re: spi/tegra114: add spi driver

re: spi/tegra114: add spi driver

[Dan Carpenter]

Hello Laxman Dewangan,

The patch f333a331adfa: "spi/tegra114: add spi driver" from Feb 22,
2013, leads to the following static checker warning:

	drivers/spi/spi-tegra114.c:621 tegra_spi_init_dma_param()
	error: uninitialized variable 'dma_phys'.

drivers/spi/spi-tegra114.c
   583          dma_addr_t dma_phys;
                ^^^^^^^^^^^^^^^^^^^
   584          int ret;
   585          struct dma_slave_config dma_sconfig;
   586  
   587          dma_chan = dma_request_slave_channel_reason(tspi->dev,
   588                                          dma_to_memory ? "rx" : "tx");
   589          if (IS_ERR(dma_chan)) {
   590                  ret = PTR_ERR(dma_chan);
   591                  if (ret != -EPROBE_DEFER)
   592                          dev_err(tspi->dev,
   593                                  "Dma channel is not available: %d\n", ret);
   594                  return ret;
   595          }
   596  
   597          dma_buf = dma_alloc_coherent(tspi->dev, tspi->dma_buf_size,
   598                                  &dma_phys, GFP_KERNEL);

The issue is that in dma_alloc_attrs() if we can dma_alloc_from_coherent()
then dma_phys is not set.  I'm getting a couple similar errors and I'm
not certian what to do about it.

   599          if (!dma_buf) {
   600                  dev_err(tspi->dev, " Not able to allocate the dma buffer\n");
   601                  dma_release_channel(dma_chan);
   602                  return -ENOMEM;
   603          }
   604  
   605          if (dma_to_memory) {
   606                  dma_sconfig.src_addr = tspi->phys + SPI_RX_FIFO;
   607                  dma_sconfig.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
   608                  dma_sconfig.src_maxburst = 0;
   609          } else {
   610                  dma_sconfig.dst_addr = tspi->phys + SPI_TX_FIFO;
   611                  dma_sconfig.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
   612                  dma_sconfig.dst_maxburst = 0;
   613          }
   614  
   615          ret = dmaengine_slave_config(dma_chan, &dma_sconfig);
   616          if (ret)
   617                  goto scrub;
   618          if (dma_to_memory) {
   619                  tspi->rx_dma_chan = dma_chan;
   620                  tspi->rx_dma_buf = dma_buf;
   621                  tspi->rx_dma_phys = dma_phys;
                                            ^^^^^^^^
Uninitialized.

   622          } else {
   623                  tspi->tx_dma_chan = dma_chan;
   624                  tspi->tx_dma_buf = dma_buf;
   625                  tspi->tx_dma_phys = dma_phys;

regards,
dan carpenter
--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: spi/tegra114: add spi driver

[Laxman Dewangan]

On Friday 01 April 2016 07:22 PM, Dan Carpenter wrote:
> Hello Laxman Dewangan,
>
> The patch f333a331adfa: "spi/tegra114: add spi driver" from Feb 22,
> 2013, leads to the following static checker warning:
>
> 	drivers/spi/spi-tegra114.c:621 tegra_spi_init_dma_param()
> 	error: uninitialized variable 'dma_phys'.
>
> drivers/spi/spi-tegra114.c
>     583          dma_addr_t dma_phys;
>                  ^^^^^^^^^^^^^^^^^^^
>     584          int ret;
>     585          struct dma_slave_config dma_sconfig;
>     586
>     587          dma_chan = dma_request_slave_channel_reason(tspi->dev,
>     588                                          dma_to_memory ? "rx" : "tx");
>     589          if (IS_ERR(dma_chan)) {
>     590                  ret = PTR_ERR(dma_chan);
>     591                  if (ret != -EPROBE_DEFER)
>     592                          dev_err(tspi->dev,
>     593                                  "Dma channel is not available: %d\n", ret);
>     594                  return ret;
>     595          }
>     596
>     597          dma_buf = dma_alloc_coherent(tspi->dev, tspi->dma_buf_size,
>     598                                  &dma_phys, GFP_KERNEL);
>
> The issue is that in dma_alloc_attrs() if we can dma_alloc_from_coherent()
> then dma_phys is not set.  I'm getting a couple similar errors and I'm
> not certian what to do about it.
>


The issue is with compiler actually.
we have code as
         dma_buf = dma_alloc_coherent(tspi->dev, tspi->dma_buf_size,
                                 &dma_phys, GFP_KERNEL);
         if (!dma_buf) {
                 dev_err(tspi->dev, " Not able to allocate the dma 
buffer\n");
                 dma_release_channel(dma_chan);
                 return -ENOMEM;
         }

So if returned value "dma_buf" is not NULL then we can assume that 
dma_phy is initialized other wise it should return NULL.

The static check is not able to find out that calling function has 
initialized it or not.

--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: spi/tegra114: add spi driver

[Dan Carpenter]

Oh...  Duh.  I read this code but I got my variables switched around in
my head.  The static checker finds the code confusing in a different way
than I did...

Thanks for taking a look.

regards,
dan carpenter