[PATCH] serial: 8250: Fix TX deadlock when using DMA

[Raul E Rangel]

`dmaengine_terminate_async` does not guarantee that the
`__dma_tx_complete` callback will run. The callback is currently the
only place where `dma->tx_running` gets cleared. If the transaction is
canceled and the callback never runs, then `dma->tx_running` will never
get cleared and we will never schedule new TX DMA transactions again.

This change makes it so we clear `dma->tx_running` after we terminate
the DMA transaction. This is "safe" because `serial8250_tx_dma_flush`
is holding the UART port lock. The first thing the callback does is also
grab the UART port lock, so access to `dma->tx_running` is serialized.

Fixes: 9e512eaaf8f4 ("serial: 8250: Fix fifo underflow on flush")

Signed-off-by: Raul E Rangel <rrangel@xxxxxxxxxx>
---

 drivers/tty/serial/8250/8250_dma.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/drivers/tty/serial/8250/8250_dma.c b/drivers/tty/serial/8250/8250_dma.c
index bdd26c9f34bd..3b6452e759d5 100644
--- a/drivers/tty/serial/8250/8250_dma.c
+++ b/drivers/tty/serial/8250/8250_dma.c
@@ -162,7 +162,22 @@ void serial8250_tx_dma_flush(struct uart_8250_port *p)
 	 */
 	dma->tx_size = 0;
 
+	/*
+	 * We can't use `dmaengine_terminate_sync` because `uart_flush_buffer` is
+	 * holding the uart port spinlock.
+	 */
 	dmaengine_terminate_async(dma->txchan);
+
+	/*
+	 * The callback might or might not run. If it doesn't run, we need to ensure
+	 * that `tx_running` is cleared so that we can schedule new transactions.
+	 * If it does run, then the zombie callback will clear `tx_running` again
+	 * and perform a no-op since `tx_size` was cleared above.
+	 *
+	 * In either case, we ASSUME the DMA transaction will terminate before we
+	 * issue a new `serial8250_tx_dma`.
+	 */
+	dma->tx_running = 0;
 }
 
 int serial8250_rx_dma(struct uart_8250_port *p)
-- 
2.53.0.rc2.204.g2597b5adb4-goog