[PATCH 0/3] Make DMA slave ID 0 invalid to simplify platform data

[PATCH 0/3] Make DMA slave ID 0 invalid to simplify platform data

[Guennadi Liakhovetski]

Currently platforms, using the SDHI TMIO driver, have to specify negative 
DMA slave IDs to force PIO mode. If default 0 is left, the driver 
recognises it as a valid DMA slave ID and tries to acquire channels. This 
fails then, because the driver cannot allocate two channels with the same 
ID, but it would be better to just make 0 an invalid ID to prevent this.

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/

[PATCH 1/3] sh: mark DMA slave ID 0 as invalid

[Guennadi Liakhovetski]

This makes it possible to leave DMA slave IDs in the platform data
at default 0 value without hitting DMA channel allocation error paths.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
---
 arch/sh/include/cpu-sh4/cpu/sh7722.h |    1 +
 arch/sh/include/cpu-sh4/cpu/sh7724.h |    1 +
 arch/sh/include/cpu-sh4/cpu/sh7757.h |    1 +
 3 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/arch/sh/include/cpu-sh4/cpu/sh7722.h b/arch/sh/include/cpu-sh4/cpu/sh7722.h
index 7a5b8a3..bd06227 100644
--- a/arch/sh/include/cpu-sh4/cpu/sh7722.h
+++ b/arch/sh/include/cpu-sh4/cpu/sh7722.h
@@ -236,6 +236,7 @@ enum {
 };
 
 enum {
+	SHDMA_SLAVE_INVALID,
 	SHDMA_SLAVE_SCIF0_TX,
 	SHDMA_SLAVE_SCIF0_RX,
 	SHDMA_SLAVE_SCIF1_TX,
diff --git a/arch/sh/include/cpu-sh4/cpu/sh7724.h b/arch/sh/include/cpu-sh4/cpu/sh7724.h
index 7eb4359..3daef8e 100644
--- a/arch/sh/include/cpu-sh4/cpu/sh7724.h
+++ b/arch/sh/include/cpu-sh4/cpu/sh7724.h
@@ -285,6 +285,7 @@ enum {
 };
 
 enum {
+	SHDMA_SLAVE_INVALID,
 	SHDMA_SLAVE_SCIF0_TX,
 	SHDMA_SLAVE_SCIF0_RX,
 	SHDMA_SLAVE_SCIF1_TX,
diff --git a/arch/sh/include/cpu-sh4/cpu/sh7757.h b/arch/sh/include/cpu-sh4/cpu/sh7757.h
index 05b8196..41f9f8b 100644
--- a/arch/sh/include/cpu-sh4/cpu/sh7757.h
+++ b/arch/sh/include/cpu-sh4/cpu/sh7757.h
@@ -252,6 +252,7 @@ enum {
 };
 
 enum {
+	SHDMA_SLAVE_INVALID,
 	SHDMA_SLAVE_SDHI_TX,
 	SHDMA_SLAVE_SDHI_RX,
 	SHDMA_SLAVE_MMCIF_TX,
-- 
1.7.2.5

[PATCH 2/3] ARM: mach-shmobile: mark DMA slave ID 0 as invalid

[Guennadi Liakhovetski]

This makes it possible to leave DMA slave IDs in the platform data
at default 0 value without hitting DMA channel allocation error paths.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
---
 arch/arm/mach-shmobile/include/mach/sh7372.h |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/arch/arm/mach-shmobile/include/mach/sh7372.h b/arch/arm/mach-shmobile/include/mach/sh7372.h
index 6a290b7..75076b1 100644
--- a/arch/arm/mach-shmobile/include/mach/sh7372.h
+++ b/arch/arm/mach-shmobile/include/mach/sh7372.h
@@ -436,6 +436,7 @@ enum {
 
 /* DMA slave IDs */
 enum {
+	SHDMA_SLAVE_INVALID,
 	SHDMA_SLAVE_SCIF0_TX,
 	SHDMA_SLAVE_SCIF0_RX,
 	SHDMA_SLAVE_SCIF1_TX,
-- 
1.7.2.5

[PATCH 3/3] mmc: SDHI: DMA slave ID 0 is invalid

[Guennadi Liakhovetski]

Don't try to allocate DMA resources, if the platform didn't specify
positive DMA slave IDs.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
---
 drivers/mmc/host/sh_mobile_sdhi.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/mmc/host/sh_mobile_sdhi.c b/drivers/mmc/host/sh_mobile_sdhi.c
index c05d699..e646741 100644
--- a/drivers/mmc/host/sh_mobile_sdhi.c
+++ b/drivers/mmc/host/sh_mobile_sdhi.c
@@ -132,7 +132,7 @@ static int __devinit sh_mobile_sdhi_probe(struct platform_device *pdev)
 		mmc_data->ocr_mask = p->tmio_ocr_mask;
 		mmc_data->capabilities |= p->tmio_caps;
 
-		if (p->dma_slave_tx >= 0 && p->dma_slave_rx >= 0) {
+		if (p->dma_slave_tx > 0 && p->dma_slave_rx > 0) {
 			priv->param_tx.slave_id = p->dma_slave_tx;
 			priv->param_rx.slave_id = p->dma_slave_rx;
 			priv->dma_priv.chan_priv_tx = &priv->param_tx;
-- 
1.7.2.5

Re: [PATCH 3/3] mmc: SDHI: DMA slave ID 0 is invalid

[Guennadi Liakhovetski]

Hi Chris

Can we please get this patch too in for 3.0? It is a part of a 
patch-series, of which all others havs already been applied. This specific 
patch makes the behaviour of the SDHI driver consistent with other DMA 
users on sh-mobile, namely, that default platform data with no DMA 
information provided immediately switches the driver to the PIO mode. This 
patch fixes this logical error of trying to acquire DMA channels with 
invalid configuration information and avoids ugly error messages in the 
kernel log.

Thanks
Guennadi

On Tue, 24 May 2011, Guennadi Liakhovetski wrote:

> Don't try to allocate DMA resources, if the platform didn't specify
> positive DMA slave IDs.
> 
> Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
> ---
>  drivers/mmc/host/sh_mobile_sdhi.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/mmc/host/sh_mobile_sdhi.c b/drivers/mmc/host/sh_mobile_sdhi.c
> index c05d699..e646741 100644
> --- a/drivers/mmc/host/sh_mobile_sdhi.c
> +++ b/drivers/mmc/host/sh_mobile_sdhi.c
> @@ -132,7 +132,7 @@ static int __devinit sh_mobile_sdhi_probe(struct platform_device *pdev)
>  		mmc_data->ocr_mask = p->tmio_ocr_mask;
>  		mmc_data->capabilities |= p->tmio_caps;
>  
> -		if (p->dma_slave_tx >= 0 && p->dma_slave_rx >= 0) {
> +		if (p->dma_slave_tx > 0 && p->dma_slave_rx > 0) {
>  			priv->param_tx.slave_id = p->dma_slave_tx;
>  			priv->param_rx.slave_id = p->dma_slave_rx;
>  			priv->dma_priv.chan_priv_tx = &priv->param_tx;
> -- 
> 1.7.2.5
> 
> 

---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/

Re: [PATCH 3/3] mmc: SDHI: DMA slave ID 0 is invalid

[Chris Ball]

Hi Guennadi,

On Mon, Jun 20 2011, Guennadi Liakhovetski wrote:
> Can we please get this patch too in for 3.0? It is a part of a 
> patch-series, of which all others havs already been applied. This specific 
> patch makes the behaviour of the SDHI driver consistent with other DMA 
> users on sh-mobile, namely, that default platform data with no DMA 
> information provided immediately switches the driver to the PIO mode. This 
> patch fixes this logical error of trying to acquire DMA channels with 
> invalid configuration information and avoids ugly error messages in the 
> kernel log.
>
> Thanks
> Guennadi
>
> On Tue, 24 May 2011, Guennadi Liakhovetski wrote:
>
>> Don't try to allocate DMA resources, if the platform didn't specify
>> positive DMA slave IDs.
>> 
>> Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
>> ---
>>  drivers/mmc/host/sh_mobile_sdhi.c |    2 +-
>>  1 files changed, 1 insertions(+), 1 deletions(-)
>> 
>> diff --git a/drivers/mmc/host/sh_mobile_sdhi.c b/drivers/mmc/host/sh_mobile_sdhi.c
>> index c05d699..e646741 100644
>> --- a/drivers/mmc/host/sh_mobile_sdhi.c
>> +++ b/drivers/mmc/host/sh_mobile_sdhi.c
>> @@ -132,7 +132,7 @@ static int __devinit sh_mobile_sdhi_probe(struct platform_device *pdev)
>>  		mmc_data->ocr_mask = p->tmio_ocr_mask;
>>  		mmc_data->capabilities |= p->tmio_caps;
>>  
>> -		if (p->dma_slave_tx >= 0 && p->dma_slave_rx >= 0) {
>> +		if (p->dma_slave_tx > 0 && p->dma_slave_rx > 0) {
>>  			priv->param_tx.slave_id = p->dma_slave_tx;
>>  			priv->param_rx.slave_id = p->dma_slave_rx;
>>  			priv->dma_priv.chan_priv_tx = &priv->param_tx;
>> -- 
>> 1.7.2.5

Thanks, pushed to mmc-next for 3.0.

- Chris.
-- 
Chris Ball   <cjb@laptop.org>   <http://printf.net/>
One Laptop Per Child

Re: [PATCH 0/3] Make DMA slave ID 0 invalid to simplify platform data

[Paul Mundt]

On Tue, May 24, 2011 at 12:23:55PM +0200, Guennadi Liakhovetski wrote:
> Currently platforms, using the SDHI TMIO driver, have to specify negative 
> DMA slave IDs to force PIO mode. If default 0 is left, the driver 
> recognises it as a valid DMA slave ID and tries to acquire channels. This 
> fails then, because the driver cannot allocate two channels with the same 
> ID, but it would be better to just make 0 an invalid ID to prevent this.
> 
I suppose the ordering for this will be adding the invalid IDs for all of
the CPUs first and then updating the SDHI driver accordingly during -rc2
or something like that.

Re: [PATCH 0/3] Make DMA slave ID 0 invalid to simplify platform data

[Guennadi Liakhovetski]

On Wed, 25 May 2011, Paul Mundt wrote:

> On Tue, May 24, 2011 at 12:23:55PM +0200, Guennadi Liakhovetski wrote:
> > Currently platforms, using the SDHI TMIO driver, have to specify negative 
> > DMA slave IDs to force PIO mode. If default 0 is left, the driver 
> > recognises it as a valid DMA slave ID and tries to acquire channels. This 
> > fails then, because the driver cannot allocate two channels with the same 
> > ID, but it would be better to just make 0 an invalid ID to prevent this.
> > 
> I suppose the ordering for this will be adding the invalid IDs for all of
> the CPUs first and then updating the SDHI driver accordingly during -rc2
> or something like that.

Yes, otherwise it would break SDHI DMA on sh7757, where one of the IDs == 
0.

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/