[bug report] tty: serial: fsl_lpuart: fix DMA mapping

[bug report] tty: serial: fsl_lpuart: fix DMA mapping

[Dan Carpenter]

Hello Michael Walle,

This is a semi-automatic email about new static checker warnings.

The patch a092ab25fdaa: "tty: serial: fsl_lpuart: fix DMA mapping" 
from Mar 6, 2020, leads to the following Smatch complaint:

    drivers/tty/serial/fsl_lpuart.c:1237 lpuart_dma_rx_free()
    error: we previously assumed 'chan' could be null (see line 1234)

drivers/tty/serial/fsl_lpuart.c
  1228  static void lpuart_dma_rx_free(struct uart_port *port)
  1229  {
  1230          struct lpuart_port *sport = container_of(port,
  1231                                          struct lpuart_port, port);
  1232          struct dma_chan *chan = sport->dma_rx_chan;
  1233	
  1234		if (chan)
                    ^^^^
The patch adds a check

  1235			dmaengine_terminate_all(chan);
  1236	
  1237		dma_unmap_sg(chan->device->dev, &sport->rx_sgl, 1, DMA_FROM_DEVICE);
                             ^^^^^^
and an unchecked dereference.

  1238		kfree(sport->rx_ring.buf);
  1239		sport->rx_ring.tail = 0;

regards,
dan carpenter

Re: [bug report] tty: serial: fsl_lpuart: fix DMA mapping

[Michael Walle]

Am 2020-04-06 16:42, schrieb Dan Carpenter:
> Hello Michael Walle,
> 
> This is a semi-automatic email about new static checker warnings.

Thanks, fix is pending, see here:
   
https://lore.kernel.org/linux-serial/20200403174942.9594-1-michael@walle.cc/

-michael

> 
> The patch a092ab25fdaa: "tty: serial: fsl_lpuart: fix DMA mapping"
> from Mar 6, 2020, leads to the following Smatch complaint:
> 
>     drivers/tty/serial/fsl_lpuart.c:1237 lpuart_dma_rx_free()
>     error: we previously assumed 'chan' could be null (see line 1234)
> 
> drivers/tty/serial/fsl_lpuart.c
>   1228  static void lpuart_dma_rx_free(struct uart_port *port)
>   1229  {
>   1230          struct lpuart_port *sport = container_of(port,
>   1231                                          struct lpuart_port, 
> port);
>   1232          struct dma_chan *chan = sport->dma_rx_chan;
>   1233
>   1234		if (chan)
>                     ^^^^
> The patch adds a check
> 
>   1235			dmaengine_terminate_all(chan);
>   1236
>   1237		dma_unmap_sg(chan->device->dev, &sport->rx_sgl, 1, 
> DMA_FROM_DEVICE);
>                              ^^^^^^
> and an unchecked dereference.
> 
>   1238		kfree(sport->rx_ring.buf);
>   1239		sport->rx_ring.tail = 0;
> 
> regards,
> dan carpenter