Re: [PATCH RFT RESEND linux-next] c6x: dma-mapping: supportdebug_dma_mapping_error

[Shuah Khan]

On Fri, 2012-11-02 at 16:15 -0400, Mark Salter wrote:
> On Fri, 2012-11-02 at 13:53 -0600, Shuah Khan wrote:
> > On Fri, 2012-11-02 at 15:10 -0400, Mark Salter wrote:
> > > On Fri, 2012-11-02 at 10:44 -0600, Shuah Khan wrote:
> > > > On Fri, 2012-10-26 at 09:40 -0600, Shuah Khan wrote:
> > > > > Add support for debug_dma_mapping_error() call to avoid warning from
> > > > > debug_dma_unmap() interface when it checks for mapping error checked
> > > > > status. Without this patch, device driver failed to check map error
> > > > > warning is generated.
> > > > > 
> > > > > Signed-off-by: Shuah Khan <shuah.khan@xxxxxx>
> > > > > ---
> > > > >  arch/c6x/include/asm/dma-mapping.h |    1 +
> > > > >  1 file changed, 1 insertion(+)
> > > 
> > > > Would you like to this patch go through c6x arch tree or linux-next?
> > > > Please let me know your preference.
> > > 
> > > I tried to test this but I get a build error with CONFIG_DMA_API_DEBUG:
> > > 
> > > /linux-next/lib/dma-debug.c: In function 'has_mapping_error':
> > > /linux-next/lib/dma-debug.c:863:15: error: implicit declaration of function 'get_dma_ops' [-Werror=implicit-function-declaration]
> > > /linux-next/lib/dma-debug.c:863:34: warning: initialization makes pointer from integer without a cast [enabled by default]
> > > 
> > > C6X (along with some other architectures) doesn't have a get_dma_ops()
> > > function defined.
> > 
> > That is a problem I didn't think about. I did a check and looks like c6x
> > and frv are the only ones that don't have get_dma_ops() defined. frv is
> 
> By my count, there are 14 architectures with get_dma_ops() and 14
> without.
Right. I should have explained more. The following archs

arch/avr32/include/asm/dma-mapping.h
arch/blackfin/include/asm/dma-mapping.h
arch/cris/include/asm/dma-mapping.h
arch/mn10300/include/asm/dma-mapping.h
arch/parisc/include/asm/dma-mapping.h
arch/xtensa/include/asm/dma-mapping.h

define dma_map_page() and dma_map_single() and not call
debug_dma_map_page() interface. There is no risk of mis-matched debug
and non-debug mapping and mapping error checks like in the case of other
archs and c6x.

> > in a different category as it doesn't use dma_debug interfaces. IN the
> > case c6x, now with my change to add debug_dma_mapping_error(), we will
> > start seeing warnings since dma_map_page() and dma_map_single() are
> > debugged with a call to debug_dma_map_page() and the corresponding
> > dma_mapping_error() interface doesn't call debug_dma_mapping_error()
> > interface
> > 
> > - Does adding get_dma_ops() make sense? Doesn't look like c6x exports
> > dma_ops?
> > 
> > Any other ideas?
> 
> I'm not sure. I don't know what get_dma_ops() does and it doesn't seem
> to be documented anywhere.

It returns pointer to dma_ops like the one on alpha:

static inline struct dma_map_ops *get_dma_ops(struct device *dev)
{
        return dma_ops;
}

c6x doesn't define dma_ops looks like. Is that correct? Returning null
from get_dma_ops() is not an option as get_dma_ops() return is assumed
to be not null.

Thanks,
-- Shuah


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/