[v1] smsc911x: add second read of EEPROM mac when possible corruption seen

[PATCH v1] smsc911x: add second read of EEPROM mac when possible corruption seen

Posted by Colin Foster 1 month ago

When the EEPROM MAC is read by way of ADDRH, it can return all 0s the
first time. Subsequent reads succeed.

Re-read the ADDRH when this behaviour is observed, in an attempt to
correctly apply the EEPROM MAC address.

Signed-off-by: Colin Foster <colin.foster@in-advantage.com>
---
 drivers/net/ethernet/smsc/smsc911x.c | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/smsc/smsc911x.c b/drivers/net/ethernet/smsc/smsc911x.c
index a2e511912e6a9..63ed221edc00a 100644
--- a/drivers/net/ethernet/smsc/smsc911x.c
+++ b/drivers/net/ethernet/smsc/smsc911x.c
@@ -2162,8 +2162,20 @@ static const struct net_device_ops smsc911x_netdev_ops = {
 static void smsc911x_read_mac_address(struct net_device *dev)
 {
 	struct smsc911x_data *pdata = netdev_priv(dev);
-	u32 mac_high16 = smsc911x_mac_read(pdata, ADDRH);
-	u32 mac_low32 = smsc911x_mac_read(pdata, ADDRL);
+	u32 mac_high16, mac_low32;
+
+	mac_high16 = smsc911x_mac_read(pdata, ADDRH);
+	mac_low32 = smsc911x_mac_read(pdata, ADDRL);
+
+	/*
+	 * The first mac_read always returns 0. Re-read it to get the
+	 * full MAC
+	 */
+	if (mac_high16 == 0) {
+		SMSC_TRACE(pdata, probe, "Re-read MAC ADDRH\n");
+		mac_high16 = smsc911x_mac_read(pdata, ADDRH);
+	}
+
 	u8 addr[ETH_ALEN];
 
 	addr[0] = (u8)(mac_low32);
-- 
2.43.0

Re: [PATCH v1] smsc911x: add second read of EEPROM mac when possible corruption seen

Posted by Jakub Kicinski 1 month ago

On Thu, 28 Aug 2025 16:44:52 -0500 Colin Foster wrote:
> When the EEPROM MAC is read by way of ADDRH, it can return all 0s the
> first time. Subsequent reads succeed.
> 
> Re-read the ADDRH when this behaviour is observed, in an attempt to
> correctly apply the EEPROM MAC address.

Please name the device, and FW version if applicable, on which you
observe the issue.

> Signed-off-by: Colin Foster <colin.foster@in-advantage.com>
> ---
>  drivers/net/ethernet/smsc/smsc911x.c | 16 ++++++++++++++--
>  1 file changed, 14 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/ethernet/smsc/smsc911x.c b/drivers/net/ethernet/smsc/smsc911x.c
> index a2e511912e6a9..63ed221edc00a 100644
> --- a/drivers/net/ethernet/smsc/smsc911x.c
> +++ b/drivers/net/ethernet/smsc/smsc911x.c
> @@ -2162,8 +2162,20 @@ static const struct net_device_ops smsc911x_netdev_ops = {
>  static void smsc911x_read_mac_address(struct net_device *dev)
>  {
>  	struct smsc911x_data *pdata = netdev_priv(dev);
> -	u32 mac_high16 = smsc911x_mac_read(pdata, ADDRH);
> -	u32 mac_low32 = smsc911x_mac_read(pdata, ADDRL);
> +	u32 mac_high16, mac_low32;
> +
> +	mac_high16 = smsc911x_mac_read(pdata, ADDRH);
> +	mac_low32 = smsc911x_mac_read(pdata, ADDRL);
> +
> +	/*

nit: netdev multi-line comment style doesn't place /* on a separate
line:
	

> +	 * The first mac_read always returns 0. Re-read it to get the
> +	 * full MAC

Always? Strange, why did nobody notice until now?

> +	 */
> +	if (mac_high16 == 0) {
> +		SMSC_TRACE(pdata, probe, "Re-read MAC ADDRH\n");
> +		mac_high16 = smsc911x_mac_read(pdata, ADDRH);
> +	}

> 	u8 addr[ETH_ALEN];

Please don't add code in the middle of variable declarations
-- 
pw-bot: cr

Re: [PATCH v1] smsc911x: add second read of EEPROM mac when possible corruption seen

Posted by Colin Foster 1 month ago

Hi Jakub,

On Mon, Sep 01, 2025 at 01:57:12PM -0700, Jakub Kicinski wrote:
> On Thu, 28 Aug 2025 16:44:52 -0500 Colin Foster wrote:
> > When the EEPROM MAC is read by way of ADDRH, it can return all 0s the
> > first time. Subsequent reads succeed.
> > 
> > Re-read the ADDRH when this behaviour is observed, in an attempt to
> > correctly apply the EEPROM MAC address.
> 
> Please name the device, and FW version if applicable, on which you
> observe the issue.

I'll add that to the commit message. FWIW it is the Phytec PCM049 SOM.

> 
> > Signed-off-by: Colin Foster <colin.foster@in-advantage.com>
> > ---
> >  drivers/net/ethernet/smsc/smsc911x.c | 16 ++++++++++++++--
> >  1 file changed, 14 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/net/ethernet/smsc/smsc911x.c b/drivers/net/ethernet/smsc/smsc911x.c
> > index a2e511912e6a9..63ed221edc00a 100644
> > --- a/drivers/net/ethernet/smsc/smsc911x.c
> > +++ b/drivers/net/ethernet/smsc/smsc911x.c
> > @@ -2162,8 +2162,20 @@ static const struct net_device_ops smsc911x_netdev_ops = {
> >  static void smsc911x_read_mac_address(struct net_device *dev)
> >  {
> >  	struct smsc911x_data *pdata = netdev_priv(dev);
> > -	u32 mac_high16 = smsc911x_mac_read(pdata, ADDRH);
> > -	u32 mac_low32 = smsc911x_mac_read(pdata, ADDRL);
> > +	u32 mac_high16, mac_low32;
> > +
> > +	mac_high16 = smsc911x_mac_read(pdata, ADDRH);
> > +	mac_low32 = smsc911x_mac_read(pdata, ADDRL);
> > +
> > +	/*
> 
> nit: netdev multi-line comment style doesn't place /* on a separate
> line:

Apologies - that shouldn't have slipped through.

> 	
> 
> > +	 * The first mac_read always returns 0. Re-read it to get the
> > +	 * full MAC
> 
> Always? Strange, why did nobody notice until now?

For me it is 100% reproduceable. The first read is always 0. I've added
delays in case timing was the issue. I've swapped ADDRH and ADDRL and
the opposite effect happened (where the first four MAC octets were
zero). Re-reads always succeed.

Without the patch, the last two MAC octets are always zero.

We didn't notice it until we started hooking multiple devices on the
same network.

If there is anyone else running this hardware, I'd love verification.
Its an SMSC9221.

That's a long way of saying "I don't know" unfortunately.

> 
> > +	 */
> > +	if (mac_high16 == 0) {
> > +		SMSC_TRACE(pdata, probe, "Re-read MAC ADDRH\n");
> > +		mac_high16 = smsc911x_mac_read(pdata, ADDRH);
> > +	}
> 
> > 	u8 addr[ETH_ALEN];
> 
> Please don't add code in the middle of variable declarations

Ack.

> -- 
> pw-bot: cr

Re: [PATCH v1] smsc911x: add second read of EEPROM mac when possible corruption seen

Posted by Jakub Kicinski 1 month ago

On Tue, 2 Sep 2025 07:31:13 -0500 Colin Foster wrote:
> > > +	 * The first mac_read always returns 0. Re-read it to get the
> > > +	 * full MAC  
> > 
> > Always? Strange, why did nobody notice until now?  
> 
> For me it is 100% reproduceable. The first read is always 0. I've added
> delays in case timing was the issue. I've swapped ADDRH and ADDRL and
> the opposite effect happened (where the first four MAC octets were
> zero). Re-reads always succeed.
> 
> Without the patch, the last two MAC octets are always zero.
> 
> We didn't notice it until we started hooking multiple devices on the
> same network.
> 
> If there is anyone else running this hardware, I'd love verification.
> Its an SMSC9221.
> 
> That's a long way of saying "I don't know" unfortunately.

Right, I think we should avoid saying "always" in the comment then.
Let's weasel word it a little bit given the uncertainty..