[PATCH] ARM: samsung: improve static dma_mask definition

[PATCH] ARM: samsung: improve static dma_mask definition

[Arnd Bergmann]

When no DMA master devices are part of the kernel configuration,
we get a warning about the unused dma mask definition:

arch/arm/plat-samsung/devs.c:71:12: error: 'samsung_device_dma_mask' defined but not used [-Werror=unused-variable]
 static u64 samsung_device_dma_mask = DMA_BIT_MASK(32);

We could simply mark this as __maybe_unused to shut up that warning,
but a nicer solution seems to be to have a separate mask for each
device. The advantage is that a driver that happens to call
dma_set_mask() on one device doesn't implicitly change the mask
for the other devices as well. This is more of a theoretical
problem, as obviously nothing does it for the devices in this
file (or they would have always been broken), but it feels
cleaner that way.

The definition works by creating an array in place so we can take
the address of it and let the compiler generate a hidden symbol
for it at compile time.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 arch/arm/plat-samsung/devs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/arm/plat-samsung/devs.c b/arch/arm/plat-samsung/devs.c
index 84baa16f4c0b..e93aa6734147 100644
--- a/arch/arm/plat-samsung/devs.c
+++ b/arch/arm/plat-samsung/devs.c
@@ -68,7 +68,7 @@
 #include <linux/platform_data/asoc-s3c.h>
 #include <linux/platform_data/spi-s3c64xx.h>
 
-static u64 samsung_device_dma_mask = DMA_BIT_MASK(32);
+#define samsung_device_dma_mask (*((u64[]) { DMA_BIT_MASK(32) }))
 
 /* AC97 */
 #ifdef CONFIG_CPU_S3C2440
-- 
2.7.0

[PATCH] ARM: samsung: improve static dma_mask definition

[Krzysztof Kozlowski]

On 05/12/2016 10:38 AM, Arnd Bergmann wrote:
> When no DMA master devices are part of the kernel configuration,
> we get a warning about the unused dma mask definition:
> 
> arch/arm/plat-samsung/devs.c:71:12: error: 'samsung_device_dma_mask' defined but not used [-Werror=unused-variable]
>  static u64 samsung_device_dma_mask = DMA_BIT_MASK(32);
> 
> We could simply mark this as __maybe_unused to shut up that warning,
> but a nicer solution seems to be to have a separate mask for each
> device. The advantage is that a driver that happens to call
> dma_set_mask() on one device doesn't implicitly change the mask
> for the other devices as well. This is more of a theoretical
> problem, as obviously nothing does it for the devices in this
> file (or they would have always been broken), but it feels
> cleaner that way.
> 
> The definition works by creating an array in place so we can take
> the address of it and let the compiler generate a hidden symbol
> for it at compile time.
> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
>  arch/arm/plat-samsung/devs.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

Thanks, although does not look as critical fix but less warnings makes
spotting right issues faster so applied for v4.7.

Best regards,
Krzysztof