[PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Manikanta Mylavarapu]

In existing socs, I2C serial engine is sourced from XO (19.2MHz).
Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).

The existing map table is based on 19.2MHz. This patch incorporate
the clock map table to derive the SCL clock from the 32MHz source
clock frequency.

Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
---
Changes in v2:
	- Dropped IPQ5424 from the commit title
	- Added else part to assign geni_i2c_clk_map_19p2mhz to itr
	- Dropped MHZ macro and used HZ_PER_MHZ macro
	- Expanded SE to serial engine
	- Added the reason for 32MHz clock in commit message

 drivers/i2c/busses/i2c-qcom-geni.c | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
index 212336f724a6..22f2a0d83641 100644
--- a/drivers/i2c/busses/i2c-qcom-geni.c
+++ b/drivers/i2c/busses/i2c-qcom-geni.c
@@ -16,6 +16,7 @@
 #include <linux/pm_runtime.h>
 #include <linux/soc/qcom/geni-se.h>
 #include <linux/spinlock.h>
+#include <linux/units.h>
 
 #define SE_I2C_TX_TRANS_LEN		0x26c
 #define SE_I2C_RX_TRANS_LEN		0x270
@@ -146,18 +147,30 @@ struct geni_i2c_clk_fld {
  * clk_freq_out = t / t_cycle
  * source_clock = 19.2 MHz
  */
-static const struct geni_i2c_clk_fld geni_i2c_clk_map[] = {
+static const struct geni_i2c_clk_fld geni_i2c_clk_map_19p2mhz[] = {
 	{KHZ(100), 7, 10, 11, 26},
 	{KHZ(400), 2,  5, 12, 24},
 	{KHZ(1000), 1, 3,  9, 18},
 };
 
+/* source_clock = 32 MHz */
+static const struct geni_i2c_clk_fld geni_i2c_clk_map_32mhz[] = {
+	{KHZ(100), 7, 14, 18, 40},
+	{KHZ(400), 4,  3, 11, 20},
+	{KHZ(1000), 4, 3,  6, 15},
+};
+
 static int geni_i2c_clk_map_idx(struct geni_i2c_dev *gi2c)
 {
 	int i;
-	const struct geni_i2c_clk_fld *itr = geni_i2c_clk_map;
+	const struct geni_i2c_clk_fld *itr;
+
+	if (clk_get_rate(gi2c->se.clk) == 32 * HZ_PER_MHZ)
+		itr = geni_i2c_clk_map_32mhz;
+	else
+		itr = geni_i2c_clk_map_19p2mhz;
 
-	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map); i++, itr++) {
+	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map_19p2mhz); i++, itr++) {
 		if (itr->clk_freq_out == gi2c->clk_freq_out) {
 			gi2c->clk_fld = itr;
 			return 0;
-- 
2.34.1

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Mukesh Kumar Savaliya]

Hi Manikanta, My suggestion:

i2c: qcom-geni: Support systems with 32MHz serial engine clock
[Derive i2c clock from 32MHz Sourced clock]

On 9/26/2024 9:13 AM, Manikanta Mylavarapu wrote:
> In existing socs, I2C serial engine is sourced from XO (19.2MHz).
> Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
> 
> The existing map table is based on 19.2MHz. This patch incorporate
incorporate/s/incorporates
> the clock map table to derive the SCL clock from the 32MHz source
> clock frequency.
> 
> Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
> ---
> Changes in v2:
> 	- Dropped IPQ5424 from the commit title
> 	- Added else part to assign geni_i2c_clk_map_19p2mhz to itr
> 	- Dropped MHZ macro and used HZ_PER_MHZ macro
> 	- Expanded SE to serial engine
> 	- Added the reason for 32MHz clock in commit message
> 
>   drivers/i2c/busses/i2c-qcom-geni.c | 19 ++++++++++++++++---
>   1 file changed, 16 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> index 212336f724a6..22f2a0d83641 100644
> --- a/drivers/i2c/busses/i2c-qcom-geni.c
> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> @@ -16,6 +16,7 @@
>   #include <linux/pm_runtime.h>
>   #include <linux/soc/qcom/geni-se.h>
>   #include <linux/spinlock.h>
> +#include <linux/units.h>
>   
>   #define SE_I2C_TX_TRANS_LEN		0x26c
>   #define SE_I2C_RX_TRANS_LEN		0x270
> @@ -146,18 +147,30 @@ struct geni_i2c_clk_fld {
>    * clk_freq_out = t / t_cycle
>    * source_clock = 19.2 MHz
>    */
> -static const struct geni_i2c_clk_fld geni_i2c_clk_map[] = {
> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_19p2mhz[] = {
>   	{KHZ(100), 7, 10, 11, 26},
>   	{KHZ(400), 2,  5, 12, 24},
>   	{KHZ(1000), 1, 3,  9, 18},
>   };
>   
> +/* source_clock = 32 MHz */
> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_32mhz[] = {
> +	{KHZ(100), 7, 14, 18, 40},
> +	{KHZ(400), 4,  3, 11, 20},
> +	{KHZ(1000), 4, 3,  6, 15},
> +};
> +
>   static int geni_i2c_clk_map_idx(struct geni_i2c_dev *gi2c)
>   {
>   	int i;
> -	const struct geni_i2c_clk_fld *itr = geni_i2c_clk_map;
> +	const struct geni_i2c_clk_fld *itr;
> +
> +	if (clk_get_rate(gi2c->se.clk) == 32 * HZ_PER_MHZ)
> +		itr = geni_i2c_clk_map_32mhz;
> +	else
> +		itr = geni_i2c_clk_map_19p2mhz;
>   
> -	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map); i++, itr++) {
> +	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map_19p2mhz); i++, itr++) {
>   		if (itr->clk_freq_out == gi2c->clk_freq_out) {
>   			gi2c->clk_fld = itr;
>   			return 0;

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Manikanta Mylavarapu]

On 9/26/2024 11:44 AM, Mukesh Kumar Savaliya wrote:
> Hi Manikanta, My suggestion:
> 
> i2c: qcom-geni: Support systems with 32MHz serial engine clock
> [Derive i2c clock from 32MHz Sourced clock]
> 

As per Bjorn suggestion, i used this commit title "i2c: qcom-geni: Support systems with 32MHz serial engine clock".

> On 9/26/2024 9:13 AM, Manikanta Mylavarapu wrote:
>> In existing socs, I2C serial engine is sourced from XO (19.2MHz).
>> Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
>>
>> The existing map table is based on 19.2MHz. This patch incorporate
> incorporate/s/incorporates

Okay, sure.

Thanks & Regards,
Manikanta.

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Bjorn Andersson]

On Thu, Sep 26, 2024 at 11:44:20AM GMT, Mukesh Kumar Savaliya wrote:
> Hi Manikanta, My suggestion:
> 
> i2c: qcom-geni: Support systems with 32MHz serial engine clock
> [Derive i2c clock from 32MHz Sourced clock]

I'm fine with a more accurate subject, but to me your sentence indicates
"there's a choice of where the i2c clock is sourced from, now make it
come from the 32MHz clock".

Is this really what's going on? (Patch content and commit message
indicates no)

Is the "Sourced clock" the name of specific clock? Why does it have a
capital 'S'?

Regards,
Bjorn

> 
> On 9/26/2024 9:13 AM, Manikanta Mylavarapu wrote:
> > In existing socs, I2C serial engine is sourced from XO (19.2MHz).
> > Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
> > 
> > The existing map table is based on 19.2MHz. This patch incorporate
> incorporate/s/incorporates
> > the clock map table to derive the SCL clock from the 32MHz source
> > clock frequency.
> > 
> > Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
> > ---
> > Changes in v2:
> > 	- Dropped IPQ5424 from the commit title
> > 	- Added else part to assign geni_i2c_clk_map_19p2mhz to itr
> > 	- Dropped MHZ macro and used HZ_PER_MHZ macro
> > 	- Expanded SE to serial engine
> > 	- Added the reason for 32MHz clock in commit message
> > 
> >   drivers/i2c/busses/i2c-qcom-geni.c | 19 ++++++++++++++++---
> >   1 file changed, 16 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> > index 212336f724a6..22f2a0d83641 100644
> > --- a/drivers/i2c/busses/i2c-qcom-geni.c
> > +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> > @@ -16,6 +16,7 @@
> >   #include <linux/pm_runtime.h>
> >   #include <linux/soc/qcom/geni-se.h>
> >   #include <linux/spinlock.h>
> > +#include <linux/units.h>
> >   #define SE_I2C_TX_TRANS_LEN		0x26c
> >   #define SE_I2C_RX_TRANS_LEN		0x270
> > @@ -146,18 +147,30 @@ struct geni_i2c_clk_fld {
> >    * clk_freq_out = t / t_cycle
> >    * source_clock = 19.2 MHz
> >    */
> > -static const struct geni_i2c_clk_fld geni_i2c_clk_map[] = {
> > +static const struct geni_i2c_clk_fld geni_i2c_clk_map_19p2mhz[] = {
> >   	{KHZ(100), 7, 10, 11, 26},
> >   	{KHZ(400), 2,  5, 12, 24},
> >   	{KHZ(1000), 1, 3,  9, 18},
> >   };
> > +/* source_clock = 32 MHz */
> > +static const struct geni_i2c_clk_fld geni_i2c_clk_map_32mhz[] = {
> > +	{KHZ(100), 7, 14, 18, 40},
> > +	{KHZ(400), 4,  3, 11, 20},
> > +	{KHZ(1000), 4, 3,  6, 15},
> > +};
> > +
> >   static int geni_i2c_clk_map_idx(struct geni_i2c_dev *gi2c)
> >   {
> >   	int i;
> > -	const struct geni_i2c_clk_fld *itr = geni_i2c_clk_map;
> > +	const struct geni_i2c_clk_fld *itr;
> > +
> > +	if (clk_get_rate(gi2c->se.clk) == 32 * HZ_PER_MHZ)
> > +		itr = geni_i2c_clk_map_32mhz;
> > +	else
> > +		itr = geni_i2c_clk_map_19p2mhz;
> > -	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map); i++, itr++) {
> > +	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map_19p2mhz); i++, itr++) {
> >   		if (itr->clk_freq_out == gi2c->clk_freq_out) {
> >   			gi2c->clk_fld = itr;
> >   			return 0;
> 

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Vladimir Zapolskiy]

Hello Manikanta.

On 9/26/24 06:43, Manikanta Mylavarapu wrote:
> In existing socs, I2C serial engine is sourced from XO (19.2MHz).
> Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
> 
> The existing map table is based on 19.2MHz. This patch incorporate
> the clock map table to derive the SCL clock from the 32MHz source
> clock frequency.
> 
> Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
> ---
> Changes in v2:
> 	- Dropped IPQ5424 from the commit title
> 	- Added else part to assign geni_i2c_clk_map_19p2mhz to itr
> 	- Dropped MHZ macro and used HZ_PER_MHZ macro
> 	- Expanded SE to serial engine
> 	- Added the reason for 32MHz clock in commit message
> 
>   drivers/i2c/busses/i2c-qcom-geni.c | 19 ++++++++++++++++---
>   1 file changed, 16 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> index 212336f724a6..22f2a0d83641 100644
> --- a/drivers/i2c/busses/i2c-qcom-geni.c
> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> @@ -16,6 +16,7 @@
>   #include <linux/pm_runtime.h>
>   #include <linux/soc/qcom/geni-se.h>
>   #include <linux/spinlock.h>
> +#include <linux/units.h>
>   
>   #define SE_I2C_TX_TRANS_LEN		0x26c
>   #define SE_I2C_RX_TRANS_LEN		0x270
> @@ -146,18 +147,30 @@ struct geni_i2c_clk_fld {
>    * clk_freq_out = t / t_cycle
>    * source_clock = 19.2 MHz
>    */
> -static const struct geni_i2c_clk_fld geni_i2c_clk_map[] = {
> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_19p2mhz[] = {
>   	{KHZ(100), 7, 10, 11, 26},
>   	{KHZ(400), 2,  5, 12, 24},
>   	{KHZ(1000), 1, 3,  9, 18},
>   };
>   
> +/* source_clock = 32 MHz */
> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_32mhz[] = {
> +	{KHZ(100), 7, 14, 18, 40},
> +	{KHZ(400), 4,  3, 11, 20},
> +	{KHZ(1000), 4, 3,  6, 15},
> +};

Please double check the values.

This is what I get:
* for 100KHz: 32000000 / (40 * 7) ~ 114286, apparently 32000000 / (40 * 8) would
be a better fit, however it's unclear what would be proper t_high / t_low values,
* for 400KHz: it seems good,
* for 1000KHz: 32000000 / (15 * 4) ~ 533333, which is almost 1/2 of the wanted
bus frequency, so this one looks very wrong.

Do you have any ideas how to get better bus frequency settings?

> +
>   static int geni_i2c_clk_map_idx(struct geni_i2c_dev *gi2c)
>   {
>   	int i;
> -	const struct geni_i2c_clk_fld *itr = geni_i2c_clk_map;
> +	const struct geni_i2c_clk_fld *itr;
> +
> +	if (clk_get_rate(gi2c->se.clk) == 32 * HZ_PER_MHZ)
> +		itr = geni_i2c_clk_map_32mhz;
> +	else
> +		itr = geni_i2c_clk_map_19p2mhz;
>   
> -	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map); i++, itr++) {
> +	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map_19p2mhz); i++, itr++) {

Struct/data organization here is so weak, that here it's implicitly assumed
that sizes of two arrays are equal. It could be kept as is of course, just
pointing to it.

>   		if (itr->clk_freq_out == gi2c->clk_freq_out) {
>   			gi2c->clk_fld = itr;
>   			return 0;

--
Best wishes,
Vladimir

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Bjorn Andersson]

On Thu, Sep 26, 2024 at 01:28:37PM GMT, Vladimir Zapolskiy wrote:
> On 9/26/24 06:43, Manikanta Mylavarapu wrote:
> > diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
[..]
> > +
> >   static int geni_i2c_clk_map_idx(struct geni_i2c_dev *gi2c)
> >   {
> >   	int i;
> > -	const struct geni_i2c_clk_fld *itr = geni_i2c_clk_map;
> > +	const struct geni_i2c_clk_fld *itr;
> > +
> > +	if (clk_get_rate(gi2c->se.clk) == 32 * HZ_PER_MHZ)
> > +		itr = geni_i2c_clk_map_32mhz;
> > +	else
> > +		itr = geni_i2c_clk_map_19p2mhz;
> > -	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map); i++, itr++) {
> > +	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map_19p2mhz); i++, itr++) {
> 
> Struct/data organization here is so weak, that here it's implicitly assumed
> that sizes of two arrays are equal. It could be kept as is of course, just
> pointing to it.
> 

Thanks for pointing this out, Vladimir.

I'd prefer we fix it by adding a sentinel value to the arrays - because
this was only spotted due to the rename, the next guy will not be so
lucky.

Regards,
Bjorn

> >   		if (itr->clk_freq_out == gi2c->clk_freq_out) {
> >   			gi2c->clk_fld = itr;
> >   			return 0;
> 
> --
> Best wishes,
> Vladimir
> 

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Manikanta Mylavarapu]

On 9/26/2024 6:18 PM, Bjorn Andersson wrote:
> On Thu, Sep 26, 2024 at 01:28:37PM GMT, Vladimir Zapolskiy wrote:
>> On 9/26/24 06:43, Manikanta Mylavarapu wrote:
>>> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> [..]
>>> +
>>>   static int geni_i2c_clk_map_idx(struct geni_i2c_dev *gi2c)
>>>   {
>>>   	int i;
>>> -	const struct geni_i2c_clk_fld *itr = geni_i2c_clk_map;
>>> +	const struct geni_i2c_clk_fld *itr;
>>> +
>>> +	if (clk_get_rate(gi2c->se.clk) == 32 * HZ_PER_MHZ)
>>> +		itr = geni_i2c_clk_map_32mhz;
>>> +	else
>>> +		itr = geni_i2c_clk_map_19p2mhz;
>>> -	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map); i++, itr++) {
>>> +	for (i = 0; i < ARRAY_SIZE(geni_i2c_clk_map_19p2mhz); i++, itr++) {
>>
>> Struct/data organization here is so weak, that here it's implicitly assumed
>> that sizes of two arrays are equal. It could be kept as is of course, just
>> pointing to it.
>>
> 
> Thanks for pointing this out, Vladimir.
> 
> I'd prefer we fix it by adding a sentinel value to the arrays - because
> this was only spotted due to the rename, the next guy will not be so
> lucky.
> 
> Regards,
> Bjorn
> 

Okay, sure.

Thanks & Regards,
Manikanta.

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Manikanta Mylavarapu]

On 9/26/2024 3:58 PM, Vladimir Zapolskiy wrote:
> Hello Manikanta.
> 
> On 9/26/24 06:43, Manikanta Mylavarapu wrote:
>> In existing socs, I2C serial engine is sourced from XO (19.2MHz).
>> Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
>>
>> The existing map table is based on 19.2MHz. This patch incorporate
>> the clock map table to derive the SCL clock from the 32MHz source
>> clock frequency.
>>
>> Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
>> ---
>> Changes in v2:
>>     - Dropped IPQ5424 from the commit title
>>     - Added else part to assign geni_i2c_clk_map_19p2mhz to itr
>>     - Dropped MHZ macro and used HZ_PER_MHZ macro
>>     - Expanded SE to serial engine
>>     - Added the reason for 32MHz clock in commit message
>>
>>   drivers/i2c/busses/i2c-qcom-geni.c | 19 ++++++++++++++++---
>>   1 file changed, 16 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
>> index 212336f724a6..22f2a0d83641 100644
>> --- a/drivers/i2c/busses/i2c-qcom-geni.c
>> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
>> @@ -16,6 +16,7 @@
>>   #include <linux/pm_runtime.h>
>>   #include <linux/soc/qcom/geni-se.h>
>>   #include <linux/spinlock.h>
>> +#include <linux/units.h>
>>     #define SE_I2C_TX_TRANS_LEN        0x26c
>>   #define SE_I2C_RX_TRANS_LEN        0x270
>> @@ -146,18 +147,30 @@ struct geni_i2c_clk_fld {
>>    * clk_freq_out = t / t_cycle
>>    * source_clock = 19.2 MHz
>>    */
>> -static const struct geni_i2c_clk_fld geni_i2c_clk_map[] = {
>> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_19p2mhz[] = {
>>       {KHZ(100), 7, 10, 11, 26},
>>       {KHZ(400), 2,  5, 12, 24},
>>       {KHZ(1000), 1, 3,  9, 18},
>>   };
>>   +/* source_clock = 32 MHz */
>> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_32mhz[] = {
>> +    {KHZ(100), 7, 14, 18, 40},
>> +    {KHZ(400), 4,  3, 11, 20},
>> +    {KHZ(1000), 4, 3,  6, 15},
>> +};
> 
> Please double check the values.
> 
> This is what I get:
> * for 100KHz: 32000000 / (40 * 7) ~ 114286, apparently 32000000 / (40 * 8) would
> be a better fit, however it's unclear what would be proper t_high / t_low values,
> * for 400KHz: it seems good,
> * for 1000KHz: 32000000 / (15 * 4) ~ 533333, which is almost 1/2 of the wanted
> bus frequency, so this one looks very wrong.
> 
> Do you have any ideas how to get better bus frequency settings?
> 

Thanks for pointing this out.

I will double check and get back with the proper data.

Thanks & Regards,
Manikanta.

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Manikanta Mylavarapu]

On 9/29/2024 12:06 PM, Manikanta Mylavarapu wrote:
> 
> 
> On 9/26/2024 3:58 PM, Vladimir Zapolskiy wrote:
>> Hello Manikanta.
>>
>> On 9/26/24 06:43, Manikanta Mylavarapu wrote:
>>> In existing socs, I2C serial engine is sourced from XO (19.2MHz).
>>> Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
>>>
>>> The existing map table is based on 19.2MHz. This patch incorporate
>>> the clock map table to derive the SCL clock from the 32MHz source
>>> clock frequency.
>>>
>>> Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
>>> ---
>>> Changes in v2:
>>>     - Dropped IPQ5424 from the commit title
>>>     - Added else part to assign geni_i2c_clk_map_19p2mhz to itr
>>>     - Dropped MHZ macro and used HZ_PER_MHZ macro
>>>     - Expanded SE to serial engine
>>>     - Added the reason for 32MHz clock in commit message
>>>
>>>   drivers/i2c/busses/i2c-qcom-geni.c | 19 ++++++++++++++++---
>>>   1 file changed, 16 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
>>> index 212336f724a6..22f2a0d83641 100644
>>> --- a/drivers/i2c/busses/i2c-qcom-geni.c
>>> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
>>> @@ -16,6 +16,7 @@
>>>   #include <linux/pm_runtime.h>
>>>   #include <linux/soc/qcom/geni-se.h>
>>>   #include <linux/spinlock.h>
>>> +#include <linux/units.h>
>>>     #define SE_I2C_TX_TRANS_LEN        0x26c
>>>   #define SE_I2C_RX_TRANS_LEN        0x270
>>> @@ -146,18 +147,30 @@ struct geni_i2c_clk_fld {
>>>    * clk_freq_out = t / t_cycle
>>>    * source_clock = 19.2 MHz
>>>    */
>>> -static const struct geni_i2c_clk_fld geni_i2c_clk_map[] = {
>>> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_19p2mhz[] = {
>>>       {KHZ(100), 7, 10, 11, 26},
>>>       {KHZ(400), 2,  5, 12, 24},
>>>       {KHZ(1000), 1, 3,  9, 18},
>>>   };
>>>   +/* source_clock = 32 MHz */
>>> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_32mhz[] = {
>>> +    {KHZ(100), 7, 14, 18, 40},
>>> +    {KHZ(400), 4,  3, 11, 20},
>>> +    {KHZ(1000), 4, 3,  6, 15},
>>> +};
>>
>> Please double check the values.
>>
>> This is what I get:
>> * for 100KHz: 32000000 / (40 * 7) ~ 114286, apparently 32000000 / (40 * 8) would
>> be a better fit, however it's unclear what would be proper t_high / t_low values,
>> * for 400KHz: it seems good,
>> * for 1000KHz: 32000000 / (15 * 4) ~ 533333, which is almost 1/2 of the wanted
>> bus frequency, so this one looks very wrong.
>>
>> Do you have any ideas how to get better bus frequency settings?
>>
> 
> Thanks for pointing this out.
> 
> I will double check and get back with the proper data.
> 
> Thanks & Regards,
> Manikanta.
> 

Based on Qualcomm's internal hardware programming guide, below values need to be used for 100K & 1000K.
{KHZ(100), 8, 14, 18, 40}
{KHZ(1000), 2, 3,  6, 15}

I will update these values in next version.

Thanks & Regards,
Manikanta.

Re: [PATCH v2 1/1] i2c: qcom-geni: Support systems with 32MHz serial engine clock

[Vladimir Zapolskiy]

On 9/30/24 17:45, Manikanta Mylavarapu wrote:
> 
> 
> On 9/29/2024 12:06 PM, Manikanta Mylavarapu wrote:
>>
>>
>> On 9/26/2024 3:58 PM, Vladimir Zapolskiy wrote:
>>> Hello Manikanta.
>>>
>>> On 9/26/24 06:43, Manikanta Mylavarapu wrote:
>>>> In existing socs, I2C serial engine is sourced from XO (19.2MHz).
>>>> Where as in IPQ5424, I2C serial engine is sourced from GPLL0 (32MHz).
>>>>
>>>> The existing map table is based on 19.2MHz. This patch incorporate
>>>> the clock map table to derive the SCL clock from the 32MHz source
>>>> clock frequency.
>>>>
>>>> Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@quicinc.com>
>>>> ---
>>>> Changes in v2:
>>>>      - Dropped IPQ5424 from the commit title
>>>>      - Added else part to assign geni_i2c_clk_map_19p2mhz to itr
>>>>      - Dropped MHZ macro and used HZ_PER_MHZ macro
>>>>      - Expanded SE to serial engine
>>>>      - Added the reason for 32MHz clock in commit message
>>>>
>>>>    drivers/i2c/busses/i2c-qcom-geni.c | 19 ++++++++++++++++---
>>>>    1 file changed, 16 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
>>>> index 212336f724a6..22f2a0d83641 100644
>>>> --- a/drivers/i2c/busses/i2c-qcom-geni.c
>>>> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
>>>> @@ -16,6 +16,7 @@
>>>>    #include <linux/pm_runtime.h>
>>>>    #include <linux/soc/qcom/geni-se.h>
>>>>    #include <linux/spinlock.h>
>>>> +#include <linux/units.h>
>>>>      #define SE_I2C_TX_TRANS_LEN        0x26c
>>>>    #define SE_I2C_RX_TRANS_LEN        0x270
>>>> @@ -146,18 +147,30 @@ struct geni_i2c_clk_fld {
>>>>     * clk_freq_out = t / t_cycle
>>>>     * source_clock = 19.2 MHz
>>>>     */
>>>> -static const struct geni_i2c_clk_fld geni_i2c_clk_map[] = {
>>>> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_19p2mhz[] = {
>>>>        {KHZ(100), 7, 10, 11, 26},
>>>>        {KHZ(400), 2,  5, 12, 24},
>>>>        {KHZ(1000), 1, 3,  9, 18},
>>>>    };
>>>>    +/* source_clock = 32 MHz */
>>>> +static const struct geni_i2c_clk_fld geni_i2c_clk_map_32mhz[] = {
>>>> +    {KHZ(100), 7, 14, 18, 40},
>>>> +    {KHZ(400), 4,  3, 11, 20},
>>>> +    {KHZ(1000), 4, 3,  6, 15},
>>>> +};
>>>
>>> Please double check the values.
>>>
>>> This is what I get:
>>> * for 100KHz: 32000000 / (40 * 7) ~ 114286, apparently 32000000 / (40 * 8) would
>>> be a better fit, however it's unclear what would be proper t_high / t_low values,
>>> * for 400KHz: it seems good,
>>> * for 1000KHz: 32000000 / (15 * 4) ~ 533333, which is almost 1/2 of the wanted
>>> bus frequency, so this one looks very wrong.
>>>
>>> Do you have any ideas how to get better bus frequency settings?
>>>
>>
>> Thanks for pointing this out.
>>
>> I will double check and get back with the proper data.
>>
>> Thanks & Regards,
>> Manikanta.
>>
> 
> Based on Qualcomm's internal hardware programming guide, below values need to be used for 100K & 1000K.
> {KHZ(100), 8, 14, 18, 40}
> {KHZ(1000), 2, 3,  6, 15}
> 
> I will update these values in next version.

Great! These values make much more sence, thank you.

--
Best wishes,
Vladimir