Re: [RFCv2 ben-wpan] fw: add trac reg to tx done

[Miquel Raynal <miquel.raynal@bootlin.com>]

Hi Alexander,

aahringo@redhat.com wrote on Mon,  5 Sep 2022 20:12:18 -0400:

> This patch adds an additional byte for tx done for the trac status
> register which can then evaluated by the atusb Linux driver.
> 
> We need to move the transition to RX_AACK_ON after reading the trac
> register value. The current optimization by switchting to RX_AACK_ON
> after invoking transceiver transmission could have the side effect that
> the trac register is not read out before a frame is received. Receiving
> another frame will overwrite the trac register. We can only switch to
> RX_AACK_ON state after we read out the trac register.
> ---
> Not tested yet.
> 
>  atusb/fw/mac.c | 24 ++++++++++++------------
>  1 file changed, 12 insertions(+), 12 deletions(-)
> 
> diff --git a/atusb/fw/mac.c b/atusb/fw/mac.c
> index 835002c..165ce30 100644
> --- a/atusb/fw/mac.c
> +++ b/atusb/fw/mac.c
> @@ -32,7 +32,7 @@ static uint8_t tx_buf[MAX_PSDU];
>  static uint8_t tx_size = 0;
>  static bool txing = 0;
>  static bool queued_tx_ack = 0;
> -static uint8_t next_seq, this_seq, queued_seq;
> +static uint8_t next_seq, this_seq, this_data[2], queued_data[2];
>  
>  
>  /* ----- Receive buffer management ----------------------------------------- */
> @@ -65,7 +65,8 @@ static void usb_next(void)
>  	}
>  
>  	if (queued_tx_ack) {
> -		usb_send(&eps[1], &queued_seq, 1, tx_ack_done, NULL);
> +		usb_send(&eps[1], queued_data, sizeof(queued_data),
> +			 tx_ack_done, NULL);
>  		queued_tx_ack = 0;	
>  	}
>  }
> @@ -124,11 +125,17 @@ static bool handle_irq(void)
>  
>  	if (txing) {
>  		if (eps[1].state == EP_IDLE) {
> -			usb_send(&eps[1], &this_seq, 1, tx_ack_done, NULL);
> +			this_data[0] = this_seq;
> +			this_data[1] = reg_read(REG_TRX_STATE);
> +			usb_send(&eps[1], this_data, sizeof(this_data),
> +				 tx_ack_done, NULL);
>  		} else {
>  			queued_tx_ack = 1;
> -			queued_seq = this_seq;
> +			queued_data[0] = this_seq;
> +			queued_data[1] = reg_read(REG_TRX_STATE);
>  		}
> +		change_state(TRX_CMD_PLL_ON);
> +		change_state(TRX_CMD_RX_AACK_ON);
>  		txing = 0;
>  		return 1;
>  	}
> @@ -215,13 +222,6 @@ static void do_tx(void *user)
>  
>  	txing = 1;
>  	this_seq = next_seq;
> -
> -	/*
> -	 * Wait until we reach BUSY_TX_ARET, so that we command the transition to
> -	 * RX_AACK_ON which will be executed upon TX completion.
> -	 */
> -	change_state(TRX_CMD_PLL_ON);
> -	change_state(TRX_CMD_RX_AACK_ON);

Mmmh, I see. Good catch.

And this patch works.

>  }
>  
>  
> @@ -242,7 +242,7 @@ void mac_reset(void)
>  	txing = 0;
>  	queued_tx_ack = 0;
>  	rx_in = rx_out = 0;
> -	next_seq = this_seq = queued_seq = 0;
> +	next_seq = this_seq = queued_data[0], queued_data[1] = 0;
>  
>  	/* enable CRC and PHY_RSSI (with RX_CRC_VALID) in SPI status return */
>  	reg_write(REG_TRX_CTRL_1,


Thanks,
Miquèl