On Sat, 21 Nov 2009, Dan Williams wrote:
> On Sat, Nov 21, 2009 at 4:52 AM, Julia Lawall <julia@xxxxxxx> wrote:
> > From: Julia Lawall <julia@xxxxxxx>
> >
> > Error handling code following a kzalloc should free the allocated data.
> >
>
> Yes, but there is another buglet here:
>
> > /* get platform data */
> > if (!pdev->dev.platform_data)
> > - goto shdev_err;
> > + goto platform_err;
>
> "err" is still zero at this point so the probe does not fail.
>
> How about the following instead?
OK, that is indeed both better and simpler.
thanks,
julia
> diff --git a/drivers/dma/shdma.c b/drivers/dma/shdma.c
> index b3b065c..034ecf0 100644
> --- a/drivers/dma/shdma.c
> +++ b/drivers/dma/shdma.c
> @@ -640,17 +640,16 @@ static int __init sh_dmae_probe(struct
> platform_device *pdev)
> #endif
> struct sh_dmae_device *shdev;
>
> + /* get platform data */
> + if (!pdev->dev.platform_data)
> + return -ENODEV;
> +
> shdev = kzalloc(sizeof(struct sh_dmae_device), GFP_KERNEL);
> if (!shdev) {
> dev_err(&pdev->dev, "No enough memory\n");
> - err = -ENOMEM;
> - goto shdev_err;
> + return -ENOMEM;
> }
>
> - /* get platform data */
> - if (!pdev->dev.platform_data)
> - goto shdev_err;
> -
> /* platform data */
> memcpy(&shdev->pdata, pdev->dev.platform_data,
> sizeof(struct sh_dmae_pdata));
> @@ -722,7 +721,6 @@ eirq_err:
> rst_err:
> kfree(shdev);
>
> -shdev_err:
> return err;
> }
>
