diff mbox series

[v2] ASoC: qcom: soundwire: Add support for controlling audio CGCR from HLOS

Message ID 1652877755-25120-1-git-send-email-quic_srivasam@quicinc.com (mailing list archive)
State Accepted
Commit 32882881078bd8f8fae47ff69c102d9e691f5bb9
Headers show
Series [v2] ASoC: qcom: soundwire: Add support for controlling audio CGCR from HLOS | expand

Commit Message

Srinivasa Rao Mandadapu May 18, 2022, 12:42 p.m. UTC
Add support for controlling soundwire audio CGCR interface using clock
framework to make hclk ungating with software. As per new hardware
changes, software has to always ungate hclk if soundwire is operational
and keep it running. This requirement is for latest LPASS chipsets for
RX, TX and WSA path to work.

Signed-off-by: Srinivasa Rao Mandadapu <quic_srivasam@quicinc.com>
---
This patch set depends on:
    -- Clock driver patches for CGCR reset control support.
	--- https://patchwork.kernel.org/project/linux-arm-msm/list/?series=638002
	--- https://patchwork.kernel.org/project/linux-arm-msm/list/?series=637998	

Changes since v1:
    -- Add audio cgcr reset control in runtime PM resume handler.
    -- Update dependency list.

 drivers/soundwire/qcom.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

Comments

Stephen Boyd May 21, 2022, 3:13 a.m. UTC | #1
Quoting Srinivasa Rao Mandadapu (2022-05-18 05:42:35)
> diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c
> index da1ad7e..445e481 100644
> --- a/drivers/soundwire/qcom.c
> +++ b/drivers/soundwire/qcom.c
> @@ -1333,6 +1337,10 @@ static int qcom_swrm_probe(struct platform_device *pdev)
>         ctrl->bus.compute_params = &qcom_swrm_compute_params;
>         ctrl->bus.clk_stop_timeout = 300;
>
> +       ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, "swr_audio_cgcr");
> +       if (IS_ERR(ctrl->audio_cgcr))
> +               dev_err(dev, "Failed to get audio_cgcr reset required for soundwire-v1.6.0\n");

Why is there no return on error here? Is the reset optional?

> +
>         ret = qcom_swrm_get_port_config(ctrl);
>         if (ret)
>                 goto err_clk;
Srinivasa Rao Mandadapu May 24, 2022, 10:49 a.m. UTC | #2
On 5/21/2022 8:43 AM, Stephen Boyd wrote:
Thanks for your time Stephen!!!
> Quoting Srinivasa Rao Mandadapu (2022-05-18 05:42:35)
>> diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c
>> index da1ad7e..445e481 100644
>> --- a/drivers/soundwire/qcom.c
>> +++ b/drivers/soundwire/qcom.c
>> @@ -1333,6 +1337,10 @@ static int qcom_swrm_probe(struct platform_device *pdev)
>>          ctrl->bus.compute_params = &qcom_swrm_compute_params;
>>          ctrl->bus.clk_stop_timeout = 300;
>>
>> +       ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, "swr_audio_cgcr");
>> +       if (IS_ERR(ctrl->audio_cgcr))
>> +               dev_err(dev, "Failed to get audio_cgcr reset required for soundwire-v1.6.0\n");
> Why is there no return on error here? Is the reset optional?
Yes it's optional. For older platforms this is not required.
>
>> +
>>          ret = qcom_swrm_get_port_config(ctrl);
>>          if (ret)
>>                  goto err_clk;
Matthias Kaehlcke May 31, 2022, 10:32 p.m. UTC | #3
On Tue, May 24, 2022 at 04:19:47PM +0530, Srinivasa Rao Mandadapu wrote:
> 
> On 5/21/2022 8:43 AM, Stephen Boyd wrote:
> Thanks for your time Stephen!!!
> > Quoting Srinivasa Rao Mandadapu (2022-05-18 05:42:35)
> > > diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c
> > > index da1ad7e..445e481 100644
> > > --- a/drivers/soundwire/qcom.c
> > > +++ b/drivers/soundwire/qcom.c
> > > @@ -1333,6 +1337,10 @@ static int qcom_swrm_probe(struct platform_device *pdev)
> > >          ctrl->bus.compute_params = &qcom_swrm_compute_params;
> > >          ctrl->bus.clk_stop_timeout = 300;
> > > 
> > > +       ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, "swr_audio_cgcr");
> > > +       if (IS_ERR(ctrl->audio_cgcr))
> > > +               dev_err(dev, "Failed to get audio_cgcr reset required for soundwire-v1.6.0\n");
> > Why is there no return on error here? Is the reset optional?
> Yes it's optional. For older platforms this is not required.

If it's optional then either there should be no error message, or the
error message should only be logged when the version is >= 1.6.0. There
are few things worse than a kernel log riddled with misleading error
messages.
Srinivasa Rao Mandadapu June 1, 2022, 12:57 p.m. UTC | #4
On 6/1/2022 4:02 AM, Matthias Kaehlcke wrote:
Thanks for Your Time Matthias!!!
> On Tue, May 24, 2022 at 04:19:47PM +0530, Srinivasa Rao Mandadapu wrote:
>> On 5/21/2022 8:43 AM, Stephen Boyd wrote:
>> Thanks for your time Stephen!!!
>>> Quoting Srinivasa Rao Mandadapu (2022-05-18 05:42:35)
>>>> diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c
>>>> index da1ad7e..445e481 100644
>>>> --- a/drivers/soundwire/qcom.c
>>>> +++ b/drivers/soundwire/qcom.c
>>>> @@ -1333,6 +1337,10 @@ static int qcom_swrm_probe(struct platform_device *pdev)
>>>>           ctrl->bus.compute_params = &qcom_swrm_compute_params;
>>>>           ctrl->bus.clk_stop_timeout = 300;
>>>>
>>>> +       ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, "swr_audio_cgcr");
>>>> +       if (IS_ERR(ctrl->audio_cgcr))
>>>> +               dev_err(dev, "Failed to get audio_cgcr reset required for soundwire-v1.6.0\n");
>>> Why is there no return on error here? Is the reset optional?
>> Yes it's optional. For older platforms this is not required.
> If it's optional then either there should be no error message, or the
> error message should only be logged when the version is >= 1.6.0. There
> are few things worse than a kernel log riddled with misleading error
> messages.

In that case, it can be done like below. Kindly let me know your opinion 
on this.

if (ctrl->version >= 0x01060000) {
     ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, 
"swr_audio_cgcr");
         if (IS_ERR(ctrl->audio_cgcr)) {
             dev_err(dev, "Failed to get audio_cgcr reset required for 
soundwire-v1.6.0\n");
             ret = PTR_ERR(ctrl->audio_cgcr);
             goto err_clk;
         }
     }
Srinivas Kandagatla June 1, 2022, 1:06 p.m. UTC | #5
On 01/06/2022 13:57, Srinivasa Rao Mandadapu wrote:
> 
> On 6/1/2022 4:02 AM, Matthias Kaehlcke wrote:
> Thanks for Your Time Matthias!!!
>> On Tue, May 24, 2022 at 04:19:47PM +0530, Srinivasa Rao Mandadapu wrote:
>>> On 5/21/2022 8:43 AM, Stephen Boyd wrote:
>>> Thanks for your time Stephen!!!
>>>> Quoting Srinivasa Rao Mandadapu (2022-05-18 05:42:35)
>>>>> diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c
>>>>> index da1ad7e..445e481 100644
>>>>> --- a/drivers/soundwire/qcom.c
>>>>> +++ b/drivers/soundwire/qcom.c
>>>>> @@ -1333,6 +1337,10 @@ static int qcom_swrm_probe(struct 
>>>>> platform_device *pdev)
>>>>>           ctrl->bus.compute_params = &qcom_swrm_compute_params;
>>>>>           ctrl->bus.clk_stop_timeout = 300;
>>>>>
>>>>> +       ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, 
>>>>> "swr_audio_cgcr");
>>>>> +       if (IS_ERR(ctrl->audio_cgcr))
>>>>> +               dev_err(dev, "Failed to get audio_cgcr reset 
>>>>> required for soundwire-v1.6.0\n");
>>>> Why is there no return on error here? Is the reset optional?
>>> Yes it's optional. For older platforms this is not required.
>> If it's optional then either there should be no error message, or the
>> error message should only be logged when the version is >= 1.6.0. There
>> are few things worse than a kernel log riddled with misleading error
>> messages.
> 
> In that case, it can be done like below. Kindly let me know your opinion 
> on this.
> 
> if (ctrl->version >= 0x01060000) {

This is not true 1.7+ variants do not require anything as such.

Why not add a flag in struct qcom_swrm_data and pass it as part of 
of_match data specific to this version?

--srini
>      ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, 
> "swr_audio_cgcr");
>          if (IS_ERR(ctrl->audio_cgcr)) {
>              dev_err(dev, "Failed to get audio_cgcr reset required for 
> soundwire-v1.6.0\n");
>              ret = PTR_ERR(ctrl->audio_cgcr);
>              goto err_clk;
>          }
>      }
>
Srinivasa Rao Mandadapu June 1, 2022, 1:15 p.m. UTC | #6
On 6/1/2022 6:36 PM, Srinivas Kandagatla wrote:
Thanks for Your time Srini!!
>
>
> On 01/06/2022 13:57, Srinivasa Rao Mandadapu wrote:
>>
>> On 6/1/2022 4:02 AM, Matthias Kaehlcke wrote:
>> Thanks for Your Time Matthias!!!
>>> On Tue, May 24, 2022 at 04:19:47PM +0530, Srinivasa Rao Mandadapu 
>>> wrote:
>>>> On 5/21/2022 8:43 AM, Stephen Boyd wrote:
>>>> Thanks for your time Stephen!!!
>>>>> Quoting Srinivasa Rao Mandadapu (2022-05-18 05:42:35)
>>>>>> diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c
>>>>>> index da1ad7e..445e481 100644
>>>>>> --- a/drivers/soundwire/qcom.c
>>>>>> +++ b/drivers/soundwire/qcom.c
>>>>>> @@ -1333,6 +1337,10 @@ static int qcom_swrm_probe(struct 
>>>>>> platform_device *pdev)
>>>>>>           ctrl->bus.compute_params = &qcom_swrm_compute_params;
>>>>>>           ctrl->bus.clk_stop_timeout = 300;
>>>>>>
>>>>>> +       ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, 
>>>>>> "swr_audio_cgcr");
>>>>>> +       if (IS_ERR(ctrl->audio_cgcr))
>>>>>> +               dev_err(dev, "Failed to get audio_cgcr reset 
>>>>>> required for soundwire-v1.6.0\n");
>>>>> Why is there no return on error here? Is the reset optional?
>>>> Yes it's optional. For older platforms this is not required.
>>> If it's optional then either there should be no error message, or the
>>> error message should only be logged when the version is >= 1.6.0. There
>>> are few things worse than a kernel log riddled with misleading error
>>> messages.
>>
>> In that case, it can be done like below. Kindly let me know your 
>> opinion on this.
>>
>> if (ctrl->version >= 0x01060000) {
>
> This is not true 1.7+ variants do not require anything as such.

I think it applies for all upcoming versions as Qualcomm Hardware team. 
Here is the not from HW Team.

*/Issue: /*

HW CTL on AHB interface to SWR/Codec slave is default ON, the AHB clock 
is ungated for programmable cycles per transaction

(from valid htrans until slave ready or hysteresis count) and therefore 
hclk is not free-running.

Without a free-running HCLK, interrupt, CMD FIFO, slave status and other 
similar logic will have functional issues.

SWR has internal clock gating for address accesses but some logic is 
running on HCLK

*/Fix: /*

Registers to control HW_CTL is accessible via HLOS – so software has to 
always ungate hclk if soundwire is operational and keep it running

*/CSR: /*

SoundWire TX/RX.WSA :

Register: *LPASS_SWR_TX/RX/WSA_CONFIG_CGCR* | 0xXXXC010

	
	
	
	

0x3

	

0x1

*Bits*

	

*Field Name*

	

*Access*

	

*Description*

	

*Current PoR*

	

*SW Update*

1

	

HW_CTL

	

	

HW Dynamic Clock Gating Control Register 1: HW Controlled 0: SW Controlled

	

1

	

0

0

	

CLK_ENABLE

	

	

Enabling the clock when in SW Controlled Mode 1: Clock Enabled 0: Clock 
Disabled

	

1

	

1

>
> Why not add a flag in struct qcom_swrm_data and pass it as part of 
> of_match data specific to this version?
>
> --srini
>>      ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, 
>> "swr_audio_cgcr");
>>          if (IS_ERR(ctrl->audio_cgcr)) {
>>              dev_err(dev, "Failed to get audio_cgcr reset required 
>> for soundwire-v1.6.0\n");
>>              ret = PTR_ERR(ctrl->audio_cgcr);
>>              goto err_clk;
>>          }
>>      }
>>
Srinivas Kandagatla June 1, 2022, 1:42 p.m. UTC | #7
On 01/06/2022 14:15, Srinivasa Rao Mandadapu wrote:
>>>>>>> +       ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, 
>>>>>>> "swr_audio_cgcr");
>>>>>>> +       if (IS_ERR(ctrl->audio_cgcr))
>>>>>>> +               dev_err(dev, "Failed to get audio_cgcr reset 
>>>>>>> required for soundwire-v1.6.0\n");
>>>>>> Why is there no return on error here? Is the reset optional?
>>>>> Yes it's optional. For older platforms this is not required.
>>>> If it's optional then either there should be no error message, or the
>>>> error message should only be logged when the version is >= 1.6.0. There
>>>> are few things worse than a kernel log riddled with misleading error
>>>> messages.
>>>
>>> In that case, it can be done like below. Kindly let me know your 
>>> opinion on this.
>>>
>>> if (ctrl->version >= 0x01060000) {
>>
>> This is not true 1.7+ variants do not require anything as such.
> 
> I think it applies for all upcoming versions as Qualcomm Hardware team. 
> Here is the not from HW Team.

Am testing sm8450 which has 1.7.0 and it does not require/have such control.

I dont understand what is the issue in adding a flag to
struct qcom_swrm_data.

This should give finer control rather than matching anything > 1.6.


--srini
>
Matthias Kaehlcke June 1, 2022, 4:55 p.m. UTC | #8
On Wed, Jun 01, 2022 at 02:42:30PM +0100, Srinivas Kandagatla wrote:
> 
> 
> On 01/06/2022 14:15, Srinivasa Rao Mandadapu wrote:
> > > > > > > > +       ctrl->audio_cgcr =
> > > > > > > > devm_reset_control_get_exclusive(dev,
> > > > > > > > "swr_audio_cgcr");
> > > > > > > > +       if (IS_ERR(ctrl->audio_cgcr))
> > > > > > > > +               dev_err(dev, "Failed to get
> > > > > > > > audio_cgcr reset required for
> > > > > > > > soundwire-v1.6.0\n");
> > > > > > > Why is there no return on error here? Is the reset optional?
> > > > > > Yes it's optional. For older platforms this is not required.
> > > > > If it's optional then either there should be no error message, or the
> > > > > error message should only be logged when the version is >= 1.6.0. There
> > > > > are few things worse than a kernel log riddled with misleading error
> > > > > messages.
> > > > 
> > > > In that case, it can be done like below. Kindly let me know your
> > > > opinion on this.
> > > > 
> > > > if (ctrl->version >= 0x01060000) {
> > > 
> > > This is not true 1.7+ variants do not require anything as such.
> > 
> > I think it applies for all upcoming versions as Qualcomm Hardware team.
> > Here is the not from HW Team.
> 
> Am testing sm8450 which has 1.7.0 and it does not require/have such control.
> 
> I dont understand what is the issue in adding a flag to
> struct qcom_swrm_data.
> 
> This should give finer control rather than matching anything > 1.6.

I agree, a flag seems a suitable option.
Mark Brown June 7, 2022, 10:45 a.m. UTC | #9
On Wed, 18 May 2022 18:12:35 +0530, Srinivasa Rao Mandadapu wrote:
> Add support for controlling soundwire audio CGCR interface using clock
> framework to make hclk ungating with software. As per new hardware
> changes, software has to always ungate hclk if soundwire is operational
> and keep it running. This requirement is for latest LPASS chipsets for
> RX, TX and WSA path to work.
> 
> 
> [...]

Applied to

   https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git for-next

Thanks!

[1/1] ASoC: qcom: soundwire: Add support for controlling audio CGCR from HLOS
      commit: 32882881078bd8f8fae47ff69c102d9e691f5bb9

All being well this means that it will be integrated into the linux-next
tree (usually sometime in the next 24 hours) and sent to Linus during
the next merge window (or sooner if it is a bug fix), however if
problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing
and review of the tree, please engage with people reporting problems and
send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they
should be sent as incremental updates against current git, existing
patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying
to this mail.

Thanks,
Mark
diff mbox series

Patch

diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c
index da1ad7e..445e481 100644
--- a/drivers/soundwire/qcom.c
+++ b/drivers/soundwire/qcom.c
@@ -13,6 +13,7 @@ 
 #include <linux/of_device.h>
 #include <linux/pm_runtime.h>
 #include <linux/regmap.h>
+#include <linux/reset.h>
 #include <linux/slab.h>
 #include <linux/pm_wakeirq.h>
 #include <linux/slimbus.h>
@@ -142,6 +143,7 @@  struct qcom_swrm_ctrl {
 	struct device *dev;
 	struct regmap *regmap;
 	void __iomem *mmio;
+	struct reset_control *audio_cgcr;
 #ifdef CONFIG_DEBUG_FS
 	struct dentry *debugfs;
 #endif
@@ -656,6 +658,8 @@  static int qcom_swrm_init(struct qcom_swrm_ctrl *ctrl)
 	val = FIELD_PREP(SWRM_MCP_FRAME_CTRL_BANK_ROW_CTRL_BMSK, ctrl->rows_index);
 	val |= FIELD_PREP(SWRM_MCP_FRAME_CTRL_BANK_COL_CTRL_BMSK, ctrl->cols_index);
 
+	reset_control_reset(ctrl->audio_cgcr);
+
 	ctrl->reg_write(ctrl, SWRM_MCP_FRAME_CTRL_BANK_ADDR(0), val);
 
 	/* Enable Auto enumeration */
@@ -1333,6 +1337,10 @@  static int qcom_swrm_probe(struct platform_device *pdev)
 	ctrl->bus.compute_params = &qcom_swrm_compute_params;
 	ctrl->bus.clk_stop_timeout = 300;
 
+	ctrl->audio_cgcr = devm_reset_control_get_exclusive(dev, "swr_audio_cgcr");
+	if (IS_ERR(ctrl->audio_cgcr))
+		dev_err(dev, "Failed to get audio_cgcr reset required for soundwire-v1.6.0\n");
+
 	ret = qcom_swrm_get_port_config(ctrl);
 	if (ret)
 		goto err_clk;
@@ -1486,6 +1494,8 @@  static int __maybe_unused swrm_runtime_resume(struct device *dev)
 		qcom_swrm_get_device_status(ctrl);
 		sdw_handle_slave_status(&ctrl->bus, ctrl->status);
 	} else {
+		reset_control_reset(ctrl->audio_cgcr);
+
 		ctrl->reg_write(ctrl, SWRM_MCP_BUS_CTRL, SWRM_MCP_BUS_CLK_START);
 		ctrl->reg_write(ctrl, SWRM_INTERRUPT_CLEAR,
 			SWRM_INTERRUPT_STATUS_MASTER_CLASH_DET);