result & 0x3f) {
- case MSGCODE_FRAME_DATA:
- cec_received_msg(pulse8->adap, &pulse8->rx_msg);
- break;
+ spin_lock_irqsave(&pulse8->msg_lock, flags);
+ while (pulse8->rx_msg_num) {
+ spin_unlock_irqrestore(&pulse8->msg_lock, flags);
+ cec_received_msg(pulse8->adap,
+ &pulse8->rx_msg[pulse8->rx_msg_cur_idx]);
+ spin_lock_irqsave(&pulse8->msg_lock, flags);
+ if (pulse8->rx_msg_num)
+ pulse8->rx_msg_num--;
+ pulse8->rx_msg_cur_idx =
+ (pulse8->rx_msg_cur_idx + 1) % NUM_MSGS;
}
+ spin_unlock_irqrestore(&pulse8->msg_lock, flags);
+
mutex_lock(&pulse8->lock);
status = pulse8->tx_done_status;
pulse8->tx_done_status = 0;
@@ -342,6 +355,8 @@ static irqreturn_t pulse8_interrupt(struct serio *serio, unsigned char data,
unsigned int flags)
{
struct pulse8 *pulse8 = serio_get_drvdata(serio);
+ unsigned long irq_flags;
+ unsigned int idx;
if (!pulse8->started && data != MSGSTART)
return IRQ_HANDLED;
@@ -353,7 +368,6 @@ static irqreturn_t pulse8_interrupt(struct serio *serio, unsigned char data,
data += MSGOFFSET;
pulse8->escape = false;
} else if (data == MSGEND) {
- struct cec_msg *msg = &pulse8->rx_msg;
u8 msgcode = pulse8->buf[0];
if (debug > 1)
@@ -362,19 +376,43 @@ static irqreturn_t pulse8_interrupt(struct serio *serio, unsigned char data,
pulse8->idx, pulse8->buf);
switch (msgcode & 0x3f) {
case MSGCODE_FRAME_START:
- msg->len = 1;
- msg->msg[0] = pulse8->buf[1];
- break;
+ /*
+ * Test if we are receiving a new msg when a previous
+ * message is still pending.
+ */
+ if (!(msgcode & MSGCODE_FRAME_EOM)) {
+ pulse8->new_rx_msg_len = 1;
+ pulse8->new_rx_msg[0] = pulse8->buf[1];
+ break;
+ }
+ /* fall through */
case MSGCODE_FRAME_DATA:
- if (msg->len == CEC_MAX_MSG_SIZE)
+ if (pulse8->new_rx_msg_len < CEC_MAX_MSG_SIZE)
+ pulse8->new_rx_msg[pulse8->new_rx_msg_len++] =
+ pulse8->buf[1];
+ if (!(msgcode & MSGCODE_FRAME_EOM))
break;
- msg->msg[msg->len++] = pulse8->buf[1];
- if (msgcode & MSGCODE_FRAME_EOM) {
- WARN_ON(pulse8->irq_work_result);
- pulse8->irq_work_result = msgcode;
- schedule_work(&pulse8->irq_work);
+
+ spin_lock_irqsave(&pulse8->msg_lock, irq_flags);
+ idx = (pulse8->rx_msg_cur_idx + pulse8->rx_msg_num) %
+ NUM_MSGS;
+ if (pulse8->rx_msg_num == NUM_MSGS) {
+ dev_warn(pulse8->dev,
+ "message queue is full, dropping %*ph\n",
+ pulse8->new_rx_msg_len,
+ pulse8->new_rx_msg);
+ spin_unlock_irqrestore(&pulse8->msg_lock,
+ irq_flags);
+ pulse8->new_rx_msg_len = 0;
break;
}
+ pulse8->rx_msg_num++;
+ memcpy(pulse8->rx_msg[idx].msg, pulse8->new_rx_msg,
+ pulse8->new_rx_msg_len);
+ pulse8->rx_msg[idx].len = pulse8->new_rx_msg_len;
+ spin_unlock_irqrestore(&pulse8->msg_lock, irq_flags);
+ schedule_work(&pulse8->irq_work);
+ pulse8->new_rx_msg_len = 0;
break;
case MSGCODE_TRANSMIT_SUCCEEDED:
WARN_ON(pulse8->tx_done_status);
@@ -403,6 +441,7 @@ static irqreturn_t pulse8_interrupt(struct serio *serio, unsigned char data,
case MSGCODE_LOW_ERROR:
case MSGCODE_RECEIVE_FAILED:
case MSGCODE_TIMEOUT_ERROR:
+ pulse8->new_rx_msg_len = 0;
break;
case MSGCODE_COMMAND_ACCEPTED:
case MSGCODE_COMMAND_REJECTED:
@@ -443,8 +482,10 @@ static int pulse8_cec_adap_enable(struct cec_adapter *adap, bool enable)
cmd[1] = enable;
err = pulse8_send_and_wait(pulse8, cmd, 2,
MSGCODE_COMMAND_ACCEPTED, 1);
- if (!enable)
+ if (!enable) {
+ pulse8->rx_msg_num = 0;
pulse8->tx_done_status = 0;
+ }
mutex_unlock(&pulse8->lock);
return enable ? err : 0;
}
@@ -790,6 +831,7 @@ static int pulse8_connect(struct serio *serio, struct serio_driver *drv)
INIT_WORK(&pulse8->irq_work, pulse8_irq_work_handler);
INIT_WORK(&pulse8->tx_work, pulse8_tx_work_handler);
mutex_init(&pulse8->lock);
+ spin_lock_init(&pulse8->msg_lock);
pulse8->config_pending = false;
err = serio_open(serio, drv);
--
2.23.0
