1 files changed, 389 insertions, 300 deletions

diff --git a/drivers/tty/serial/stm32-usart.c b/drivers/tty/serial/stm32-usart.c
index a1490033aa16..ad06b760cfca 100644
--- a/drivers/tty/serial/stm32-usart.c
+++ b/drivers/tty/serial/stm32-usart.c
@@ -9,6 +9,7 @@
  * Inspired by st-asc.c from STMicroelectronics (c)
  */
 
+#include <linux/bitfield.h>
 #include <linux/clk.h>
 #include <linux/console.h>
 #include <linux/delay.h>
@@ -39,60 +40,64 @@
 /* Register offsets */
 static struct stm32_usart_info __maybe_unused stm32f4_info = {
 	.ofs = {
-		.isr	= 0x00,
-		.rdr	= 0x04,
-		.tdr	= 0x04,
-		.brr	= 0x08,
-		.cr1	= 0x0c,
-		.cr2	= 0x10,
-		.cr3	= 0x14,
-		.gtpr	= 0x18,
-		.rtor	= UNDEF_REG,
-		.rqr	= UNDEF_REG,
-		.icr	= UNDEF_REG,
+		.isr		= 0x00,
+		.rdr		= 0x04,
+		.tdr		= 0x04,
+		.brr		= 0x08,
+		.cr1		= 0x0c,
+		.cr2		= 0x10,
+		.cr3		= 0x14,
+		.gtpr		= 0x18,
+		.rtor		= UNDEF_REG,
+		.rqr		= UNDEF_REG,
+		.icr		= UNDEF_REG,
+		.presc		= UNDEF_REG,
+		.hwcfgr1	= UNDEF_REG,
 	},
 	.cfg = {
 		.uart_enable_bit = 13,
 		.has_7bits_data = false,
-		.fifosize = 1,
 	}
 };
 
 static struct stm32_usart_info __maybe_unused stm32f7_info = {
 	.ofs = {
-		.cr1	= 0x00,
-		.cr2	= 0x04,
-		.cr3	= 0x08,
-		.brr	= 0x0c,
-		.gtpr	= 0x10,
-		.rtor	= 0x14,
-		.rqr	= 0x18,
-		.isr	= 0x1c,
-		.icr	= 0x20,
-		.rdr	= 0x24,
-		.tdr	= 0x28,
+		.cr1		= 0x00,
+		.cr2		= 0x04,
+		.cr3		= 0x08,
+		.brr		= 0x0c,
+		.gtpr		= 0x10,
+		.rtor		= 0x14,
+		.rqr		= 0x18,
+		.isr		= 0x1c,
+		.icr		= 0x20,
+		.rdr		= 0x24,
+		.tdr		= 0x28,
+		.presc		= UNDEF_REG,
+		.hwcfgr1	= UNDEF_REG,
 	},
 	.cfg = {
 		.uart_enable_bit = 0,
 		.has_7bits_data = true,
 		.has_swap = true,
-		.fifosize = 1,
 	}
 };
 
 static struct stm32_usart_info __maybe_unused stm32h7_info = {
 	.ofs = {
-		.cr1	= 0x00,
-		.cr2	= 0x04,
-		.cr3	= 0x08,
-		.brr	= 0x0c,
-		.gtpr	= 0x10,
-		.rtor	= 0x14,
-		.rqr	= 0x18,
-		.isr	= 0x1c,
-		.icr	= 0x20,
-		.rdr	= 0x24,
-		.tdr	= 0x28,
+		.cr1		= 0x00,
+		.cr2		= 0x04,
+		.cr3		= 0x08,
+		.brr		= 0x0c,
+		.gtpr		= 0x10,
+		.rtor		= 0x14,
+		.rqr		= 0x18,
+		.isr		= 0x1c,
+		.icr		= 0x20,
+		.rdr		= 0x24,
+		.tdr		= 0x28,
+		.presc		= 0x2c,
+		.hwcfgr1	= 0x3f0,
 	},
 	.cfg = {
 		.uart_enable_bit = 0,
@@ -100,7 +105,6 @@ static struct stm32_usart_info __maybe_unused stm32h7_info = {
 		.has_swap = true,
 		.has_wakeup = true,
 		.has_fifo = true,
-		.fifosize = 16,
 	}
 };
 
@@ -226,8 +230,6 @@ static int stm32_usart_config_rs485(struct uart_port *port, struct ktermios *ter
 
 	stm32_usart_clr_bits(port, ofs->cr1, BIT(cfg->uart_enable_bit));
 
-	rs485conf->flags |= SER_RS485_RX_DURING_TX;
-
 	if (rs485conf->flags & SER_RS485_ENABLED) {
 		cr1 = readl_relaxed(port->membase + ofs->cr1);
 		cr3 = readl_relaxed(port->membase + ofs->cr3);
@@ -252,6 +254,10 @@ static int stm32_usart_config_rs485(struct uart_port *port, struct ktermios *ter
 
 		writel_relaxed(cr3, port->membase + ofs->cr3);
 		writel_relaxed(cr1, port->membase + ofs->cr1);
+
+		if (!port->rs485_rx_during_tx_gpio)
+			rs485conf->flags |= SER_RS485_RX_DURING_TX;
+
 	} else {
 		stm32_usart_clr_bits(port, ofs->cr3,
 				     USART_CR3_DEM | USART_CR3_DEP);
@@ -285,15 +291,57 @@ static int stm32_usart_init_rs485(struct uart_port *port,
 	return uart_get_rs485_mode(port);
 }
 
-static bool stm32_usart_rx_dma_enabled(struct uart_port *port)
+static bool stm32_usart_rx_dma_started(struct stm32_port *stm32_port)
 {
-	struct stm32_port *stm32_port = to_stm32_port(port);
-	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
+	return stm32_port->rx_ch ? stm32_port->rx_dma_busy : false;
+}
+
+static void stm32_usart_rx_dma_terminate(struct stm32_port *stm32_port)
+{
+	dmaengine_terminate_async(stm32_port->rx_ch);
+	stm32_port->rx_dma_busy = false;
+}
+
+static int stm32_usart_dma_pause_resume(struct stm32_port *stm32_port,
+					struct dma_chan *chan,
+					enum dma_status expected_status,
+					int dmaengine_pause_or_resume(struct dma_chan *),
+					bool stm32_usart_xx_dma_started(struct stm32_port *),
+					void stm32_usart_xx_dma_terminate(struct stm32_port *))
+{
+	struct uart_port *port = &stm32_port->port;
+	enum dma_status dma_status;
+	int ret;
 
-	if (!stm32_port->rx_ch)
-		return false;
+	if (!stm32_usart_xx_dma_started(stm32_port))
+		return -EPERM;
 
-	return !!(readl_relaxed(port->membase + ofs->cr3) & USART_CR3_DMAR);
+	dma_status = dmaengine_tx_status(chan, chan->cookie, NULL);
+	if (dma_status != expected_status)
+		return -EAGAIN;
+
+	ret = dmaengine_pause_or_resume(chan);
+	if (ret) {
+		dev_err(port->dev, "DMA failed with error code: %d\n", ret);
+		stm32_usart_xx_dma_terminate(stm32_port);
+	}
+	return ret;
+}
+
+static int stm32_usart_rx_dma_pause(struct stm32_port *stm32_port)
+{
+	return stm32_usart_dma_pause_resume(stm32_port, stm32_port->rx_ch,
+					    DMA_IN_PROGRESS, dmaengine_pause,
+					    stm32_usart_rx_dma_started,
+					    stm32_usart_rx_dma_terminate);
+}
+
+static int stm32_usart_rx_dma_resume(struct stm32_port *stm32_port)
+{
+	return stm32_usart_dma_pause_resume(stm32_port, stm32_port->rx_ch,
+					    DMA_PAUSED, dmaengine_resume,
+					    stm32_usart_rx_dma_started,
+					    stm32_usart_rx_dma_terminate);
 }
 
 /* Return true when data is pending (in pio mode), and false when no data is pending. */
@@ -306,7 +354,7 @@ static bool stm32_usart_pending_rx_pio(struct uart_port *port, u32 *sr)
 	/* Get pending characters in RDR or FIFO */
 	if (*sr & USART_SR_RXNE) {
 		/* Get all pending characters from the RDR or the FIFO when using interrupts */
-		if (!stm32_usart_rx_dma_enabled(port))
+		if (!stm32_usart_rx_dma_started(stm32_port))
 			return true;
 
 		/* Handle only RX data errors when using DMA */
@@ -317,7 +365,7 @@ static bool stm32_usart_pending_rx_pio(struct uart_port *port, u32 *sr)
 	return false;
 }
 
-static unsigned long stm32_usart_get_char_pio(struct uart_port *port)
+static u8 stm32_usart_get_char_pio(struct uart_port *port)
 {
 	struct stm32_port *stm32_port = to_stm32_port(port);
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
@@ -334,10 +382,9 @@ static unsigned int stm32_usart_receive_chars_pio(struct uart_port *port)
 {
 	struct stm32_port *stm32_port = to_stm32_port(port);
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
-	unsigned long c;
 	unsigned int size = 0;
 	u32 sr;
-	char flag;
+	u8 c, flag;
 
 	while (stm32_usart_pending_rx_pio(port, &sr)) {
 		sr |= USART_SR_DUMMY_RX;
@@ -452,11 +499,12 @@ static unsigned int stm32_usart_receive_chars(struct uart_port *port, bool force
 	u32 sr;
 	unsigned int size = 0;
 
-	if (stm32_usart_rx_dma_enabled(port) || force_dma_flush) {
+	if (stm32_usart_rx_dma_started(stm32_port) || force_dma_flush) {
 		rx_dma_status = dmaengine_tx_status(stm32_port->rx_ch,
 						    stm32_port->rx_ch->cookie,
 						    &stm32_port->rx_dma_state);
-		if (rx_dma_status == DMA_IN_PROGRESS) {
+		if (rx_dma_status == DMA_IN_PROGRESS ||
+		    rx_dma_status == DMA_PAUSED) {
 			/* Empty DMA buffer */
 			size = stm32_usart_receive_chars_dma(port);
 			sr = readl_relaxed(port->membase + ofs->isr);
@@ -472,8 +520,7 @@ static unsigned int stm32_usart_receive_chars(struct uart_port *port, bool force
 			}
 		} else {
 			/* Disable RX DMA */
-			dmaengine_terminate_async(stm32_port->rx_ch);
-			stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAR);
+			stm32_usart_rx_dma_terminate(stm32_port);
 			/* Fall back to interrupt mode */
 			dev_dbg(port->dev, "DMA error, fallback to irq mode\n");
 			size = stm32_usart_receive_chars_pio(port);
@@ -485,6 +532,76 @@ static unsigned int stm32_usart_receive_chars(struct uart_port *port, bool force
 	return size;
 }
 
+static void stm32_usart_rx_dma_complete(void *arg)
+{
+	struct uart_port *port = arg;
+	struct tty_port *tport = &port->state->port;
+	unsigned int size;
+	unsigned long flags;
+
+	uart_port_lock_irqsave(port, &flags);
+	size = stm32_usart_receive_chars(port, false);
+	uart_unlock_and_check_sysrq_irqrestore(port, flags);
+	if (size)
+		tty_flip_buffer_push(tport);
+}
+
+static int stm32_usart_rx_dma_start_or_resume(struct uart_port *port)
+{
+	struct stm32_port *stm32_port = to_stm32_port(port);
+	struct dma_async_tx_descriptor *desc;
+	enum dma_status rx_dma_status;
+	int ret;
+
+	if (stm32_port->throttled)
+		return 0;
+
+	if (stm32_port->rx_dma_busy) {
+		rx_dma_status = dmaengine_tx_status(stm32_port->rx_ch,
+						    stm32_port->rx_ch->cookie,
+						    NULL);
+		if (rx_dma_status == DMA_IN_PROGRESS)
+			return 0;
+
+		if (rx_dma_status == DMA_PAUSED && !stm32_usart_rx_dma_resume(stm32_port))
+			return 0;
+
+		dev_err(port->dev, "DMA failed : status error.\n");
+		stm32_usart_rx_dma_terminate(stm32_port);
+	}
+
+	stm32_port->rx_dma_busy = true;
+
+	stm32_port->last_res = RX_BUF_L;
+	/* Prepare a DMA cyclic transaction */
+	desc = dmaengine_prep_dma_cyclic(stm32_port->rx_ch,
+					 stm32_port->rx_dma_buf,
+					 RX_BUF_L, RX_BUF_P,
+					 DMA_DEV_TO_MEM,
+					 DMA_PREP_INTERRUPT);
+	if (!desc) {
+		dev_err(port->dev, "rx dma prep cyclic failed\n");
+		stm32_port->rx_dma_busy = false;
+		return -ENODEV;
+	}
+
+	desc->callback = stm32_usart_rx_dma_complete;
+	desc->callback_param = port;
+
+	/* Push current DMA transaction in the pending queue */
+	ret = dma_submit_error(dmaengine_submit(desc));
+	if (ret) {
+		dmaengine_terminate_sync(stm32_port->rx_ch);
+		stm32_port->rx_dma_busy = false;
+		return ret;
+	}
+
+	/* Issue pending DMA requests */
+	dma_async_issue_pending(stm32_port->rx_ch);
+
+	return 0;
+}
+
 static void stm32_usart_tx_dma_terminate(struct stm32_port *stm32_port)
 {
 	dmaengine_terminate_async(stm32_port->tx_ch);
@@ -503,27 +620,34 @@ static bool stm32_usart_tx_dma_started(struct stm32_port *stm32_port)
 	return stm32_port->tx_dma_busy;
 }
 
-static bool stm32_usart_tx_dma_enabled(struct stm32_port *stm32_port)
+static int stm32_usart_tx_dma_pause(struct stm32_port *stm32_port)
 {
-	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
+	return stm32_usart_dma_pause_resume(stm32_port, stm32_port->tx_ch,
+					    DMA_IN_PROGRESS, dmaengine_pause,
+					    stm32_usart_tx_dma_started,
+					    stm32_usart_tx_dma_terminate);
+}
 
-	return !!(readl_relaxed(stm32_port->port.membase + ofs->cr3) & USART_CR3_DMAT);
+static int stm32_usart_tx_dma_resume(struct stm32_port *stm32_port)
+{
+	return stm32_usart_dma_pause_resume(stm32_port, stm32_port->tx_ch,
+					    DMA_PAUSED, dmaengine_resume,
+					    stm32_usart_tx_dma_started,
+					    stm32_usart_tx_dma_terminate);
 }
 
 static void stm32_usart_tx_dma_complete(void *arg)
 {
 	struct uart_port *port = arg;
 	struct stm32_port *stm32port = to_stm32_port(port);
-	const struct stm32_usart_offsets *ofs = &stm32port->info->ofs;
 	unsigned long flags;
 
-	stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
 	stm32_usart_tx_dma_terminate(stm32port);
 
 	/* Let's see if we have pending data to send */
-	spin_lock_irqsave(&port->lock, flags);
+	uart_port_lock_irqsave(port, &flags);
 	stm32_usart_transmit_chars(port);
-	spin_unlock_irqrestore(&port->lock, flags);
+	uart_port_unlock_irqrestore(port, flags);
 }
 
 static void stm32_usart_tx_interrupt_enable(struct uart_port *port)
@@ -549,20 +673,6 @@ static void stm32_usart_tc_interrupt_enable(struct uart_port *port)
 	stm32_usart_set_bits(port, ofs->cr1, USART_CR1_TCIE);
 }
 
-static void stm32_usart_rx_dma_complete(void *arg)
-{
-	struct uart_port *port = arg;
-	struct tty_port *tport = &port->state->port;
-	unsigned int size;
-	unsigned long flags;
-
-	spin_lock_irqsave(&port->lock, flags);
-	size = stm32_usart_receive_chars(port, false);
-	uart_unlock_and_check_sysrq_irqrestore(port, flags);
-	if (size)
-		tty_flip_buffer_push(tport);
-}
-
 static void stm32_usart_tx_interrupt_disable(struct uart_port *port)
 {
 	struct stm32_port *stm32_port = to_stm32_port(port);
@@ -586,21 +696,23 @@ static void stm32_usart_transmit_chars_pio(struct uart_port *port)
 {
 	struct stm32_port *stm32_port = to_stm32_port(port);
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
-	struct circ_buf *xmit = &port->state->xmit;
+	struct tty_port *tport = &port->state->port;
 
-	if (stm32_usart_tx_dma_enabled(stm32_port))
-		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
+	while (1) {
+		unsigned char ch;
 
-	while (!uart_circ_empty(xmit)) {
 		/* Check that TDR is empty before filling FIFO */
 		if (!(readl_relaxed(port->membase + ofs->isr) & USART_SR_TXE))
 			break;
-		writel_relaxed(xmit->buf[xmit->tail], port->membase + ofs->tdr);
-		uart_xmit_advance(port, 1);
+
+		if (!uart_fifo_get(port, &ch))
+			break;
+
+		writel_relaxed(ch, port->membase + ofs->tdr);
 	}
 
 	/* rely on TXE irq (mask or unmask) for sending remaining data */
-	if (uart_circ_empty(xmit))
+	if (kfifo_is_empty(&tport->xmit_fifo))
 		stm32_usart_tx_interrupt_disable(port);
 	else
 		stm32_usart_tx_interrupt_enable(port);
@@ -609,36 +721,20 @@ static void stm32_usart_transmit_chars_pio(struct uart_port *port)
 static void stm32_usart_transmit_chars_dma(struct uart_port *port)
 {
 	struct stm32_port *stm32port = to_stm32_port(port);
-	const struct stm32_usart_offsets *ofs = &stm32port->info->ofs;
-	struct circ_buf *xmit = &port->state->xmit;
+	struct tty_port *tport = &port->state->port;
 	struct dma_async_tx_descriptor *desc = NULL;
 	unsigned int count;
+	int ret;
 
 	if (stm32_usart_tx_dma_started(stm32port)) {
-		if (!stm32_usart_tx_dma_enabled(stm32port))
-			stm32_usart_set_bits(port, ofs->cr3, USART_CR3_DMAT);
+		ret = stm32_usart_tx_dma_resume(stm32port);
+		if (ret < 0 && ret != -EAGAIN)
+			goto fallback_err;
 		return;
 	}
 
-	count = uart_circ_chars_pending(xmit);
-
-	if (count > TX_BUF_L)
-		count = TX_BUF_L;
-
-	if (xmit->tail < xmit->head) {
-		memcpy(&stm32port->tx_buf[0], &xmit->buf[xmit->tail], count);
-	} else {
-		size_t one = UART_XMIT_SIZE - xmit->tail;
-		size_t two;
-
-		if (one > count)
-			one = count;
-		two = count - one;
-
-		memcpy(&stm32port->tx_buf[0], &xmit->buf[xmit->tail], one);
-		if (two)
-			memcpy(&stm32port->tx_buf[one], &xmit->buf[0], two);
-	}
+	count =	kfifo_out_peek(&tport->xmit_fifo, &stm32port->tx_buf[0],
+			TX_BUF_L);
 
 	desc = dmaengine_prep_slave_single(stm32port->tx_ch,
 					   stm32port->tx_dma_buf,
@@ -661,8 +757,10 @@ static void stm32_usart_transmit_chars_dma(struct uart_port *port)
 	desc->callback_param = port;
 
 	/* Push current DMA TX transaction in the pending queue */
-	if (dma_submit_error(dmaengine_submit(desc))) {
-		/* dma no yet started, safe to free resources */
+	/* DMA no yet started, safe to free resources */
+	ret = dma_submit_error(dmaengine_submit(desc));
+	if (ret) {
+		dev_err(port->dev, "DMA failed with error code: %d\n", ret);
 		stm32_usart_tx_dma_terminate(stm32port);
 		goto fallback_err;
 	}
@@ -670,8 +768,6 @@ static void stm32_usart_transmit_chars_dma(struct uart_port *port)
 	/* Issue pending DMA TX requests */
 	dma_async_issue_pending(stm32port->tx_ch);
 
-	stm32_usart_set_bits(port, ofs->cr3, USART_CR3_DMAT);
-
 	uart_xmit_advance(port, count);
 
 	return;
@@ -684,21 +780,21 @@ static void stm32_usart_transmit_chars(struct uart_port *port)
 {
 	struct stm32_port *stm32_port = to_stm32_port(port);
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
-	struct circ_buf *xmit = &port->state->xmit;
+	struct tty_port *tport = &port->state->port;
 	u32 isr;
 	int ret;
 
 	if (!stm32_port->hw_flow_control &&
-	    port->rs485.flags & SER_RS485_ENABLED) {
-		stm32_port->txdone = false;
+	    port->rs485.flags & SER_RS485_ENABLED &&
+	    (port->x_char ||
+	     !(kfifo_is_empty(&tport->xmit_fifo) || uart_tx_stopped(port)))) {
 		stm32_usart_tc_interrupt_disable(port);
 		stm32_usart_rs485_rts_enable(port);
 	}
 
 	if (port->x_char) {
-		if (stm32_usart_tx_dma_started(stm32_port) &&
-		    stm32_usart_tx_dma_enabled(stm32_port))
-			stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
+		/* dma terminate may have been called in case of dma pause failure */
+		stm32_usart_tx_dma_pause(stm32_port);
 
 		/* Check that TDR is empty before filling FIFO */
 		ret =
@@ -712,12 +808,13 @@ static void stm32_usart_transmit_chars(struct uart_port *port)
 		writel_relaxed(port->x_char, port->membase + ofs->tdr);
 		port->x_char = 0;
 		port->icount.tx++;
-		if (stm32_usart_tx_dma_started(stm32_port))
-			stm32_usart_set_bits(port, ofs->cr3, USART_CR3_DMAT);
+
+		/* dma terminate may have been called in case of dma resume failure */
+		stm32_usart_tx_dma_resume(stm32_port);
 		return;
 	}
 
-	if (uart_circ_empty(xmit) || uart_tx_stopped(port)) {
+	if (kfifo_is_empty(&tport->xmit_fifo) || uart_tx_stopped(port)) {
 		stm32_usart_tx_interrupt_disable(port);
 		return;
 	}
@@ -732,14 +829,13 @@ static void stm32_usart_transmit_chars(struct uart_port *port)
 	else
 		stm32_usart_transmit_chars_pio(port);
 
-	if (uart_circ_chars_pending(xmit) < WAKEUP_CHARS)
+	if (kfifo_len(&tport->xmit_fifo) < WAKEUP_CHARS)
 		uart_write_wakeup(port);
 
-	if (uart_circ_empty(xmit)) {
+	if (kfifo_is_empty(&tport->xmit_fifo)) {
 		stm32_usart_tx_interrupt_disable(port);
 		if (!stm32_port->hw_flow_control &&
 		    port->rs485.flags & SER_RS485_ENABLED) {
-			stm32_port->txdone = true;
 			stm32_usart_tc_interrupt_enable(port);
 		}
 	}
@@ -753,6 +849,7 @@ static irqreturn_t stm32_usart_interrupt(int irq, void *ptr)
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 	u32 sr;
 	unsigned int size;
+	irqreturn_t ret = IRQ_NONE;
 
 	sr = readl_relaxed(port->membase + ofs->isr);
 
@@ -761,11 +858,14 @@ static irqreturn_t stm32_usart_interrupt(int irq, void *ptr)
 	    (sr & USART_SR_TC)) {
 		stm32_usart_tc_interrupt_disable(port);
 		stm32_usart_rs485_rts_disable(port);
+		ret = IRQ_HANDLED;
 	}
 
-	if ((sr & USART_SR_RTOF) && ofs->icr != UNDEF_REG)
+	if ((sr & USART_SR_RTOF) && ofs->icr != UNDEF_REG) {
 		writel_relaxed(USART_ICR_RTOCF,
 			       port->membase + ofs->icr);
+		ret = IRQ_HANDLED;
+	}
 
 	if ((sr & USART_SR_WUF) && ofs->icr != UNDEF_REG) {
 		/* Clear wake up flag and disable wake up interrupt */
@@ -774,6 +874,7 @@ static irqreturn_t stm32_usart_interrupt(int irq, void *ptr)
 		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_WUFIE);
 		if (irqd_is_wakeup_set(irq_get_irq_data(port->irq)))
 			pm_wakeup_event(tport->tty->dev, 0);
+		ret = IRQ_HANDLED;
 	}
 
 	/*
@@ -781,46 +882,35 @@ static irqreturn_t stm32_usart_interrupt(int irq, void *ptr)
 	 * line has been masked by HW and rx data are stacking in FIFO.
 	 */
 	if (!stm32_port->throttled) {
-		if (((sr & USART_SR_RXNE) && !stm32_usart_rx_dma_enabled(port)) ||
-		    ((sr & USART_SR_ERR_MASK) && stm32_usart_rx_dma_enabled(port))) {
-			spin_lock(&port->lock);
+		if (((sr & USART_SR_RXNE) && !stm32_usart_rx_dma_started(stm32_port)) ||
+		    ((sr & USART_SR_ERR_MASK) && stm32_usart_rx_dma_started(stm32_port))) {
+			uart_port_lock(port);
 			size = stm32_usart_receive_chars(port, false);
 			uart_unlock_and_check_sysrq(port);
 			if (size)
 				tty_flip_buffer_push(tport);
+			ret = IRQ_HANDLED;
 		}
 	}
 
 	if ((sr & USART_SR_TXE) && !(stm32_port->tx_ch)) {
-		spin_lock(&port->lock);
+		uart_port_lock(port);
 		stm32_usart_transmit_chars(port);
-		spin_unlock(&port->lock);
+		uart_port_unlock(port);
+		ret = IRQ_HANDLED;
 	}
 
-	if (stm32_usart_rx_dma_enabled(port))
-		return IRQ_WAKE_THREAD;
-	else
-		return IRQ_HANDLED;
-}
-
-static irqreturn_t stm32_usart_threaded_interrupt(int irq, void *ptr)
-{
-	struct uart_port *port = ptr;
-	struct tty_port *tport = &port->state->port;
-	struct stm32_port *stm32_port = to_stm32_port(port);
-	unsigned int size;
-	unsigned long flags;
-
 	/* Receiver timeout irq for DMA RX */
-	if (!stm32_port->throttled) {
-		spin_lock_irqsave(&port->lock, flags);
+	if (stm32_usart_rx_dma_started(stm32_port) && !stm32_port->throttled) {
+		uart_port_lock(port);
 		size = stm32_usart_receive_chars(port, false);
-		uart_unlock_and_check_sysrq_irqrestore(port, flags);
+		uart_unlock_and_check_sysrq(port);
 		if (size)
 			tty_flip_buffer_push(tport);
+		ret = IRQ_HANDLED;
 	}
 
-	return IRQ_HANDLED;
+	return ret;
 }
 
 static void stm32_usart_set_mctrl(struct uart_port *port, unsigned int mctrl)
@@ -854,18 +944,18 @@ static void stm32_usart_enable_ms(struct uart_port *port)
 
 static void stm32_usart_disable_ms(struct uart_port *port)
 {
-	mctrl_gpio_disable_ms(to_stm32_port(port)->gpios);
+	mctrl_gpio_disable_ms_sync(to_stm32_port(port)->gpios);
 }
 
 /* Transmit stop */
 static void stm32_usart_stop_tx(struct uart_port *port)
 {
 	struct stm32_port *stm32_port = to_stm32_port(port);
-	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 
 	stm32_usart_tx_interrupt_disable(port);
-	if (stm32_usart_tx_dma_started(stm32_port) && stm32_usart_tx_dma_enabled(stm32_port))
-		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
+
+	/* dma terminate may have been called in case of dma pause failure */
+	stm32_usart_tx_dma_pause(stm32_port);
 
 	stm32_usart_rs485_rts_disable(port);
 }
@@ -873,12 +963,10 @@ static void stm32_usart_stop_tx(struct uart_port *port)
 /* There are probably characters waiting to be transmitted. */
 static void stm32_usart_start_tx(struct uart_port *port)
 {
-	struct circ_buf *xmit = &port->state->xmit;
+	struct tty_port *tport = &port->state->port;
 
-	if (uart_circ_empty(xmit) && !port->x_char) {
-		stm32_usart_rs485_rts_disable(port);
+	if (kfifo_is_empty(&tport->xmit_fifo) && !port->x_char)
 		return;
-	}
 
 	stm32_usart_rs485_rts_enable(port);
 
@@ -889,12 +977,9 @@ static void stm32_usart_start_tx(struct uart_port *port)
 static void stm32_usart_flush_buffer(struct uart_port *port)
 {
 	struct stm32_port *stm32_port = to_stm32_port(port);
-	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 
-	if (stm32_port->tx_ch) {
+	if (stm32_port->tx_ch)
 		stm32_usart_tx_dma_terminate(stm32_port);
-		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
-	}
 }
 
 /* Throttle the remote when input buffer is about to overflow. */
@@ -904,21 +989,20 @@ static void stm32_usart_throttle(struct uart_port *port)
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 	unsigned long flags;
 
-	spin_lock_irqsave(&port->lock, flags);
+	uart_port_lock_irqsave(port, &flags);
 
 	/*
-	 * Disable DMA request line if enabled, so the RX data gets queued into the FIFO.
+	 * Pause DMA transfer, so the RX data gets queued into the FIFO.
 	 * Hardware flow control is triggered when RX FIFO is full.
 	 */
-	if (stm32_usart_rx_dma_enabled(port))
-		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAR);
+	stm32_usart_rx_dma_pause(stm32_port);
 
 	stm32_usart_clr_bits(port, ofs->cr1, stm32_port->cr1_irq);
 	if (stm32_port->cr3_irq)
 		stm32_usart_clr_bits(port, ofs->cr3, stm32_port->cr3_irq);
 
 	stm32_port->throttled = true;
-	spin_unlock_irqrestore(&port->lock, flags);
+	uart_port_unlock_irqrestore(port, flags);
 }
 
 /* Unthrottle the remote, the input buffer can now accept data. */
@@ -928,20 +1012,21 @@ static void stm32_usart_unthrottle(struct uart_port *port)
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 	unsigned long flags;
 
-	spin_lock_irqsave(&port->lock, flags);
+	uart_port_lock_irqsave(port, &flags);
 	stm32_usart_set_bits(port, ofs->cr1, stm32_port->cr1_irq);
 	if (stm32_port->cr3_irq)
 		stm32_usart_set_bits(port, ofs->cr3, stm32_port->cr3_irq);
 
+	stm32_port->throttled = false;
+
 	/*
-	 * Switch back to DMA mode (re-enable DMA request line).
+	 * Switch back to DMA mode (resume DMA).
 	 * Hardware flow control is stopped when FIFO is not full any more.
 	 */
 	if (stm32_port->rx_ch)
-		stm32_usart_set_bits(port, ofs->cr3, USART_CR3_DMAR);
+		stm32_usart_rx_dma_start_or_resume(port);
 
-	stm32_port->throttled = false;
-	spin_unlock_irqrestore(&port->lock, flags);
+	uart_port_unlock_irqrestore(port, flags);
 }
 
 /* Receive stop */
@@ -951,59 +1036,27 @@ static void stm32_usart_stop_rx(struct uart_port *port)
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 
 	/* Disable DMA request line. */
-	if (stm32_port->rx_ch)
-		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAR);
+	stm32_usart_rx_dma_pause(stm32_port);
 
 	stm32_usart_clr_bits(port, ofs->cr1, stm32_port->cr1_irq);
 	if (stm32_port->cr3_irq)
 		stm32_usart_clr_bits(port, ofs->cr3, stm32_port->cr3_irq);
 }
 
-/* Handle breaks - ignored by us */
 static void stm32_usart_break_ctl(struct uart_port *port, int break_state)
 {
-}
-
-static int stm32_usart_start_rx_dma_cyclic(struct uart_port *port)
-{
 	struct stm32_port *stm32_port = to_stm32_port(port);
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
-	struct dma_async_tx_descriptor *desc;
-	int ret;
-
-	stm32_port->last_res = RX_BUF_L;
-	/* Prepare a DMA cyclic transaction */
-	desc = dmaengine_prep_dma_cyclic(stm32_port->rx_ch,
-					 stm32_port->rx_dma_buf,
-					 RX_BUF_L, RX_BUF_P,
-					 DMA_DEV_TO_MEM,
-					 DMA_PREP_INTERRUPT);
-	if (!desc) {
-		dev_err(port->dev, "rx dma prep cyclic failed\n");
-		return -ENODEV;
-	}
-
-	desc->callback = stm32_usart_rx_dma_complete;
-	desc->callback_param = port;
-
-	/* Push current DMA transaction in the pending queue */
-	ret = dma_submit_error(dmaengine_submit(desc));
-	if (ret) {
-		dmaengine_terminate_sync(stm32_port->rx_ch);
-		return ret;
-	}
+	unsigned long flags;
 
-	/* Issue pending DMA requests */
-	dma_async_issue_pending(stm32_port->rx_ch);
+	uart_port_lock_irqsave(port, &flags);
 
-	/*
-	 * DMA request line not re-enabled at resume when port is throttled.
-	 * It will be re-enabled by unthrottle ops.
-	 */
-	if (!stm32_port->throttled)
-		stm32_usart_set_bits(port, ofs->cr3, USART_CR3_DMAR);
+	if (break_state)
+		stm32_usart_set_bits(port, ofs->rqr, USART_RQR_SBKRQ);
+	else
+		stm32_usart_clr_bits(port, ofs->rqr, USART_RQR_SBKRQ);
 
-	return 0;
+	uart_port_unlock_irqrestore(port, flags);
 }
 
 static int stm32_usart_startup(struct uart_port *port)
@@ -1015,10 +1068,8 @@ static int stm32_usart_startup(struct uart_port *port)
 	u32 val;
 	int ret;
 
-	ret = request_threaded_irq(port->irq, stm32_usart_interrupt,
-				   stm32_usart_threaded_interrupt,
-				   IRQF_ONESHOT | IRQF_NO_SUSPEND,
-				   name, port);
+	ret = request_irq(port->irq, stm32_usart_interrupt,
+			  IRQF_NO_SUSPEND, name, port);
 	if (ret)
 		return ret;
 
@@ -1027,13 +1078,14 @@ static int stm32_usart_startup(struct uart_port *port)
 		val |= USART_CR2_SWAP;
 		writel_relaxed(val, port->membase + ofs->cr2);
 	}
+	stm32_port->throttled = false;
 
 	/* RX FIFO Flush */
 	if (ofs->rqr != UNDEF_REG)
 		writel_relaxed(USART_RQR_RXFRQ, port->membase + ofs->rqr);
 
 	if (stm32_port->rx_ch) {
-		ret = stm32_usart_start_rx_dma_cyclic(port);
+		ret = stm32_usart_rx_dma_start_or_resume(port);
 		if (ret) {
 			free_irq(port->irq, port);
 			return ret;
@@ -1055,12 +1107,12 @@ static void stm32_usart_shutdown(struct uart_port *port)
 	u32 val, isr;
 	int ret;
 
-	if (stm32_usart_tx_dma_enabled(stm32_port))
-		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
-
 	if (stm32_usart_tx_dma_started(stm32_port))
 		stm32_usart_tx_dma_terminate(stm32_port);
 
+	if (stm32_port->tx_ch)
+		stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
+
 	/* Disable modem control interrupts */
 	stm32_usart_disable_ms(port);
 
@@ -1079,8 +1131,10 @@ static void stm32_usart_shutdown(struct uart_port *port)
 		dev_err(port->dev, "Transmission is not complete\n");
 
 	/* Disable RX DMA. */
-	if (stm32_port->rx_ch)
-		dmaengine_terminate_async(stm32_port->rx_ch);
+	if (stm32_port->rx_ch) {
+		stm32_usart_rx_dma_terminate(stm32_port);
+		dmaengine_synchronize(stm32_port->rx_ch);
+	}
 
 	/* flush RX & TX FIFO */
 	if (ofs->rqr != UNDEF_REG)
@@ -1092,6 +1146,8 @@ static void stm32_usart_shutdown(struct uart_port *port)
 	free_irq(port->irq, port);
 }
 
+static const unsigned int stm32_usart_presc_val[] = {1, 2, 4, 6, 8, 10, 12, 16, 32, 64, 128, 256};
+
 static void stm32_usart_set_termios(struct uart_port *port,
 				    struct ktermios *termios,
 				    const struct ktermios *old)
@@ -1100,19 +1156,21 @@ static void stm32_usart_set_termios(struct uart_port *port,
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 	const struct stm32_usart_config *cfg = &stm32_port->info->cfg;
 	struct serial_rs485 *rs485conf = &port->rs485;
-	unsigned int baud, bits;
+	unsigned int baud, bits, uart_clk, uart_clk_pres;
 	u32 usartdiv, mantissa, fraction, oversampling;
 	tcflag_t cflag = termios->c_cflag;
-	u32 cr1, cr2, cr3, isr;
+	u32 cr1, cr2, cr3, isr, brr, presc;
 	unsigned long flags;
 	int ret;
 
 	if (!stm32_port->hw_flow_control)
 		cflag &= ~CRTSCTS;
 
-	baud = uart_get_baud_rate(port, termios, old, 0, port->uartclk / 8);
+	uart_clk = clk_get_rate(stm32_port->clk);
 
-	spin_lock_irqsave(&port->lock, flags);
+	baud = uart_get_baud_rate(port, termios, old, 0, uart_clk / 8);
+
+	uart_port_lock_irqsave(port, &flags);
 
 	ret = readl_relaxed_poll_timeout_atomic(port->membase + ofs->isr,
 						isr,
@@ -1212,27 +1270,48 @@ static void stm32_usart_set_termios(struct uart_port *port,
 		cr3 |= USART_CR3_CTSE | USART_CR3_RTSE;
 	}
 
-	usartdiv = DIV_ROUND_CLOSEST(port->uartclk, baud);
+	for (presc = 0; presc <= USART_PRESC_MAX; presc++) {
+		uart_clk_pres = DIV_ROUND_CLOSEST(uart_clk, stm32_usart_presc_val[presc]);
+		usartdiv = DIV_ROUND_CLOSEST(uart_clk_pres, baud);
 
-	/*
-	 * The USART supports 16 or 8 times oversampling.
-	 * By default we prefer 16 times oversampling, so that the receiver
-	 * has a better tolerance to clock deviations.
-	 * 8 times oversampling is only used to achieve higher speeds.
-	 */
-	if (usartdiv < 16) {
-		oversampling = 8;
-		cr1 |= USART_CR1_OVER8;
-		stm32_usart_set_bits(port, ofs->cr1, USART_CR1_OVER8);
-	} else {
-		oversampling = 16;
-		cr1 &= ~USART_CR1_OVER8;
-		stm32_usart_clr_bits(port, ofs->cr1, USART_CR1_OVER8);
+		/*
+		 * The USART supports 16 or 8 times oversampling.
+		 * By default we prefer 16 times oversampling, so that the receiver
+		 * has a better tolerance to clock deviations.
+		 * 8 times oversampling is only used to achieve higher speeds.
+		 */
+		if (usartdiv < 16) {
+			oversampling = 8;
+			cr1 |= USART_CR1_OVER8;
+			stm32_usart_set_bits(port, ofs->cr1, USART_CR1_OVER8);
+		} else {
+			oversampling = 16;
+			cr1 &= ~USART_CR1_OVER8;
+			stm32_usart_clr_bits(port, ofs->cr1, USART_CR1_OVER8);
+		}
+
+		mantissa = (usartdiv / oversampling) << USART_BRR_DIV_M_SHIFT;
+		fraction = usartdiv % oversampling;
+		brr = mantissa | fraction;
+
+		if (FIELD_FIT(USART_BRR_MASK, brr)) {
+			if (ofs->presc != UNDEF_REG) {
+				port->uartclk = uart_clk_pres;
+				writel_relaxed(presc, port->membase + ofs->presc);
+			} else if (presc) {
+				/* We need a prescaler but we don't have it (STM32F4, STM32F7) */
+				dev_err(port->dev,
+					"unable to set baudrate, input clock is too high");
+			}
+			break;
+		} else if (presc == USART_PRESC_MAX) {
+			/* Even with prescaler and brr at max value we can't set baudrate */
+			dev_err(port->dev, "unable to set baudrate, input clock is too high");
+			break;
+		}
 	}
 
-	mantissa = (usartdiv / oversampling) << USART_BRR_DIV_M_SHIFT;
-	fraction = usartdiv % oversampling;
-	writel_relaxed(mantissa | fraction, port->membase + ofs->brr);
+	writel_relaxed(brr, port->membase + ofs->brr);
 
 	uart_update_timeout(port, cflag, baud);
 
@@ -1271,6 +1350,9 @@ static void stm32_usart_set_termios(struct uart_port *port,
 		cr3 |= USART_CR3_DDRE;
 	}
 
+	if (stm32_port->tx_ch)
+		cr3 |= USART_CR3_DMAT;
+
 	if (rs485conf->flags & SER_RS485_ENABLED) {
 		stm32_usart_config_reg_rs485(&cr1, &cr3,
 					     rs485conf->delay_rts_before_send,
@@ -1300,7 +1382,7 @@ static void stm32_usart_set_termios(struct uart_port *port,
 	writel_relaxed(cr1, port->membase + ofs->cr1);
 
 	stm32_usart_set_bits(port, ofs->cr1, BIT(cfg->uart_enable_bit));
-	spin_unlock_irqrestore(&port->lock, flags);
+	uart_port_unlock_irqrestore(port, flags);
 
 	/* Handle modem control interrupts */
 	if (UART_ENABLE_MS(port, termios->c_cflag))
@@ -1350,9 +1432,9 @@ static void stm32_usart_pm(struct uart_port *port, unsigned int state,
 		pm_runtime_get_sync(port->dev);
 		break;
 	case UART_PM_STATE_OFF:
-		spin_lock_irqsave(&port->lock, flags);
+		uart_port_lock_irqsave(port, &flags);
 		stm32_usart_clr_bits(port, ofs->cr1, BIT(cfg->uart_enable_bit));
-		spin_unlock_irqrestore(&port->lock, flags);
+		uart_port_unlock_irqrestore(port, flags);
 		pm_runtime_put_sync(port->dev);
 		break;
 	}
@@ -1413,37 +1495,57 @@ static const struct uart_ops stm32_uart_ops = {
 #endif /* CONFIG_CONSOLE_POLL */
 };
 
-/*
- * STM32H7 RX & TX FIFO threshold configuration (CR3 RXFTCFG / TXFTCFG)
- * Note: 1 isn't a valid value in RXFTCFG / TXFTCFG. In this case,
- * RXNEIE / TXEIE can be used instead of threshold irqs: RXFTIE / TXFTIE.
- * So, RXFTCFG / TXFTCFG bitfields values are encoded as array index + 1.
- */
-static const u32 stm32h7_usart_fifo_thresh_cfg[] = { 1, 2, 4, 8, 12, 14, 16 };
+struct stm32_usart_thresh_ratio {
+	int mul;
+	int div;
+};
 
-static void stm32_usart_get_ftcfg(struct platform_device *pdev, const char *p,
-				  int *ftcfg)
+static const struct stm32_usart_thresh_ratio stm32h7_usart_fifo_thresh_cfg[] = {
+	{1, 8}, {1, 4}, {1, 2}, {3, 4}, {7, 8}, {1, 1} };
+
+static int stm32_usart_get_thresh_value(u32 fifo_size, int index)
 {
-	u32 bytes, i;
+	return fifo_size * stm32h7_usart_fifo_thresh_cfg[index].mul /
+		stm32h7_usart_fifo_thresh_cfg[index].div;
+}
 
-	/* DT option to get RX & TX FIFO threshold (default to 8 bytes) */
+static int stm32_usart_get_ftcfg(struct platform_device *pdev, struct stm32_port *stm32port,
+				 const char *p, int *ftcfg)
+{
+	const struct stm32_usart_offsets *ofs = &stm32port->info->ofs;
+	u32 bytes, i, cfg8;
+	int fifo_size;
+
+	if (WARN_ON(ofs->hwcfgr1 == UNDEF_REG))
+		return 1;
+
+	cfg8 = FIELD_GET(USART_HWCFGR1_CFG8,
+			 readl_relaxed(stm32port->port.membase + ofs->hwcfgr1));
+
+	/* On STM32H7, hwcfgr is not present, so returned value will be 0 */
+	fifo_size = cfg8 ? 1 << cfg8 : STM32H7_USART_FIFO_SIZE;
+
+	/* DT option to get RX & TX FIFO threshold (default to half fifo size) */
 	if (of_property_read_u32(pdev->dev.of_node, p, &bytes))
-		bytes = 8;
+		bytes = fifo_size / 2;
+
+	if (bytes < stm32_usart_get_thresh_value(fifo_size, 0)) {
+		*ftcfg = -EINVAL;
+		return fifo_size;
+	}
 
-	for (i = 0; i < ARRAY_SIZE(stm32h7_usart_fifo_thresh_cfg); i++)
-		if (stm32h7_usart_fifo_thresh_cfg[i] >= bytes)
+	for (i = 0; i < ARRAY_SIZE(stm32h7_usart_fifo_thresh_cfg); i++) {
+		if (stm32_usart_get_thresh_value(fifo_size, i) >= bytes)
 			break;
+	}
 	if (i >= ARRAY_SIZE(stm32h7_usart_fifo_thresh_cfg))
 		i = ARRAY_SIZE(stm32h7_usart_fifo_thresh_cfg) - 1;
 
-	dev_dbg(&pdev->dev, "%s set to %d bytes\n", p,
-		stm32h7_usart_fifo_thresh_cfg[i]);
+	dev_dbg(&pdev->dev, "%s set to %d/%d bytes\n", p,
+		stm32_usart_get_thresh_value(fifo_size, i), fifo_size);
 
-	/* Provide FIFO threshold ftcfg (1 is invalid: threshold irq unused) */
-	if (i)
-		*ftcfg = i - 1;
-	else
-		*ftcfg = -EINVAL;
+	*ftcfg = i;
+	return fifo_size;
 }
 
 static void stm32_usart_deinit_port(struct stm32_port *stm32port)
@@ -1473,7 +1575,6 @@ static int stm32_usart_init_port(struct stm32_port *stm32port,
 	port->flags	= UPF_BOOT_AUTOCONF;
 	port->ops	= &stm32_uart_ops;
 	port->dev	= &pdev->dev;
-	port->fifosize	= stm32port->info->cfg.fifosize;
 	port->has_sysrq = IS_ENABLED(CONFIG_SERIAL_STM32_CONSOLE);
 	port->irq = irq;
 	port->rs485_config = stm32_usart_config_rs485;
@@ -1489,14 +1590,6 @@ static int stm32_usart_init_port(struct stm32_port *stm32port,
 	stm32port->swap = stm32port->info->cfg.has_swap &&
 		of_property_read_bool(pdev->dev.of_node, "rx-tx-swap");
 
-	stm32port->fifoen = stm32port->info->cfg.has_fifo;
-	if (stm32port->fifoen) {
-		stm32_usart_get_ftcfg(pdev, "rx-threshold",
-				      &stm32port->rxftcfg);
-		stm32_usart_get_ftcfg(pdev, "tx-threshold",
-				      &stm32port->txftcfg);
-	}
-
 	port->membase = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
 	if (IS_ERR(port->membase))
 		return PTR_ERR(port->membase);
@@ -1519,6 +1612,15 @@ static int stm32_usart_init_port(struct stm32_port *stm32port,
 		goto err_clk;
 	}
 
+	stm32port->fifoen = stm32port->info->cfg.has_fifo;
+	if (stm32port->fifoen) {
+		stm32_usart_get_ftcfg(pdev, stm32port, "rx-threshold", &stm32port->rxftcfg);
+		port->fifosize = stm32_usart_get_ftcfg(pdev, stm32port, "tx-threshold",
+						       &stm32port->txftcfg);
+	} else {
+		port->fifosize = 1;
+	}
+
 	stm32port->gpios = mctrl_gpio_init(&stm32port->port, 0);
 	if (IS_ERR(stm32port->gpios)) {
 		ret = PTR_ERR(stm32port->gpios);
@@ -1601,13 +1703,6 @@ static int stm32_usart_of_dma_rx_probe(struct stm32_port *stm32port,
 	struct dma_slave_config config;
 	int ret;
 
-	/*
-	 * Using DMA and threaded handler for the console could lead to
-	 * deadlocks.
-	 */
-	if (uart_console(port))
-		return -ENODEV;
-
 	stm32port->rx_buf = dma_alloc_coherent(dev, RX_BUF_L,
 					       &stm32port->rx_dma_buf,
 					       GFP_KERNEL);
@@ -1769,29 +1864,21 @@ err_dma_rx:
 	return ret;
 }
 
-static int stm32_usart_serial_remove(struct platform_device *pdev)
+static void stm32_usart_serial_remove(struct platform_device *pdev)
 {
 	struct uart_port *port = platform_get_drvdata(pdev);
 	struct stm32_port *stm32_port = to_stm32_port(port);
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
-	int err;
 	u32 cr3;
 
 	pm_runtime_get_sync(&pdev->dev);
-	err = uart_remove_one_port(&stm32_usart_driver, port);
-	if (err)
-		return(err);
+	uart_remove_one_port(&stm32_usart_driver, port);
 
 	pm_runtime_disable(&pdev->dev);
 	pm_runtime_set_suspended(&pdev->dev);
 	pm_runtime_put_noidle(&pdev->dev);
 
 	stm32_usart_clr_bits(port, ofs->cr1, USART_CR1_PEIE);
-	cr3 = readl_relaxed(port->membase + ofs->cr3);
-	cr3 &= ~USART_CR3_EIE;
-	cr3 &= ~USART_CR3_DMAR;
-	cr3 &= ~USART_CR3_DDRE;
-	writel_relaxed(cr3, port->membase + ofs->cr3);
 
 	if (stm32_port->tx_ch) {
 		stm32_usart_of_dma_tx_remove(stm32_port, pdev);
@@ -1803,7 +1890,12 @@ static int stm32_usart_serial_remove(struct platform_device *pdev)
 		dma_release_channel(stm32_port->rx_ch);
 	}
 
-	stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAT);
+	cr3 = readl_relaxed(port->membase + ofs->cr3);
+	cr3 &= ~USART_CR3_EIE;
+	cr3 &= ~USART_CR3_DMAR;
+	cr3 &= ~USART_CR3_DMAT;
+	cr3 &= ~USART_CR3_DDRE;
+	writel_relaxed(cr3, port->membase + ofs->cr3);
 
 	if (stm32_port->wakeup_src) {
 		dev_pm_clear_wake_irq(&pdev->dev);
@@ -1811,8 +1903,6 @@ static int stm32_usart_serial_remove(struct platform_device *pdev)
 	}
 
 	stm32_usart_deinit_port(stm32_port);
-
-	return 0;
 }
 
 static void __maybe_unused stm32_usart_console_putchar(struct uart_port *port, unsigned char ch)
@@ -1845,9 +1935,9 @@ static void stm32_usart_console_write(struct console *co, const char *s,
 	int locked = 1;
 
 	if (oops_in_progress)
-		locked = spin_trylock_irqsave(&port->lock, flags);
+		locked = uart_port_trylock_irqsave(port, &flags);
 	else
-		spin_lock_irqsave(&port->lock, flags);
+		uart_port_lock_irqsave(port, &flags);
 
 	/* Save and disable interrupts, enable the transmitter */
 	old_cr1 = readl_relaxed(port->membase + ofs->cr1);
@@ -1861,7 +1951,7 @@ static void stm32_usart_console_write(struct console *co, const char *s,
 	writel_relaxed(old_cr1, port->membase + ofs->cr1);
 
 	if (locked)
-		spin_unlock_irqrestore(&port->lock, flags);
+		uart_port_unlock_irqrestore(port, flags);
 }
 
 static int stm32_usart_console_setup(struct console *co, char *options)
@@ -1975,7 +2065,7 @@ static int __maybe_unused stm32_usart_serial_en_wakeup(struct uart_port *port,
 	const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
 	struct tty_port *tport = &port->state->port;
 	int ret;
-	unsigned int size;
+	unsigned int size = 0;
 	unsigned long flags;
 
 	if (!stm32_port->wakeup_src || !tty_port_initialized(tport))
@@ -1996,12 +2086,11 @@ static int __maybe_unused stm32_usart_serial_en_wakeup(struct uart_port *port,
 		 * low-power mode.
 		 */
 		if (stm32_port->rx_ch) {
-			spin_lock_irqsave(&port->lock, flags);
-			/* Avoid race with RX IRQ when DMAR is cleared */
-			stm32_usart_clr_bits(port, ofs->cr3, USART_CR3_DMAR);
+			uart_port_lock_irqsave(port, &flags);
 			/* Poll data from DMA RX buffer if any */
-			size = stm32_usart_receive_chars(port, true);
-			dmaengine_terminate_async(stm32_port->rx_ch);
+			if (!stm32_usart_rx_dma_pause(stm32_port))
+				size += stm32_usart_receive_chars(port, true);
+			stm32_usart_rx_dma_terminate(stm32_port);
 			uart_unlock_and_check_sysrq_irqrestore(port, flags);
 			if (size)
 				tty_flip_buffer_push(tport);
@@ -2011,7 +2100,7 @@ static int __maybe_unused stm32_usart_serial_en_wakeup(struct uart_port *port,
 		stm32_usart_receive_chars(port, false);
 	} else {
 		if (stm32_port->rx_ch) {
-			ret = stm32_usart_start_rx_dma_cyclic(port);
+			ret = stm32_usart_rx_dma_start_or_resume(port);
 			if (ret)
 				return ret;
 		}