[PATCH] drm: work around dma_addr_t/resource_size_t mixup warning

[PATCH] drm: work around dma_addr_t/resource_size_t mixup warning

[Arnd Bergmann]

On configurations with 64-bit dma_addr_t but 32-bit resource_size_t,
there is now a warning:

drivers/gpu/drm/drm_bufs.c: In function 'drm_addmap_core':
drivers/gpu/drm/drm_bufs.c:328:8: error: passing argument 3 of 'dma_alloc_coherent' from incompatible pointer type [-Werror=incompatible-pointer-types]
  328 |        &map->offset,
      |        ^~~~~~~~~~~~
      |        |
      |        resource_size_t * {aka unsigned int *}
In file included from include/linux/pci-dma-compat.h:8,
                 from include/linux/pci.h:2392,
                 from include/drm/drm_pci.h:35,
                 from drivers/gpu/drm/drm_bufs.c:46:
include/linux/dma-mapping.h:642:15: note: expected 'dma_addr_t *' {aka 'long long unsigned int *'} but argument is of type 'resource_size_t *' {aka 'unsigned int *'}
  642 |   dma_addr_t *dma_handle, gfp_t gfp)
      |   ~~~~~~~~~~~~^~~~~~~~~~

I have no idea if this is safe on targets that may need a high DMA address,
or why we store a DMA address token in a resource_size_t in the first place,
but using a temporary variable avoids the warning.

Fixes: 8e4ff9b56957 ("drm: Remove the dma_alloc_coherent wrapper for internal usage")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 drivers/gpu/drm/drm_bufs.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/drm_bufs.c b/drivers/gpu/drm/drm_bufs.c
index dcabf5698333..0fbe65c62f1e 100644
--- a/drivers/gpu/drm/drm_bufs.c
+++ b/drivers/gpu/drm/drm_bufs.c
@@ -149,6 +149,7 @@ static int drm_addmap_core(struct drm_device *dev, resource_size_t offset,
 {
 	struct drm_local_map *map;
 	struct drm_map_list *list;
+	dma_addr_t dma_addr;
 	unsigned long user_token;
 	int ret;
 
@@ -325,8 +326,9 @@ static int drm_addmap_core(struct drm_device *dev, resource_size_t offset,
 		 * need to point to a 64bit variable first. */
 		map->handle = dma_alloc_coherent(&dev->pdev->dev,
 						 map->size,
-						 &map->offset,
+						 &dma_addr,
 						 GFP_KERNEL);
+		map->offset = (resource_size_t)dma_addr;
 		if (!map->handle) {
 			kfree(map);
 			return -ENOMEM;
-- 
2.26.0

Re: [PATCH] drm: work around dma_addr_t/resource_size_t mixup warning

[Daniel Vetter]

The fix got stuck a bit, I just pushed it out, should make it to the
next linux-next:

commit b2ecb89c27a4fd110187e0afeca70557215f55a1 (drm-misc-next-fixes)
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Apr 2 22:59:26 2020 +0100

    drm/legacy: Fix type for drm_local_map.offset

Cheers, Daniel

On Wed, Apr 8, 2020 at 9:12 PM Arnd Bergmann <arnd@arndb.de> wrote:
>
> On configurations with 64-bit dma_addr_t but 32-bit resource_size_t,
> there is now a warning:
>
> drivers/gpu/drm/drm_bufs.c: In function 'drm_addmap_core':
> drivers/gpu/drm/drm_bufs.c:328:8: error: passing argument 3 of 'dma_alloc_coherent' from incompatible pointer type [-Werror=incompatible-pointer-types]
>   328 |        &map->offset,
>       |        ^~~~~~~~~~~~
>       |        |
>       |        resource_size_t * {aka unsigned int *}
> In file included from include/linux/pci-dma-compat.h:8,
>                  from include/linux/pci.h:2392,
>                  from include/drm/drm_pci.h:35,
>                  from drivers/gpu/drm/drm_bufs.c:46:
> include/linux/dma-mapping.h:642:15: note: expected 'dma_addr_t *' {aka 'long long unsigned int *'} but argument is of type 'resource_size_t *' {aka 'unsigned int *'}
>   642 |   dma_addr_t *dma_handle, gfp_t gfp)
>       |   ~~~~~~~~~~~~^~~~~~~~~~
>
> I have no idea if this is safe on targets that may need a high DMA address,
> or why we store a DMA address token in a resource_size_t in the first place,
> but using a temporary variable avoids the warning.
>
> Fixes: 8e4ff9b56957 ("drm: Remove the dma_alloc_coherent wrapper for internal usage")
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
>  drivers/gpu/drm/drm_bufs.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/drm_bufs.c b/drivers/gpu/drm/drm_bufs.c
> index dcabf5698333..0fbe65c62f1e 100644
> --- a/drivers/gpu/drm/drm_bufs.c
> +++ b/drivers/gpu/drm/drm_bufs.c
> @@ -149,6 +149,7 @@ static int drm_addmap_core(struct drm_device *dev, resource_size_t offset,
>  {
>         struct drm_local_map *map;
>         struct drm_map_list *list;
> +       dma_addr_t dma_addr;
>         unsigned long user_token;
>         int ret;
>
> @@ -325,8 +326,9 @@ static int drm_addmap_core(struct drm_device *dev, resource_size_t offset,
>                  * need to point to a 64bit variable first. */
>                 map->handle = dma_alloc_coherent(&dev->pdev->dev,
>                                                  map->size,
> -                                                &map->offset,
> +                                                &dma_addr,
>                                                  GFP_KERNEL);
> +               map->offset = (resource_size_t)dma_addr;
>                 if (!map->handle) {
>                         kfree(map);
>                         return -ENOMEM;
> --
> 2.26.0
>


-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch