diff mbox series

[4/9] arm64: dma-mapping: export symbol arch_setup_dma_ops

Message ID 1507761269-7017-5-git-send-email-jim2101024@gmail.com
State Changes Requested
Headers show
Series [1/9] SOC: brcmstb: add memory API | expand

Commit Message

Jim Quinlan Oct. 11, 2017, 10:34 p.m. UTC
The BrcmSTB driver needs to get ahold of a pointer to swiotlb_dma_ops.
However, that variable is defined as static.  Instead, we use
arch_setup_dma_ops() to get the pointer to swiotlb_dma_ops.  Since
we also want our driver to be a separate module, we need to
export this function.

Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
---
 arch/arm64/mm/dma-mapping.c | 1 +
 1 file changed, 1 insertion(+)

Comments

Robin Murphy Oct. 12, 2017, 5:06 p.m. UTC | #1
On 11/10/17 23:34, Jim Quinlan wrote:
> The BrcmSTB driver needs to get ahold of a pointer to swiotlb_dma_ops.
> However, that variable is defined as static.  Instead, we use
> arch_setup_dma_ops() to get the pointer to swiotlb_dma_ops.  Since
> we also want our driver to be a separate module, we need to
> export this function.

NAK. Retrieve the platform-assigned ops from the device via
get_dma_ops() and stash them in your drvdata or wherever before you
replace them. Don't go poking around arch code internals directly from a
driver.

Robin.

> Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
> ---
>  arch/arm64/mm/dma-mapping.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/arm64/mm/dma-mapping.c b/arch/arm64/mm/dma-mapping.c
> index 614af88..dae572f 100644
> --- a/arch/arm64/mm/dma-mapping.c
> +++ b/arch/arm64/mm/dma-mapping.c
> @@ -936,3 +936,4 @@ void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
>  	}
>  #endif
>  }
> +EXPORT_SYMBOL(arch_setup_dma_ops);
>
Jim Quinlan Oct. 12, 2017, 6:15 p.m. UTC | #2
On Thu, Oct 12, 2017 at 1:06 PM, Robin Murphy <robin.murphy@arm.com> wrote:
> On 11/10/17 23:34, Jim Quinlan wrote:
>> The BrcmSTB driver needs to get ahold of a pointer to swiotlb_dma_ops.
>> However, that variable is defined as static.  Instead, we use
>> arch_setup_dma_ops() to get the pointer to swiotlb_dma_ops.  Since
>> we also want our driver to be a separate module, we need to
>> export this function.
>
> NAK. Retrieve the platform-assigned ops from the device via
> get_dma_ops() and stash them in your drvdata or wherever before you
> replace them. Don't go poking around arch code internals directly from a
> driver.
Will fix (and drop the commit).

>
> Robin.
>
>> Signed-off-by: Jim Quinlan <jim2101024@gmail.com>
>> ---
>>  arch/arm64/mm/dma-mapping.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/arch/arm64/mm/dma-mapping.c b/arch/arm64/mm/dma-mapping.c
>> index 614af88..dae572f 100644
>> --- a/arch/arm64/mm/dma-mapping.c
>> +++ b/arch/arm64/mm/dma-mapping.c
>> @@ -936,3 +936,4 @@ void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
>>       }
>>  #endif
>>  }
>> +EXPORT_SYMBOL(arch_setup_dma_ops);
>>
>
diff mbox series

Patch

diff --git a/arch/arm64/mm/dma-mapping.c b/arch/arm64/mm/dma-mapping.c
index 614af88..dae572f 100644
--- a/arch/arm64/mm/dma-mapping.c
+++ b/arch/arm64/mm/dma-mapping.c
@@ -936,3 +936,4 @@  void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
 	}
 #endif
 }
+EXPORT_SYMBOL(arch_setup_dma_ops);