[PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Meghana Malladi]

When using DP83869 in force 100Base-Tx mode, the PHY is required
to have robust Auto-MDIX feature enabled from register 1Eh.
Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.

[1]: https://www.ti.com/lit/ds/symlink/dp83869hm.pdf
Signed-off-by: Meghana Malladi <m-malladi@ti.com>
---
 drivers/net/phy/dp83869.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/drivers/net/phy/dp83869.c b/drivers/net/phy/dp83869.c
index 1f381d7b13ff..70d759c45c6a 100644
--- a/drivers/net/phy/dp83869.c
+++ b/drivers/net/phy/dp83869.c
@@ -110,6 +110,7 @@
 
 /* CFG4 bits */
 #define DP83869_INT_OE	BIT(7)
+#define DP83869_CFG4_ENABLE_AUTO_MDIX_100FD	BIT(9)
 
 /* OP MODE */
 #define DP83869_OP_MODE_MII			BIT(5)
@@ -900,6 +901,23 @@ static int dp83869_phy_reset(struct phy_device *phydev)
 	return dp83869_config_init(phydev);
 }
 
+static void dp83869_link_change_notify(struct phy_device *phydev)
+{
+	int cfg4;
+
+	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
+	 * to have robust Auto-MDIX feature enabled
+	 */
+	if (phydev->autoneg == AUTONEG_DISABLE &&
+	    phydev->speed == SPEED_100 &&
+	    phydev->duplex == DUPLEX_FULL) {
+		cfg4 = phy_read(phydev, DP83869_CFG4);
+		if (cfg4 >= 0) {
+			cfg4 |= DP83869_CFG4_ENABLE_AUTO_MDIX_100FD;
+			phy_write(phydev, DP83869_CFG4, cfg4);
+		}
+	}
+}
 
 #define DP83869_PHY_DRIVER(_id, _name)				\
 {								\
@@ -918,6 +936,7 @@ static int dp83869_phy_reset(struct phy_device *phydev)
 	.set_wol	= dp83869_set_wol,			\
 	.suspend	= genphy_suspend,			\
 	.resume		= genphy_resume,			\
+	.link_change_notify = dp83869_link_change_notify,	\
 }
 
 static struct phy_driver dp83869_driver[] = {

base-commit: e9a5073a98d940837cbb95e71eed1f28f48e7b30
-- 
2.43.0

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Russell King (Oracle)]

On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
> When using DP83869 in force 100Base-Tx mode, the PHY is required
> to have robust Auto-MDIX feature enabled from register 1Eh.
> Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
> 
> [1]: https://www.ti.com/lit/ds/symlink/dp83869hm.pdf
> Signed-off-by: Meghana Malladi <m-malladi@ti.com>
> ---
>  drivers/net/phy/dp83869.c | 19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
> 
> diff --git a/drivers/net/phy/dp83869.c b/drivers/net/phy/dp83869.c
> index 1f381d7b13ff..70d759c45c6a 100644
> --- a/drivers/net/phy/dp83869.c
> +++ b/drivers/net/phy/dp83869.c
> @@ -110,6 +110,7 @@
>  
>  /* CFG4 bits */
>  #define DP83869_INT_OE	BIT(7)
> +#define DP83869_CFG4_ENABLE_AUTO_MDIX_100FD	BIT(9)
>  
>  /* OP MODE */
>  #define DP83869_OP_MODE_MII			BIT(5)
> @@ -900,6 +901,23 @@ static int dp83869_phy_reset(struct phy_device *phydev)
>  	return dp83869_config_init(phydev);
>  }
>  
> +static void dp83869_link_change_notify(struct phy_device *phydev)
> +{
> +	int cfg4;
> +
> +	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
> +	 * to have robust Auto-MDIX feature enabled
> +	 */
> +	if (phydev->autoneg == AUTONEG_DISABLE &&
> +	    phydev->speed == SPEED_100 &&
> +	    phydev->duplex == DUPLEX_FULL) {
> +		cfg4 = phy_read(phydev, DP83869_CFG4);
> +		if (cfg4 >= 0) {
> +			cfg4 |= DP83869_CFG4_ENABLE_AUTO_MDIX_100FD;
> +			phy_write(phydev, DP83869_CFG4, cfg4);
> +		}
> +	}

Shouldn't this kind of configuration be done via .config_aneg() ?

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Meghana Malladi]

Hi Russell,

On 2/11/26 18:56, Russell King (Oracle) wrote:
> On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
>> When using DP83869 in force 100Base-Tx mode, the PHY is required
>> to have robust Auto-MDIX feature enabled from register 1Eh.
>> Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
>>
>> [1]: https://www.ti.com/lit/ds/symlink/dp83869hm.pdf
>> Signed-off-by: Meghana Malladi <m-malladi@ti.com>
>> ---
>>   drivers/net/phy/dp83869.c | 19 +++++++++++++++++++
>>   1 file changed, 19 insertions(+)
>>
>> diff --git a/drivers/net/phy/dp83869.c b/drivers/net/phy/dp83869.c
>> index 1f381d7b13ff..70d759c45c6a 100644
>> --- a/drivers/net/phy/dp83869.c
>> +++ b/drivers/net/phy/dp83869.c
>> @@ -110,6 +110,7 @@
>>   
>>   /* CFG4 bits */
>>   #define DP83869_INT_OE	BIT(7)
>> +#define DP83869_CFG4_ENABLE_AUTO_MDIX_100FD	BIT(9)
>>   
>>   /* OP MODE */
>>   #define DP83869_OP_MODE_MII			BIT(5)
>> @@ -900,6 +901,23 @@ static int dp83869_phy_reset(struct phy_device *phydev)
>>   	return dp83869_config_init(phydev);
>>   }
>>   
>> +static void dp83869_link_change_notify(struct phy_device *phydev)
>> +{
>> +	int cfg4;
>> +
>> +	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
>> +	 * to have robust Auto-MDIX feature enabled
>> +	 */
>> +	if (phydev->autoneg == AUTONEG_DISABLE &&
>> +	    phydev->speed == SPEED_100 &&
>> +	    phydev->duplex == DUPLEX_FULL) {
>> +		cfg4 = phy_read(phydev, DP83869_CFG4);
>> +		if (cfg4 >= 0) {
>> +			cfg4 |= DP83869_CFG4_ENABLE_AUTO_MDIX_100FD;
>> +			phy_write(phydev, DP83869_CFG4, cfg4);
>> +		}
>> +	}
> 
> Shouldn't this kind of configuration be done via .config_aneg() ?
> 

AFAIK, in case of fixed mode, auto-neg might not be triggered all the 
time. In the phy_ethtool_ksettings_set() function, auto-neg is initiated 
only if the PHY hasn't started. PHY drivers are new to me, please do 
correct me if I am wrong.

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Andrew Lunn]

On Wed, Feb 11, 2026 at 07:23:12PM +0530, Meghana Malladi wrote:
> Hi Russell,
> 
> On 2/11/26 18:56, Russell King (Oracle) wrote:
> > On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
> > > When using DP83869 in force 100Base-Tx mode, the PHY is required
> > > to have robust Auto-MDIX feature enabled from register 1Eh.
> > > Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
> > > 
> > > [1]: https://www.ti.com/lit/ds/symlink/dp83869hm.pdf
> > > Signed-off-by: Meghana Malladi <m-malladi@ti.com>
> > > ---
> > >   drivers/net/phy/dp83869.c | 19 +++++++++++++++++++
> > >   1 file changed, 19 insertions(+)
> > > 
> > > diff --git a/drivers/net/phy/dp83869.c b/drivers/net/phy/dp83869.c
> > > index 1f381d7b13ff..70d759c45c6a 100644
> > > --- a/drivers/net/phy/dp83869.c
> > > +++ b/drivers/net/phy/dp83869.c
> > > @@ -110,6 +110,7 @@
> > >   /* CFG4 bits */
> > >   #define DP83869_INT_OE	BIT(7)
> > > +#define DP83869_CFG4_ENABLE_AUTO_MDIX_100FD	BIT(9)
> > >   /* OP MODE */
> > >   #define DP83869_OP_MODE_MII			BIT(5)
> > > @@ -900,6 +901,23 @@ static int dp83869_phy_reset(struct phy_device *phydev)
> > >   	return dp83869_config_init(phydev);
> > >   }
> > > +static void dp83869_link_change_notify(struct phy_device *phydev)
> > > +{
> > > +	int cfg4;
> > > +
> > > +	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
> > > +	 * to have robust Auto-MDIX feature enabled
> > > +	 */
> > > +	if (phydev->autoneg == AUTONEG_DISABLE &&
> > > +	    phydev->speed == SPEED_100 &&
> > > +	    phydev->duplex == DUPLEX_FULL) {
> > > +		cfg4 = phy_read(phydev, DP83869_CFG4);
> > > +		if (cfg4 >= 0) {
> > > +			cfg4 |= DP83869_CFG4_ENABLE_AUTO_MDIX_100FD;
> > > +			phy_write(phydev, DP83869_CFG4, cfg4);
> > > +		}
> > > +	}
> > 
> > Shouldn't this kind of configuration be done via .config_aneg() ?
> > 
> 
> AFAIK, in case of fixed mode, auto-neg might not be triggered all the time.
> In the phy_ethtool_ksettings_set() function, auto-neg is initiated only if
> the PHY hasn't started. PHY drivers are new to me, please do correct me if I
> am wrong.

.config_aneg() is badly named, since it is also used for forced mode. Just check
phydev->autoneg == AUTONEG_DISABLE.

	Andrew

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Russell King (Oracle)]

On Wed, Feb 11, 2026 at 07:23:12PM +0530, Meghana Malladi wrote:
> Hi Russell,
> 
> On 2/11/26 18:56, Russell King (Oracle) wrote:
> > On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
> > > When using DP83869 in force 100Base-Tx mode, the PHY is required
> > > to have robust Auto-MDIX feature enabled from register 1Eh.
> > > Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
> > > 
> > > [1]: https://www.ti.com/lit/ds/symlink/dp83869hm.pdf
> > > Signed-off-by: Meghana Malladi <m-malladi@ti.com>
> > > ---
> > >   drivers/net/phy/dp83869.c | 19 +++++++++++++++++++
> > >   1 file changed, 19 insertions(+)
> > > 
> > > diff --git a/drivers/net/phy/dp83869.c b/drivers/net/phy/dp83869.c
> > > index 1f381d7b13ff..70d759c45c6a 100644
> > > --- a/drivers/net/phy/dp83869.c
> > > +++ b/drivers/net/phy/dp83869.c
> > > @@ -110,6 +110,7 @@
> > >   /* CFG4 bits */
> > >   #define DP83869_INT_OE	BIT(7)
> > > +#define DP83869_CFG4_ENABLE_AUTO_MDIX_100FD	BIT(9)
> > >   /* OP MODE */
> > >   #define DP83869_OP_MODE_MII			BIT(5)
> > > @@ -900,6 +901,23 @@ static int dp83869_phy_reset(struct phy_device *phydev)
> > >   	return dp83869_config_init(phydev);
> > >   }
> > > +static void dp83869_link_change_notify(struct phy_device *phydev)
> > > +{
> > > +	int cfg4;
> > > +
> > > +	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
> > > +	 * to have robust Auto-MDIX feature enabled
> > > +	 */
> > > +	if (phydev->autoneg == AUTONEG_DISABLE &&
> > > +	    phydev->speed == SPEED_100 &&
> > > +	    phydev->duplex == DUPLEX_FULL) {
> > > +		cfg4 = phy_read(phydev, DP83869_CFG4);
> > > +		if (cfg4 >= 0) {
> > > +			cfg4 |= DP83869_CFG4_ENABLE_AUTO_MDIX_100FD;
> > > +			phy_write(phydev, DP83869_CFG4, cfg4);
> > > +		}
> > > +	}
> > 
> > Shouldn't this kind of configuration be done via .config_aneg() ?
> > 
> 
> AFAIK, in case of fixed mode, auto-neg might not be triggered all the time.
> In the phy_ethtool_ksettings_set() function, auto-neg is initiated only if
> the PHY hasn't started. PHY drivers are new to me, please do correct me if I
> am wrong.

In phy_ethtool_ksettings_set():

        /* Restart the PHY */
        if (phy_is_started(phydev)) {
                phydev->state = PHY_UP;
                phy_trigger_machine(phydev);
        } else {
                _phy_start_aneg(phydev);
        }

If the PHY is started, then the state machine is triggered. That
will result in phy_state_machine() being called.

This calls _phy_state_machine(), which, as phydev->state is set to
PHY_UP, will set state_work = PHY_STATE_WORK_ANEG. This will then
result in:

        if (state_work == PHY_STATE_WORK_ANEG) {
                err = _phy_start_aneg(phydev);
                func = &_phy_start_aneg;
        }

which also results in _phy_start_aneg() being called, and thus the
.config_aneg() method.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Meghana Malladi]

Hi Andrew,

On 2/11/26 19:34, Andrew Lunn wrote:
> On Wed, Feb 11, 2026 at 07:23:12PM +0530, Meghana Malladi wrote:
>> Hi Russell,
>>
>> On 2/11/26 18:56, Russell King (Oracle) wrote:
>>> On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
>>>> When using DP83869 in force 100Base-Tx mode, the PHY is required
>>>> to have robust Auto-MDIX feature enabled from register 1Eh.
>>>> Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
>>>>
>>>> [1]: https://www.ti.com/lit/ds/symlink/dp83869hm.pdf
>>>> Signed-off-by: Meghana Malladi <m-malladi@ti.com>
>>>> ---
>>>>    drivers/net/phy/dp83869.c | 19 +++++++++++++++++++
>>>>    1 file changed, 19 insertions(+)
>>>>
>>>> diff --git a/drivers/net/phy/dp83869.c b/drivers/net/phy/dp83869.c
>>>> index 1f381d7b13ff..70d759c45c6a 100644
>>>> --- a/drivers/net/phy/dp83869.c
>>>> +++ b/drivers/net/phy/dp83869.c
>>>> @@ -110,6 +110,7 @@
>>>>    /* CFG4 bits */
>>>>    #define DP83869_INT_OE	BIT(7)
>>>> +#define DP83869_CFG4_ENABLE_AUTO_MDIX_100FD	BIT(9)
>>>>    /* OP MODE */
>>>>    #define DP83869_OP_MODE_MII			BIT(5)
>>>> @@ -900,6 +901,23 @@ static int dp83869_phy_reset(struct phy_device *phydev)
>>>>    	return dp83869_config_init(phydev);
>>>>    }
>>>> +static void dp83869_link_change_notify(struct phy_device *phydev)
>>>> +{
>>>> +	int cfg4;
>>>> +
>>>> +	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
>>>> +	 * to have robust Auto-MDIX feature enabled
>>>> +	 */
>>>> +	if (phydev->autoneg == AUTONEG_DISABLE &&
>>>> +	    phydev->speed == SPEED_100 &&
>>>> +	    phydev->duplex == DUPLEX_FULL) {
>>>> +		cfg4 = phy_read(phydev, DP83869_CFG4);
>>>> +		if (cfg4 >= 0) {
>>>> +			cfg4 |= DP83869_CFG4_ENABLE_AUTO_MDIX_100FD;
>>>> +			phy_write(phydev, DP83869_CFG4, cfg4);
>>>> +		}
>>>> +	}
>>>
>>> Shouldn't this kind of configuration be done via .config_aneg() ?
>>>
>>
>> AFAIK, in case of fixed mode, auto-neg might not be triggered all the time.
>> In the phy_ethtool_ksettings_set() function, auto-neg is initiated only if
>> the PHY hasn't started. PHY drivers are new to me, please do correct me if I
>> am wrong.
> 
> .config_aneg() is badly named, since it is also used for forced mode. Just check
> phydev->autoneg == AUTONEG_DISABLE.

Ok thanks, will fix this in v2 then.

> 
> 	Andrew

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Andrew Lunn]

On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
> When using DP83869 in force 100Base-Tx mode, the PHY is required
> to have robust Auto-MDIX feature enabled from register 1Eh.
> Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.

What are the results of not enabling this? How would somebody know
they need this patch?

Is this a bug fix? Should this be back ported to stable?

> +static void dp83869_link_change_notify(struct phy_device *phydev)
> +{
> +	int cfg4;
> +
> +	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
> +	 * to have robust Auto-MDIX feature enabled
> +	 */
> +	if (phydev->autoneg == AUTONEG_DISABLE &&
> +	    phydev->speed == SPEED_100 &&
> +	    phydev->duplex == DUPLEX_FULL) {

So forced 100Mbs half duplex does not require robust robust Auto-MDIX?

What about when you change the configuration out of forced 100 Full?
Shouldn't the configuration be returned to how the user wants
auto-MDIX configured? Should you disable robust Auto-MDIX so that
"unreliable Auto-MDIX" is used?

	    Andrew

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Meghana Malladi]

Hi Andrew,

On 2/11/26 19:42, Andrew Lunn wrote:
> On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
>> When using DP83869 in force 100Base-Tx mode, the PHY is required
>> to have robust Auto-MDIX feature enabled from register 1Eh.
>> Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
> 
> What are the results of not enabling this? How would somebody know
> they need this patch?

This is the fix for 100M full duplex when auto-neg gets disabled, it 
fails to have a link up and this patch fixes it. I will add more 
description to the commit message.

> 
> Is this a bug fix? Should this be back ported to stable?
> 
>> +static void dp83869_link_change_notify(struct phy_device *phydev)
>> +{
>> +	int cfg4;
>> +
>> +	/* When using DP83869 in force 100Base-Tx mode, the PHY is required
>> +	 * to have robust Auto-MDIX feature enabled
>> +	 */
>> +	if (phydev->autoneg == AUTONEG_DISABLE &&
>> +	    phydev->speed == SPEED_100 &&
>> +	    phydev->duplex == DUPLEX_FULL) {
> 
> So forced 100Mbs half duplex does not require robust robust Auto-MDIX?

I haven't tested for half duplex as the issue was raised for full 
duplex. But based on the TRM seems like it is independent of the 
duplexity. Let me locally confirm this and fix it accordingly in v2.

> 
> What about when you change the configuration out of forced 100 Full?
> Shouldn't the configuration be returned to how the user wants
> auto-MDIX configured? Should you disable robust Auto-MDIX so that
> "unreliable Auto-MDIX" is used?

This bit is only necessary if link partners are configured to 
operational modes that not supported by normal Auto MDI/MDIX mode (when 
one of that link partners are in Forced 100Base-TX), will fix it in v2.

> 
> 	    Andrew
> 

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Andrew Lunn]

> > What about when you change the configuration out of forced 100 Full?
> > Shouldn't the configuration be returned to how the user wants
> > auto-MDIX configured? Should you disable robust Auto-MDIX so that
> > "unreliable Auto-MDIX" is used?
> 
> This bit is only necessary if link partners are configured to operational
> modes that not supported by normal Auto MDI/MDIX mode (when one of that link
> partners are in Forced 100Base-TX), will fix it in v2.

Does this bit have any effect on normal Auto MDI/MDIX mode? Can this
bit be set once during probe, rather than for a specific link mode?

Is there an errata for this behaviour/fix? If there is please
reference the number/document etc.

Since this is a fix, please add a Fixes: tag.

    Andrew

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Maxime Chevallier]

Hi Meghana,

On 12/02/2026 07:21, Meghana Malladi wrote:
> Hi Andrew,
> 
> On 2/11/26 19:42, Andrew Lunn wrote:
>> On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
>>> When using DP83869 in force 100Base-Tx mode, the PHY is required
>>> to have robust Auto-MDIX feature enabled from register 1Eh.
>>> Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
>>
>> What are the results of not enabling this? How would somebody know
>> they need this patch?
> 
> This is the fix for 100M full duplex when auto-neg gets disabled, it 
> fails to have a link up and this patch fixes it. I will add more 
> description to the commit message.

I have a board here with a DP83869 + imx93's dwmac, but I couldn't
trigger the bug locally. I forced speed, dpx and aneg settings with :

ethtool -s eth0 speed 100 duplex full autoneg off

I did the same on the link partner, but I was still able to get link-up.

I'm not saying the bug isn't real, I mostly curious about the kind of
circumstances that causes this prblem to show-up. Do you have any
clarifications on that ?

I'll be happy to give the fix a try, if I can manage to reproduce the
issue :)

Maxime

Re: [EXTERNAL] Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Malladi, Meghana]

Hi Maxime,

On 2/13/2026 12:10 AM, Maxime Chevallier wrote:
> Hi Meghana, On 12/02/2026 07: 21, Meghana Malladi wrote: > Hi Andrew, > 
>  > On 2/11/26 19: 42, Andrew Lunn wrote: >> On Wed, Feb 11, 2026 at 06:  
> 40: 32PM +0530, Meghana Malladi wrote: >>> When using DP83869 in force 
> 100Base-Tx
> ZjQcmQRYFpfptBannerStart
> This message was sent from outside of Texas Instruments.
> Do not click links or open attachments unless you recognize the source 
> of this email and know the content is safe.
> Report Suspicious
> <https://us-phishalarm-ewt.proofpoint.com/EWT/v1/G3vK! 
> uvdqvDfOXMcx4EXE_POD_-6kkowIhAz8zXLHMFm81_iVEIIhL4AZYTCCg5LBLfOFsQIbaQidTlj5ZWaMhdH1sacfSO1-r-L7$>
> ZjQcmQRYFpfptBannerEnd
> 
> Hi Meghana,
> 
> On 12/02/2026 07:21, Meghana Malladi wrote:
>> Hi Andrew,
>> 
>> On 2/11/26 19:42, Andrew Lunn wrote:
>>> On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
>>>> When using DP83869 in force 100Base-Tx mode, the PHY is required
>>>> to have robust Auto-MDIX feature enabled from register 1Eh.
>>>> Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
>>>
>>> What are the results of not enabling this? How would somebody know
>>> they need this patch?
>> 
>> This is the fix for 100M full duplex when auto-neg gets disabled, it 
>> fails to have a link up and this patch fixes it. I will add more 
>> description to the commit message.
> 
> I have a board here with a DP83869 + imx93's dwmac, but I couldn't
> trigger the bug locally. I forced speed, dpx and aneg settings with :
> 
> ethtool -s eth0 speed 100 duplex full autoneg off
> 
> I did the same on the link partner, but I was still able to get link-up.
> 

Your steps so far are correct. But I didn't do the same settings on the 
link partner. With other speeds (10M, 1G) I was seeing a link up without 
the need to force speed on the link partner side. So I was expecting the 
same with 100M as well.

> I'm not saying the bug isn't real, I mostly curious about the kind of
> circumstances that causes this prblem to show-up. Do you have any
> clarifications on that ?
> 

Also when I tried the same between DP83869 PHYs of two AM64x (TI DUTs) I 
couldn't see this issue. When I tried between DP83869 PHY and an 
external switch or PC I was able to reproduce this issue.

May be I missing lot of detail in this patch. Let me know if this helped 
you with reproducing the issue. Thanks a lot :)

> I'll be happy to give the fix a try, if I can manage to reproduce the
> issue :)
> 
> Maxime
> 
> 

-- 
Thanks,
Meghana Malladi

Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Malladi, Meghana]

Hi Andrew,

On 2/12/2026 6:36 PM, Andrew Lunn wrote:
>>> What about when you change the configuration out of forced 100 Full?
>>> Shouldn't the configuration be returned to how the user wants
>>> auto-MDIX configured? Should you disable robust Auto-MDIX so that
>>> "unreliable Auto-MDIX" is used?
>>
>> This bit is only necessary if link partners are configured to operational
>> modes that not supported by normal Auto MDI/MDIX mode (when one of that link
>> partners are in Forced 100Base-TX), will fix it in v2.
> 
> Does this bit have any effect on normal Auto MDI/MDIX mode? Can this
> bit be set once during probe, rather than for a specific link mode?
> 
> Is there an errata for this behaviour/fix? If there is please
> reference the number/document etc.
> 

I checked with the PHY team and they pointed me to this section of the 
DP83867 datasheet[1]: 7.6.25 Configuration Register 3 (CFG3) and said 
something similar is needed for DP83869 as well. There are a lot of 
similar corner cases which they haven't tested and once done, more 
details w.r.t to the possible errata will be published in the next 
datasheet revision (aimed in the next quarter). I guess it makes more 
sense to post this fix later with the published errata.

[1]: https://www.ti.com/lit/ds/symlink/dp83867ir.pdf

> Since this is a fix, please add a Fixes: tag.
> 
>      Andrew

-- 
Thanks,
Meghana Malladi

Re: [EXTERNAL] Re: [PATCH net] net: phy: dp83869: Set auto mdix bit for forced 100Base-Tx mode

[Andrew Lunn]

On Sun, Feb 15, 2026 at 08:15:23PM +0530, Malladi, Meghana wrote:
> Hi Maxime,
> 
> On 2/13/2026 12:10 AM, Maxime Chevallier wrote:
> > Hi Meghana, On 12/02/2026 07: 21, Meghana Malladi wrote: > Hi Andrew, >
> > > On 2/11/26 19: 42, Andrew Lunn wrote: >> On Wed, Feb 11, 2026 at 06: 
> > 40: 32PM +0530, Meghana Malladi wrote: >>> When using DP83869 in force
> > 100Base-Tx
> > ZjQcmQRYFpfptBannerStart
> > This message was sent from outside of Texas Instruments.
> > Do not click links or open attachments unless you recognize the source
> > of this email and know the content is safe.
> > Report Suspicious
> > <https://us-phishalarm-ewt.proofpoint.com/EWT/v1/G3vK! uvdqvDfOXMcx4EXE_POD_-6kkowIhAz8zXLHMFm81_iVEIIhL4AZYTCCg5LBLfOFsQIbaQidTlj5ZWaMhdH1sacfSO1-r-L7$>
> > ZjQcmQRYFpfptBannerEnd
> > 
> > Hi Meghana,
> > 
> > On 12/02/2026 07:21, Meghana Malladi wrote:
> > > Hi Andrew,
> > > 
> > > On 2/11/26 19:42, Andrew Lunn wrote:
> > > > On Wed, Feb 11, 2026 at 06:40:32PM +0530, Meghana Malladi wrote:
> > > > > When using DP83869 in force 100Base-Tx mode, the PHY is required
> > > > > to have robust Auto-MDIX feature enabled from register 1Eh.
> > > > > Refer to 7.4.1.2 100BASE-TX section in the TRM [1] for more details.
> > > > 
> > > > What are the results of not enabling this? How would somebody know
> > > > they need this patch?
> > > 
> > > This is the fix for 100M full duplex when auto-neg gets disabled, it
> > > fails to have a link up and this patch fixes it. I will add more
> > > description to the commit message.
> > 
> > I have a board here with a DP83869 + imx93's dwmac, but I couldn't
> > trigger the bug locally. I forced speed, dpx and aneg settings with :
> > 
> > ethtool -s eth0 speed 100 duplex full autoneg off
> > 
> > I did the same on the link partner, but I was still able to get link-up.
> > 
> 
> Your steps so far are correct. But I didn't do the same settings on the link
> partner. With other speeds (10M, 1G) I was seeing a link up without the need
> to force speed on the link partner side. So I was expecting the same with
> 100M as well.

This is not correct. If you turn autoneg off on one link partner, the
other link partner also must be forced, because it should not be able
to perform autoneg. Some PHYs will default to 10/Half when autoneg
fails, but as far as i'm known, that it not part of 802.3.

> > I'm not saying the bug isn't real, I mostly curious about the kind of
> > circumstances that causes this prblem to show-up. Do you have any
> > clarifications on that ?

Given the testing you have done, it seems to need an invalid
configuration?

> Also when I tried the same between DP83869 PHYs of two AM64x (TI DUTs) I
> couldn't see this issue. When I tried between DP83869 PHY and an external
> switch or PC I was able to reproduce this issue.

Did you force the switch to 100Full? Or was autoneg still enabled?

	Andrew