dmaengine: qcom: bam_dma: make channels/EEs optional in DT with clock

[PATCH] dmaengine: qcom: bam_dma: make channels/EEs optional in DT with clock

Posted by Stephan Gerhold 2 years, 8 months ago

If we have a BAM clock in the DT we are able to turn on the BAM
controller while probing, so there is no need to read "num-channels"
and "qcom,num-ees" from the DT. It can be read more accurately directly
from the identification registers of the BAM.

This simplifies setting up typical controlled-remotely BAM DMAs in the
DT that can be turned on via a clock (e.g. the BLSP DMA).

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
---
 drivers/dma/qcom/bam_dma.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/dma/qcom/bam_dma.c b/drivers/dma/qcom/bam_dma.c
index 1e47d27e1f81..4c3eb972039d 100644
--- a/drivers/dma/qcom/bam_dma.c
+++ b/drivers/dma/qcom/bam_dma.c
@@ -1272,7 +1272,15 @@ static int bam_dma_probe(struct platform_device *pdev)
 	bdev->powered_remotely = of_property_read_bool(pdev->dev.of_node,
 						"qcom,powered-remotely");
 
-	if (bdev->controlled_remotely || bdev->powered_remotely) {
+	if (bdev->controlled_remotely || bdev->powered_remotely)
+		bdev->bamclk = devm_clk_get_optional(bdev->dev, "bam_clk");
+	else
+		bdev->bamclk = devm_clk_get(bdev->dev, "bam_clk");
+
+	if (IS_ERR(bdev->bamclk))
+		return PTR_ERR(bdev->bamclk);
+
+	if (!bdev->bamclk) {
 		ret = of_property_read_u32(pdev->dev.of_node, "num-channels",
 					   &bdev->num_channels);
 		if (ret)
@@ -1284,14 +1292,6 @@ static int bam_dma_probe(struct platform_device *pdev)
 			dev_err(bdev->dev, "num-ees unspecified in dt\n");
 	}
 
-	if (bdev->controlled_remotely || bdev->powered_remotely)
-		bdev->bamclk = devm_clk_get_optional(bdev->dev, "bam_clk");
-	else
-		bdev->bamclk = devm_clk_get(bdev->dev, "bam_clk");
-
-	if (IS_ERR(bdev->bamclk))
-		return PTR_ERR(bdev->bamclk);
-
 	ret = clk_prepare_enable(bdev->bamclk);
 	if (ret) {
 		dev_err(bdev->dev, "failed to prepare/enable clock\n");

---
base-commit: 1c677f238f92ba0a329b7c13220f38b396872806
change-id: 20230518-bamclk-dt-d44bae47b337

Best regards,
-- 
Stephan Gerhold <stephan@gerhold.net>

Re: [PATCH] dmaengine: qcom: bam_dma: make channels/EEs optional in DT with clock

Posted by Bhupesh Sharma 2 years, 8 months ago

Hi Stephan,

On Thu, 18 May 2023 at 14:56, Stephan Gerhold <stephan@gerhold.net> wrote:
>
> If we have a BAM clock in the DT we are able to turn on the BAM
> controller while probing, so there is no need to read "num-channels"
> and "qcom,num-ees" from the DT. It can be read more accurately directly
> from the identification registers of the BAM.
>
> This simplifies setting up typical controlled-remotely BAM DMAs in the
> DT that can be turned on via a clock (e.g. the BLSP DMA).

Can you please list which qcom board(s) you tested this patch on?

Thanks,
Bhupesh

> Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
> ---
>  drivers/dma/qcom/bam_dma.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/dma/qcom/bam_dma.c b/drivers/dma/qcom/bam_dma.c
> index 1e47d27e1f81..4c3eb972039d 100644
> --- a/drivers/dma/qcom/bam_dma.c
> +++ b/drivers/dma/qcom/bam_dma.c
> @@ -1272,7 +1272,15 @@ static int bam_dma_probe(struct platform_device *pdev)
>         bdev->powered_remotely = of_property_read_bool(pdev->dev.of_node,
>                                                 "qcom,powered-remotely");
>
> -       if (bdev->controlled_remotely || bdev->powered_remotely) {
> +       if (bdev->controlled_remotely || bdev->powered_remotely)
> +               bdev->bamclk = devm_clk_get_optional(bdev->dev, "bam_clk");
> +       else
> +               bdev->bamclk = devm_clk_get(bdev->dev, "bam_clk");
> +
> +       if (IS_ERR(bdev->bamclk))
> +               return PTR_ERR(bdev->bamclk);
> +
> +       if (!bdev->bamclk) {
>                 ret = of_property_read_u32(pdev->dev.of_node, "num-channels",
>                                            &bdev->num_channels);
>                 if (ret)
> @@ -1284,14 +1292,6 @@ static int bam_dma_probe(struct platform_device *pdev)
>                         dev_err(bdev->dev, "num-ees unspecified in dt\n");
>         }
>
> -       if (bdev->controlled_remotely || bdev->powered_remotely)
> -               bdev->bamclk = devm_clk_get_optional(bdev->dev, "bam_clk");
> -       else
> -               bdev->bamclk = devm_clk_get(bdev->dev, "bam_clk");
> -
> -       if (IS_ERR(bdev->bamclk))
> -               return PTR_ERR(bdev->bamclk);
> -
>         ret = clk_prepare_enable(bdev->bamclk);
>         if (ret) {
>                 dev_err(bdev->dev, "failed to prepare/enable clock\n");
>
> ---
> base-commit: 1c677f238f92ba0a329b7c13220f38b396872806
> change-id: 20230518-bamclk-dt-d44bae47b337
>
> Best regards,
> --
> Stephan Gerhold <stephan@gerhold.net>
>

Re: [PATCH] dmaengine: qcom: bam_dma: make channels/EEs optional in DT with clock

Posted by Stephan Gerhold 2 years, 8 months ago

On Thu, May 18, 2023 at 04:43:57PM +0530, Bhupesh Sharma wrote:
> On Thu, 18 May 2023 at 14:56, Stephan Gerhold <stephan@gerhold.net> wrote:
> >
> > If we have a BAM clock in the DT we are able to turn on the BAM
> > controller while probing, so there is no need to read "num-channels"
> > and "qcom,num-ees" from the DT. It can be read more accurately directly
> > from the identification registers of the BAM.
> >
> > This simplifies setting up typical controlled-remotely BAM DMAs in the
> > DT that can be turned on via a clock (e.g. the BLSP DMA).
> 
> Can you please list which qcom board(s) you tested this patch on?
> 

It works fine at least on MSM8916/DB410c (for blsp_dma) and MDM9607
(blsp_dma and qpic_dma (for NAND)). More testing would be much
appreciated of course!

Personally I don't see much of a risk: If enabling the clock doesn't
actually enable the BAM controller, then the clock probably does not
belong to the BAM in the first place... :)

Thanks,
Stephan

Re: [PATCH] dmaengine: qcom: bam_dma: make channels/EEs optional in DT with clock

Posted by Bhupesh Sharma 2 years, 8 months ago

On Thu, 18 May 2023 at 16:51, Stephan Gerhold <stephan@gerhold.net> wrote:
>
> On Thu, May 18, 2023 at 04:43:57PM +0530, Bhupesh Sharma wrote:
> > On Thu, 18 May 2023 at 14:56, Stephan Gerhold <stephan@gerhold.net> wrote:
> > >
> > > If we have a BAM clock in the DT we are able to turn on the BAM
> > > controller while probing, so there is no need to read "num-channels"
> > > and "qcom,num-ees" from the DT. It can be read more accurately directly
> > > from the identification registers of the BAM.
> > >
> > > This simplifies setting up typical controlled-remotely BAM DMAs in the
> > > DT that can be turned on via a clock (e.g. the BLSP DMA).
> >
> > Can you please list which qcom board(s) you tested this patch on?
> >
>
> It works fine at least on MSM8916/DB410c (for blsp_dma) and MDM9607
> (blsp_dma and qpic_dma (for NAND)). More testing would be much
> appreciated of course!

I tested this yesterday on RB1/RB2, RB5 and saw no improvement, so was wondering
why exactly is this needed and which platforms are impacted.

> Personally I don't see much of a risk: If enabling the clock doesn't
> actually enable the BAM controller, then the clock probably does not
> belong to the BAM in the first place... :)

Right, but I think the commit message needs a bit more clarity to
reflect that it is now proposed to check for the bam_clk presence
earlier in the _probe flow (as compared to earlier).

Thanks.

Re: [PATCH] dmaengine: qcom: bam_dma: make channels/EEs optional in DT with clock

Posted by Stephan Gerhold 2 years, 8 months ago

On Fri, May 19, 2023 at 02:40:21PM +0530, Bhupesh Sharma wrote:
> On Thu, 18 May 2023 at 16:51, Stephan Gerhold <stephan@gerhold.net> wrote:
> >
> > On Thu, May 18, 2023 at 04:43:57PM +0530, Bhupesh Sharma wrote:
> > > On Thu, 18 May 2023 at 14:56, Stephan Gerhold <stephan@gerhold.net> wrote:
> > > >
> > > > If we have a BAM clock in the DT we are able to turn on the BAM
> > > > controller while probing, so there is no need to read "num-channels"
> > > > and "qcom,num-ees" from the DT. It can be read more accurately directly
> > > > from the identification registers of the BAM.
> > > >
> > > > This simplifies setting up typical controlled-remotely BAM DMAs in the
> > > > DT that can be turned on via a clock (e.g. the BLSP DMA).
> > >
> > > Can you please list which qcom board(s) you tested this patch on?
> > >
> >
> > It works fine at least on MSM8916/DB410c (for blsp_dma) and MDM9607
> > (blsp_dma and qpic_dma (for NAND)). More testing would be much
> > appreciated of course!
> 
> I tested this yesterday on RB1/RB2, RB5 and saw no improvement, so was wondering
> why exactly is this needed and which platforms are impacted.
> 

RB1/RB2 should be able to benefit from this for the cryptobam if you add
the rpmcc clock to it, see my reply in [1].

[1]: https://lore.kernel.org/linux-arm-msm/ZGdLCdSof027mk5u@gerhold.net/

> > Personally I don't see much of a risk: If enabling the clock doesn't
> > actually enable the BAM controller, then the clock probably does not
> > belong to the BAM in the first place... :)
> 
> Right, but I think the commit message needs a bit more clarity to
> reflect that it is now proposed to check for the bam_clk presence
> earlier in the _probe flow (as compared to earlier).
> 

Sure, I will try to clarify the commit message a bit in v2.

Thanks,
Stephan