Re: [PATCH] hw/char: disable ibex uart receive if the buffer is full

[Alexander Wagner]

On 08.03.21 14:47, Alistair Francis wrote:
  hw/char/ibex_uart.c         | 20 +++++++++++++++-----
  include/hw/char/ibex_uart.h |  4 ++++
  2 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/hw/char/ibex_uart.c b/hw/char/ibex_uart.c
index 89f1182c9b..dac09d53d6 100644
--- a/hw/char/ibex_uart.c
+++ b/hw/char/ibex_uart.c
@@ -66,7 +66,8 @@ static int ibex_uart_can_receive(void *opaque)
  {
      IbexUartState *s = opaque;

-    if (s->uart_ctrl & R_CTRL_RX_ENABLE_MASK) {
+    if ((s->uart_ctrl & R_CTRL_RX_ENABLE_MASK)
+           && !(s->uart_status & R_STATUS_RXFULL_MASK)) {
          return 1;
      }

@@ -83,6 +84,8 @@ static void ibex_uart_receive(void *opaque, const uint8_t 
*buf, int size)

      s->uart_status &= ~R_STATUS_RXIDLE_MASK;
      s->uart_status &= ~R_STATUS_RXEMPTY_MASK;
+    s->uart_status |= R_STATUS_RXFULL_MASK;
Doesn't this mean we set RXFULL on every receive? Shouldn't this check
the rx_level first?

Alistair

Thank you for having a look! :)

Yes, this is correct. The RXFULL is currently set on every receive. The 
RXFULL is used to indicate to QEMU that the device cannot receive any 
further bytes.

As the FIFO buffers are currently not yet implemented I thought it would 
make sense to behave like the OT UART could only receive one byte at a time.

Alex