On Tue, Jun 30, 2020 at 01:22:19PM +0530, Souptick Joarder wrote:
> As 3 goto level referring to same common code, those can be
> accomodated with a single goto level and renameing it to
> unpin_pages. Set the -ERRNO when returning partial mapped
> pages in more appropriate place.
> 
> Signed-off-by: Souptick Joarder <jrdr.li...@gmail.com>
> Cc: John Hubbard <jhubb...@nvidia.com>
> Cc: Bharath Vedartham <linux.b...@gmail.com>
> Cc: Dan Carpenter <dan.carpen...@oracle.com>
> ---
>  drivers/staging/kpc2000/kpc_dma/fileops.c | 24 ++++++++++++------------
>  1 file changed, 12 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/staging/kpc2000/kpc_dma/fileops.c 
> b/drivers/staging/kpc2000/kpc_dma/fileops.c
> index 8cd20ad..d21a4fd 100644
> --- a/drivers/staging/kpc2000/kpc_dma/fileops.c
> +++ b/drivers/staging/kpc2000/kpc_dma/fileops.c
> @@ -35,7 +35,7 @@ static int kpc_dma_transfer(struct dev_private_data *priv,
>                           unsigned long iov_base, size_t iov_len)
>  {
>       unsigned int i = 0;
> -     int rv = 0;
> +     int rv = 0, nr_pages = 0;
>       struct kpc_dma_device *ldev;
>       struct aio_cb_data *acd;
>       DECLARE_COMPLETION_ONSTACK(done);
> @@ -79,22 +79,27 @@ static int kpc_dma_transfer(struct dev_private_data *priv,
>       rv = pin_user_pages(iov_base, acd->page_count, FOLL_TOUCH | FOLL_WRITE, 
> acd->user_pages, NULL);
>       mmap_read_unlock(current->mm);        /*  release the semaphore */
>       if (rv != acd->page_count) {
> +             nr_pages = rv;
> +             if (rv > 0)
> +                     rv = -EFAULT;
> +
>               dev_err(&priv->ldev->pldev->dev, "Couldn't pin_user_pages 
> (%d)\n", rv);
> -             goto err_get_user_pages;
> +             goto unpin_pages;
>       }
> +     nr_pages = acd->page_count;
>  
>       // Allocate and setup the sg_table (scatterlist entries)
>       rv = sg_alloc_table_from_pages(&acd->sgt, acd->user_pages, 
> acd->page_count, iov_base & (PAGE_SIZE - 1), iov_len, GFP_KERNEL);
>       if (rv) {
>               dev_err(&priv->ldev->pldev->dev, "Couldn't alloc sg_table 
> (%d)\n", rv);
> -             goto err_alloc_sg_table;
> +             goto unpin_pages;
>       }
>  
>       // Setup the DMA mapping for all the sg entries
>       acd->mapped_entry_count = dma_map_sg(&ldev->pldev->dev, acd->sgt.sgl, 
> acd->sgt.nents, ldev->dir);
>       if (acd->mapped_entry_count <= 0) {
>               dev_err(&priv->ldev->pldev->dev, "Couldn't dma_map_sg (%d)\n", 
> acd->mapped_entry_count);
> -             goto err_dma_map_sg;
> +             goto unpin_pages;

This wasn't introduced by your patch but the most recent successful
allocation is sg_alloc_table_from_pages() so this should be goto
free_table:

free_table:
        sg_free_table(&acd->sgt);
unpin_pages:
        if (nr_pages > 0)
                unpin_user_pages(acd->user_pages, nr_pages);

Always just keep track of the most recent allocation and check that the
"goto free_foo" matches what you would expect to release the most
recent thing.

>       }
>  
>       // Calculate how many descriptors are actually needed for this transfer.
> @@ -187,15 +192,10 @@ static int kpc_dma_transfer(struct dev_private_data 
> *priv,
>       unlock_engine(ldev);
>       dma_unmap_sg(&ldev->pldev->dev, acd->sgt.sgl, acd->sgt.nents, 
> ldev->dir);
>       sg_free_table(&acd->sgt);
        ^^^^^^^^^^^^^^^^^^^^^^^^

> - err_dma_map_sg:
> - err_alloc_sg_table:
> -     unpin_user_pages(acd->user_pages, acd->page_count);
>  
> - err_get_user_pages:
> -     if (rv > 0) {
> -             unpin_user_pages(acd->user_pages, rv);
> -             rv = -EFAULT;
> -     }
> + unpin_pages:
> +     if (nr_pages > 0)
> +             unpin_user_pages(acd->user_pages, nr_pages);
>       kfree(acd->user_pages);
>   err_alloc_userpages:
>       kfree(acd);

regards,
dan carpenter

Reply via email to