serial: sh-sci: Fix TX DMA buffer flushing and workqueue races

When uart_flush_buffer() is called, the .flush_buffer() callback zeroes
the tx_dma_len field.  This may race with the work queue function
handling transmit DMA requests:

  1. If the buffer is flushed before the first DMA API call,
     dmaengine_prep_slave_single() may be called with a zero length,
     causing the DMA request to never complete, leading to messages
     like:

        rcar-dmac e7300000.dma-controller: Channel Address Error happen

     and, with debug enabled:

	sh-sci e6e88000.serial: sci_dma_tx_work_fn: ffff800639b55000: 0...0, cookie 126

     and DMA timeouts.

  2. If the buffer is flushed after the first DMA API call, but before
     the second, dma_sync_single_for_device() may be called with a zero
     length, causing the transmit data not to be flushed to RAM, and
     leading to stale data being output.

Fix this by:
  1. Letting sci_dma_tx_work_fn() return immediately if the transmit
     buffer is empty,
  2. Extending the critical section to cover all DMA preparational work,
     so tx_dma_len stays consistent for all of it,
  3. Using local copies of circ_buf.head and circ_buf.tail, to make sure
     they match the actual operation above.

Reported-by: Eugeniu Rosca <[email protected]>
Suggested-by: Yoshihiro Shimoda <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
Reviewed-by: Eugeniu Rosca <[email protected]>
Tested-by: Eugeniu Rosca <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

drivers/tty/serial/sh-sci.c

@@ -1398,6 +1398,7 @@ static void sci_dma_tx_work_fn(struct work_struct *work)
 	struct circ_buf *xmit = &port->state->xmit;
 	unsigned long flags;
 	dma_addr_t buf;
+	int head, tail;
 
 	/*
 	 * DMA is idle now.
@@ -1407,35 +1408,42 @@ static void sci_dma_tx_work_fn(struct work_struct *work)
 	 * consistent xmit buffer state.
 	 */
 	spin_lock_irq(&port->lock);
-	buf = s->tx_dma_addr + (xmit->tail & (UART_XMIT_SIZE - 1));
+	head = xmit->head;
+	tail = xmit->tail;
+	buf = s->tx_dma_addr + (tail & (UART_XMIT_SIZE - 1));
 	s->tx_dma_len = min_t(unsigned int,
-		CIRC_CNT(xmit->head, xmit->tail, UART_XMIT_SIZE),
-		CIRC_CNT_TO_END(xmit->head, xmit->tail, UART_XMIT_SIZE));
-	spin_unlock_irq(&port->lock);
+		CIRC_CNT(head, tail, UART_XMIT_SIZE),
+		CIRC_CNT_TO_END(head, tail, UART_XMIT_SIZE));
+	if (!s->tx_dma_len) {
+		/* Transmit buffer has been flushed */
+		spin_unlock_irq(&port->lock);
+		return;
+	}
 
 	desc = dmaengine_prep_slave_single(chan, buf, s->tx_dma_len,
 					   DMA_MEM_TO_DEV,
 					   DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
 	if (!desc) {
+		spin_unlock_irq(&port->lock);
 		dev_warn(port->dev, "Failed preparing Tx DMA descriptor\n");
 		goto switch_to_pio;
 	}
 
 	dma_sync_single_for_device(chan->device->dev, buf, s->tx_dma_len,
 				   DMA_TO_DEVICE);
 
-	spin_lock_irq(&port->lock);
 	desc->callback = sci_dma_tx_complete;
 	desc->callback_param = s;
-	spin_unlock_irq(&port->lock);
 	s->cookie_tx = dmaengine_submit(desc);
 	if (dma_submit_error(s->cookie_tx)) {
+		spin_unlock_irq(&port->lock);
 		dev_warn(port->dev, "Failed submitting Tx DMA descriptor\n");
 		goto switch_to_pio;
 	}
 
+	spin_unlock_irq(&port->lock);
 	dev_dbg(port->dev, "%s: %p: %d...%d, cookie %d\n",
-		__func__, xmit->buf, xmit->tail, xmit->head, s->cookie_tx);
+		__func__, xmit->buf, tail, head, s->cookie_tx);
 
 	dma_async_issue_pending(chan);
 	return;