uart_rx_fifo_available considerations

[d-a-v]

According to @jasoroony's comments and after checking the doc, it appears that:

inline size_t uart_rx_fifo_available(uart_t* uart) {
     return (USS(uart->uart_nr) >> USRXC) & 0x7F;
}

should be:

inline size_t uart_rx_fifo_available(uart_t* uart) {
     return (USS(uart->uart_nr) >> USRXC) & 0xFF;
}

reference section 11.3.7:

Rx fifo length:
(READ_PERI_REG(UART_STATUS(UART0))>>UART_RXFIFO_CNT_S)&UART_RXFIFO_CNT;

where

tools/sdk/include/uart_register.h:#define UART_RXFIFO_CNT 0x000000FF

(both USRXC and UART_RXFIFO_CNT_S are 0)

[ghost]

Good work guys, the new commit works perfectly with all my serial tests, including a simulated DOS attack while WiFi is on. I did see my S/W buffer as high as 900 characters during that, but it's still running and no errors were detected in the Serial stream.

[d-a-v]

Thanks for your feedback !

[sarfata]

Just wanted to add a comment here to say that this change fixes a crash in my project that was very easy to reproduce when there is sufficient messages being sent to the ESP8266 serial port.

With the new version of uart.c, I have not been able to reproduce the crash after hours of testing where before it would usually crash in 15 minutes (with load). The symptom was a hardware watchdog timer crash with epc in uart_isr.

🙇 🤗 🎁