net: phy: at803x: add QCA8084 ethernet phy support

  Add qca8084 PHY support, which is four-port PHY with maximum
link capability 2.5G, the features of each port is almost same
as QCA8081 and slave seed config is not needed.

There are some initialization configurations needed.
1. Configuring qca8084 related initializations including
MSE detect threshold and ADC clock edge invert.
2. Add the additional configurations for the CDT feature.

Signed-off-by: Luo Jie <quic_luoj@quicinc.com>
---
 drivers/net/phy/at803x.c | 40 +++++++++++++++++++++++++++++++++++++---
 1 file changed, 37 insertions(+), 3 deletions(-)
  

>  #define QCA8081_PHY_ID				0x004dd101
> +#define QCA8081_PHY_MASK			0xffffff00

That is an unusual mask. Please check it is correct. All you should
need its PHY_ID_MATCH_EXACT, PHY_ID_MATCH_MODEL, PHY_ID_MATCH_VENDOR.

> @@ -1767,6 +1781,20 @@ static int qca808x_config_init(struct phy_device *phydev)
>  {
>  	int ret;
>  
> +	if (phydev->phy_id == QCA8084_PHY_ID) {
> +		/* Invert ADC clock edge */
> +		ret = at803x_debug_reg_mask(phydev, QCA8084_ADC_CLK_SEL,
> +					    QCA8084_ADC_CLK_SEL_ACLK,
> +					    FIELD_PREP(QCA8084_ADC_CLK_SEL_ACLK,
> +						       QCA8084_ADC_CLK_SEL_ACLK_FALL));
> +		if (ret < 0)
> +			return ret;
> +
> +		/* Adjust MSE threshold value to avoid link issue with some link partner */
> +		return phy_write_mmd(phydev, MDIO_MMD_PMAPMD,
> +				QCA8084_MSE_THRESHOLD, QCA8084_MSE_THRESHOLD_2P5G_VAL);
> +	}
> +

Please add a qca8084_config_init() and use that from the phy_driver
structure.

>  	/* Active adc&vga on 802.3az for the link 1000M and 100M */
>  	ret = phy_modify_mmd(phydev, MDIO_MMD_PCS, QCA808X_PHY_MMD3_ADDR_CLD_CTRL7,
>  			QCA808X_8023AZ_AFE_CTRL_MASK, QCA808X_8023AZ_AFE_EN);
> @@ -1958,6 +1986,11 @@ static int qca808x_cable_test_start(struct phy_device *phydev)
>  	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807a, 0xc060);
>  	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807e, 0xb060);
>  
> +	if (phydev->phy_id == QCA8084_PHY_ID) {
> +		phy_write_mmd(phydev, MDIO_MMD_PCS, 0x8075, 0xa060);
> +		phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807f, 0x1eb0);
> +	}
> +

Please add a comment what this is doing.

>  }, {
>  	/* Qualcomm QCA8081 */
> -	PHY_ID_MATCH_EXACT(QCA8081_PHY_ID),
> -	.name			= "Qualcomm QCA8081",
> +	.phy_id			= QCA8081_PHY_ID,
> +	.phy_id_mask		= QCA8081_PHY_MASK,
> +	.name			= "Qualcomm QCA808X",

Please add a new entry for the 8084. 

       Andrew
  

On 11/3/2023 9:01 PM, Andrew Lunn wrote:
>>   #define QCA8081_PHY_ID				0x004dd101
>> +#define QCA8081_PHY_MASK			0xffffff00
> 
> That is an unusual mask. Please check it is correct. All you should
> need its PHY_ID_MATCH_EXACT, PHY_ID_MATCH_MODEL, PHY_ID_MATCH_VENDOR.

Thanks Andrew for the review.
The PHY ID of qca8084 is correct, i will update to use 
PHY_ID_MATCH_EXACT in the new added entry for qca8084.

> 
>> @@ -1767,6 +1781,20 @@ static int qca808x_config_init(struct phy_device *phydev)
>>   {
>>   	int ret;
>>   
>> +	if (phydev->phy_id == QCA8084_PHY_ID) {
>> +		/* Invert ADC clock edge */
>> +		ret = at803x_debug_reg_mask(phydev, QCA8084_ADC_CLK_SEL,
>> +					    QCA8084_ADC_CLK_SEL_ACLK,
>> +					    FIELD_PREP(QCA8084_ADC_CLK_SEL_ACLK,
>> +						       QCA8084_ADC_CLK_SEL_ACLK_FALL));
>> +		if (ret < 0)
>> +			return ret;
>> +
>> +		/* Adjust MSE threshold value to avoid link issue with some link partner */
>> +		return phy_write_mmd(phydev, MDIO_MMD_PMAPMD,
>> +				QCA8084_MSE_THRESHOLD, QCA8084_MSE_THRESHOLD_2P5G_VAL);
>> +	}
>> +
> 
> Please add a qca8084_config_init() and use that from the phy_driver
> structure.

OK.

> 
>>   	/* Active adc&vga on 802.3az for the link 1000M and 100M */
>>   	ret = phy_modify_mmd(phydev, MDIO_MMD_PCS, QCA808X_PHY_MMD3_ADDR_CLD_CTRL7,
>>   			QCA808X_8023AZ_AFE_CTRL_MASK, QCA808X_8023AZ_AFE_EN);
>> @@ -1958,6 +1986,11 @@ static int qca808x_cable_test_start(struct phy_device *phydev)
>>   	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807a, 0xc060);
>>   	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807e, 0xb060);
>>   
>> +	if (phydev->phy_id == QCA8084_PHY_ID) {
>> +		phy_write_mmd(phydev, MDIO_MMD_PCS, 0x8075, 0xa060);
>> +		phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807f, 0x1eb0);
>> +	}
>> +
> 
> Please add a comment what this is doing.

Ok, will add comments in the next patch.

> 
>>   }, {
>>   	/* Qualcomm QCA8081 */
>> -	PHY_ID_MATCH_EXACT(QCA8081_PHY_ID),
>> -	.name			= "Qualcomm QCA8081",
>> +	.phy_id			= QCA8081_PHY_ID,
>> +	.phy_id_mask		= QCA8081_PHY_MASK,
>> +	.name			= "Qualcomm QCA808X",
> 
> Please add a new entry for the 8084.
> 
>         Andrew

Will add it in the next patch, thanks.
  

On Sat, Nov 04, 2023 at 02:25:25PM +0800, Jie Luo wrote:
> 
> 
> On 11/3/2023 9:01 PM, Andrew Lunn wrote:
> > >   #define QCA8081_PHY_ID				0x004dd101
> > > +#define QCA8081_PHY_MASK			0xffffff00
> > 
> > That is an unusual mask. Please check it is correct. All you should
> > need its PHY_ID_MATCH_EXACT, PHY_ID_MATCH_MODEL, PHY_ID_MATCH_VENDOR.
> 
> Thanks Andrew for the review.
> The PHY ID of qca8084 is correct, i will update to use PHY_ID_MATCH_EXACT in
> the new added entry for qca8084.

Note, i asked about the mask, not the ID. Is PHY_ID_MATCH_EXACT maybe
too exact? Is there the option for different revisions of the PHY? Can
one entry in the table be used for multiple revisions?


    Andrew

---
pw-bot: cr
  

On Fri, Nov 03, 2023 at 08:35:37PM +0800, Luo Jie wrote:
> Add qca8084 PHY support, which is four-port PHY with maximum
> link capability 2.5G, the features of each port is almost same
> as QCA8081 and slave seed config is not needed.
> 
> There are some initialization configurations needed.
> 1. Configuring qca8084 related initializations including
> MSE detect threshold and ADC clock edge invert.
> 2. Add the additional configurations for the CDT feature.
> 
> Signed-off-by: Luo Jie <quic_luoj@quicinc.com>
> ---
>  drivers/net/phy/at803x.c | 40 +++++++++++++++++++++++++++++++++++++---
>  1 file changed, 37 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/phy/at803x.c b/drivers/net/phy/at803x.c
> index 37fb033e1c29..4124eb76d835 100644
> --- a/drivers/net/phy/at803x.c
> +++ b/drivers/net/phy/at803x.c
> @@ -176,6 +176,8 @@
>  #define AT8030_PHY_ID_MASK			0xffffffef
>  
>  #define QCA8081_PHY_ID				0x004dd101
> +#define QCA8081_PHY_MASK			0xffffff00
> +#define QCA8084_PHY_ID				0x004dd180
...
> @@ -2207,8 +2240,9 @@ static struct phy_driver at803x_driver[] = {
>  	.resume			= qca83xx_resume,
>  }, {
>  	/* Qualcomm QCA8081 */
> -	PHY_ID_MATCH_EXACT(QCA8081_PHY_ID),
> -	.name			= "Qualcomm QCA8081",
> +	.phy_id			= QCA8081_PHY_ID,
> +	.phy_id_mask		= QCA8081_PHY_MASK,
> +	.name			= "Qualcomm QCA808X",
...
> @@ -2241,7 +2275,7 @@ static struct mdio_device_id __maybe_unused atheros_tbl[] = {
>  	{ PHY_ID_MATCH_EXACT(QCA8327_A_PHY_ID) },
>  	{ PHY_ID_MATCH_EXACT(QCA8327_B_PHY_ID) },
>  	{ PHY_ID_MATCH_EXACT(QCA9561_PHY_ID) },
> -	{ PHY_ID_MATCH_EXACT(QCA8081_PHY_ID) },
> +	{ QCA8081_PHY_ID, QCA8081_PHY_MASK},

So, in summary from the above, what you're doing is using the pair of
QCA8081_PHY_ID, QCA8081_PHY_MASK to match not only QCA8081 but also
QCA8084. This is confusing.

Are there any other parts that QCA808X would correspond with which
would not be compatible with the above? E.g. QCA8082, QCA8083, QCA8088
etc.

If there are, then the correct approach would be to list them
separately in atheros_tbl, and also have separate driver entries in
at803x_driver so it's unambiguous.

If we keep this approach, then I would suggest:

#define QCA808X_PHY_ID		0x004dd100
#define QCA808X_PHY_MASK	GENMASK(31, 8)

to make it explicit that this phy ID/mask pair is matching several
devices, rather than re-using the QCA8081_PHY_ID definition.


The next point - what about the revision field which occupies bits 3:0
in these:

>  static bool qca808x_has_fast_retrain_or_slave_seed(struct phy_device *phydev)
>  {
> +	if (phydev->phy_id == QCA8084_PHY_ID)
> +		return false;
> +
...
> @@ -1767,6 +1781,20 @@ static int qca808x_config_init(struct phy_device *phydev)
>  {
>  	int ret;
>  
> +	if (phydev->phy_id == QCA8084_PHY_ID) {
> +		/* Invert ADC clock edge */
...
> @@ -1958,6 +1986,11 @@ static int qca808x_cable_test_start(struct phy_device *phydev)
>  	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807a, 0xc060);
>  	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807e, 0xb060);
>  
> +	if (phydev->phy_id == QCA8084_PHY_ID) {

Do these need to be exact matches, or should the revision field be
ignored? If so, consider using phy_id_compare(), or if you end up with
separate driver entries, consider using phydev_id_compare().

Thanks.
  

On 11/4/2023 10:19 PM, Andrew Lunn wrote:
> On Sat, Nov 04, 2023 at 02:25:25PM +0800, Jie Luo wrote:
>>
>>
>> On 11/3/2023 9:01 PM, Andrew Lunn wrote:
>>>>    #define QCA8081_PHY_ID				0x004dd101
>>>> +#define QCA8081_PHY_MASK			0xffffff00
>>>
>>> That is an unusual mask. Please check it is correct. All you should
>>> need its PHY_ID_MATCH_EXACT, PHY_ID_MATCH_MODEL, PHY_ID_MATCH_VENDOR.
>>
>> Thanks Andrew for the review.
>> The PHY ID of qca8084 is correct, i will update to use PHY_ID_MATCH_EXACT in
>> the new added entry for qca8084.
> 
> Note, i asked about the mask, not the ID. Is PHY_ID_MATCH_EXACT maybe
> too exact? Is there the option for different revisions of the PHY? Can
> one entry in the table be used for multiple revisions?
> 
> 
>      Andrew
> 
> ---
> pw-bot: cr

Sure, Andrew, qca8084 is the different model chip from qca8081, and 
there is only one PHY ID used currently for the qca8084(4-port) and 
qca8082(2-port), but we can use PHY_ID_MATCH_MODEL for new added entry 
of qca8084 and qca8082.
  

On 11/5/2023 3:35 AM, Russell King (Oracle) wrote:
> On Fri, Nov 03, 2023 at 08:35:37PM +0800, Luo Jie wrote:
>> Add qca8084 PHY support, which is four-port PHY with maximum
>> link capability 2.5G, the features of each port is almost same
>> as QCA8081 and slave seed config is not needed.
>>
>> There are some initialization configurations needed.
>> 1. Configuring qca8084 related initializations including
>> MSE detect threshold and ADC clock edge invert.
>> 2. Add the additional configurations for the CDT feature.
>>
>> Signed-off-by: Luo Jie <quic_luoj@quicinc.com>
>> ---
>>   drivers/net/phy/at803x.c | 40 +++++++++++++++++++++++++++++++++++++---
>>   1 file changed, 37 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/net/phy/at803x.c b/drivers/net/phy/at803x.c
>> index 37fb033e1c29..4124eb76d835 100644
>> --- a/drivers/net/phy/at803x.c
>> +++ b/drivers/net/phy/at803x.c
>> @@ -176,6 +176,8 @@
>>   #define AT8030_PHY_ID_MASK			0xffffffef
>>   
>>   #define QCA8081_PHY_ID				0x004dd101
>> +#define QCA8081_PHY_MASK			0xffffff00
>> +#define QCA8084_PHY_ID				0x004dd180
> ...
>> @@ -2207,8 +2240,9 @@ static struct phy_driver at803x_driver[] = {
>>   	.resume			= qca83xx_resume,
>>   }, {
>>   	/* Qualcomm QCA8081 */
>> -	PHY_ID_MATCH_EXACT(QCA8081_PHY_ID),
>> -	.name			= "Qualcomm QCA8081",
>> +	.phy_id			= QCA8081_PHY_ID,
>> +	.phy_id_mask		= QCA8081_PHY_MASK,
>> +	.name			= "Qualcomm QCA808X",
> ...
>> @@ -2241,7 +2275,7 @@ static struct mdio_device_id __maybe_unused atheros_tbl[] = {
>>   	{ PHY_ID_MATCH_EXACT(QCA8327_A_PHY_ID) },
>>   	{ PHY_ID_MATCH_EXACT(QCA8327_B_PHY_ID) },
>>   	{ PHY_ID_MATCH_EXACT(QCA9561_PHY_ID) },
>> -	{ PHY_ID_MATCH_EXACT(QCA8081_PHY_ID) },
>> +	{ QCA8081_PHY_ID, QCA8081_PHY_MASK},
> 
> So, in summary from the above, what you're doing is using the pair of
> QCA8081_PHY_ID, QCA8081_PHY_MASK to match not only QCA8081 but also
> QCA8084. This is confusing.

Yes, Russell.
qca8084 is the different PHY model compared with the existed qca8081, 
qca8084 needs the extra PCS configuration and clock configuration, which 
will be pushed to the PCS driver.

I will update to use PHY_ID_MATCH_MODEL for the new added entry of 
qca8084 to distinguish qca8081.

> 
> Are there any other parts that QCA808X would correspond with which
> would not be compatible with the above? E.g. QCA8082, QCA8083, QCA8088
> etc.

The new added PHY chip qca8082, qca8084 and qca8085 use the same PHY ID, 
so the PHY_ID_MATCH_MODEL should be able to cover the new added entry.

> 
> If there are, then the correct approach would be to list them
> separately in atheros_tbl, and also have separate driver entries in
> at803x_driver so it's unambiguous.
> 
> If we keep this approach, then I would suggest:
> 
> #define QCA808X_PHY_ID		0x004dd100
> #define QCA808X_PHY_MASK	GENMASK(31, 8)
> 
> to make it explicit that this phy ID/mask pair is matching several
> devices, rather than re-using the QCA8081_PHY_ID definition.

Thanks for the suggestion. the PHY_ID_MATCH_MODEL should match the new 
added PHY devices.

> 
> 
> The next point - what about the revision field which occupies bits 3:0
> in these:

bits[3:0] is for the revision number, currently there is only one PHY ID 
(revision number is 0)used for the new added series PHY chip.

> 
>>   static bool qca808x_has_fast_retrain_or_slave_seed(struct phy_device *phydev)
>>   {
>> +	if (phydev->phy_id == QCA8084_PHY_ID)
>> +		return false;
>> +
> ...
>> @@ -1767,6 +1781,20 @@ static int qca808x_config_init(struct phy_device *phydev)
>>   {
>>   	int ret;
>>   
>> +	if (phydev->phy_id == QCA8084_PHY_ID) {
>> +		/* Invert ADC clock edge */
> ...
>> @@ -1958,6 +1986,11 @@ static int qca808x_cable_test_start(struct phy_device *phydev)
>>   	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807a, 0xc060);
>>   	phy_write_mmd(phydev, MDIO_MMD_PCS, 0x807e, 0xb060);
>>   
>> +	if (phydev->phy_id == QCA8084_PHY_ID) {
> 
> Do these need to be exact matches, or should the revision field be
> ignored? If so, consider using phy_id_compare(), or if you end up with
> separate driver entries, consider using phydev_id_compare().
> 
> Thanks.
> 
Ok, i will double check this configuration to consider to use the 
suggested API in the next updated patch. Thanks Russell for the review.