[PATCH v2] staging: fbtft: Disable DMA support if DMA is not available

[PATCH v2] staging: fbtft: Disable DMA support if DMA is not available

[Geert Uytterhoeven]

If NO_DMA=y:

    drivers/built-in.o: In function `fbtft_framebuffer_alloc':
    (.text+0xb53cae): undefined reference to `dmam_alloc_coherent'

As DMA support is already optional, make it depend on HAS_DMA.

If !HAS_DMA, "dma" will always be false, and the compiler will optimize
away the call to dmam_alloc_coherent().

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
---
v2:
  - Explain better how it works,
  - Keep the "= false" to avoid breakage (seen not only with gcc 4.1.2,
    but also with 4.6.3 and 4.9.0).
---
 drivers/staging/fbtft/fbtft-core.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/staging/fbtft/fbtft-core.c b/drivers/staging/fbtft/fbtft-core.c
index 53b748be27124793..600dbdcaa716cd76 100644
--- a/drivers/staging/fbtft/fbtft-core.c
+++ b/drivers/staging/fbtft/fbtft-core.c
@@ -47,9 +47,14 @@ static unsigned long debug;
 module_param(debug, ulong, 0);
 MODULE_PARM_DESC(debug, "override device debug level");
 
+#ifdef CONFIG_HAS_DMA
 static bool dma = true;
 module_param(dma, bool, 0);
 MODULE_PARM_DESC(dma, "Use DMA buffer");
+#else
+/* Do not remove the "= false", gcc needs it to optimize away the DMA calls */
+static const bool dma = false;
+#endif
 
 
 void fbtft_dbg_hex(const struct device *dev, int groupsize,
-- 
1.9.1

Re: [PATCH v2] staging: fbtft: Disable DMA support if DMA is not available

[Greg Kroah-Hartman]

On Thu, Apr 23, 2015 at 08:47:15PM +0200, Geert Uytterhoeven wrote:
> If NO_DMA=y:
> 
>     drivers/built-in.o: In function `fbtft_framebuffer_alloc':
>     (.text+0xb53cae): undefined reference to `dmam_alloc_coherent'
> 
> As DMA support is already optional, make it depend on HAS_DMA.
> 
> If !HAS_DMA, "dma" will always be false, and the compiler will optimize
> away the call to dmam_alloc_coherent().

Oh that's "sneaky" and not good, let's just make this driver be
dependant on DMA and be done with it.  I don't like to rely on the
optimizer to get things to work properly.

thanks,

greg k-h

Re: [PATCH v2] staging: fbtft: Disable DMA support if DMA is not available

[Geert Uytterhoeven]

On Thu, Apr 23, 2015 at 9:17 PM, Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
> On Thu, Apr 23, 2015 at 08:47:15PM +0200, Geert Uytterhoeven wrote:
>> If NO_DMA=y:
>>
>>     drivers/built-in.o: In function `fbtft_framebuffer_alloc':
>>     (.text+0xb53cae): undefined reference to `dmam_alloc_coherent'
>>
>> As DMA support is already optional, make it depend on HAS_DMA.
>>
>> If !HAS_DMA, "dma" will always be false, and the compiler will optimize
>> away the call to dmam_alloc_coherent().
>
> Oh that's "sneaky" and not good, let's just make this driver be
> dependant on DMA and be done with it.  I don't like to rely on the
> optimizer to get things to work properly.

Alternatively, I can stick in one more #ifdef, to no longer rely on the compiler
optimization?
AFAICS, the driver itself considers DMA optional, and will work without.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH v2] staging: fbtft: Disable DMA support if DMA is not available

[Greg Kroah-Hartman]

On Fri, Apr 24, 2015 at 09:15:00AM +0200, Geert Uytterhoeven wrote:
> On Thu, Apr 23, 2015 at 9:17 PM, Greg Kroah-Hartman
> <gregkh@linuxfoundation.org> wrote:
> > On Thu, Apr 23, 2015 at 08:47:15PM +0200, Geert Uytterhoeven wrote:
> >> If NO_DMA=y:
> >>
> >>     drivers/built-in.o: In function `fbtft_framebuffer_alloc':
> >>     (.text+0xb53cae): undefined reference to `dmam_alloc_coherent'
> >>
> >> As DMA support is already optional, make it depend on HAS_DMA.
> >>
> >> If !HAS_DMA, "dma" will always be false, and the compiler will optimize
> >> away the call to dmam_alloc_coherent().
> >
> > Oh that's "sneaky" and not good, let's just make this driver be
> > dependant on DMA and be done with it.  I don't like to rely on the
> > optimizer to get things to work properly.
> 
> Alternatively, I can stick in one more #ifdef, to no longer rely on the compiler
> optimization?

That would be better.

thanks,

greg k-h