net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
Subsystems (2)
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
As per Linux's driver, ID_MODE_DIS is only set when the PHY interface is RGMII. Don't enable it for the rest of setups.
This has been seen to misconfigure RPi4's PHY when booting Linux.
--- drivers/net/bcmgenet.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/net/bcmgenet.c +4 -1
@@ -448,7 +448,10 @@ static int bcmgenet_adjust_link(struct bcmgenet_eth_priv *priv) }clrsetbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,OOB_DISABLE , -RGMII_LINK |RGMII_MODE_EN |ID_MODE_DIS ); +RGMII_LINK |RGMII_MODE_EN ); + +if (phy_dev->interface ==PHY_INTERFACE_MODE_RGMII ) +setbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,ID_MODE_DIS );writel (speed <<CMD_SPEED_SHIFT , (priv->mac_reg +UMAC_CMD ));
Signature · 1 line
2.25.0
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
Quoted 6 lines · Nicolas Saenz Julienne
As per Linux's driver,
ID_MODE_DISis only set when the PHY interface is RGMII. Don't enable it for the rest of setups.This has been seen to misconfigure RPi4's PHY when booting Linux.
Signed-off-by Nicolas Saenz Julienne <[email protected]>
I forgot to add:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200220/9d3f1075/attachment-0001.sig>
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
Quoted 4 lines · Nicolas Saenz Julienne
Quoted 3 lines · Nicolas Saenz Julienne
As per Linux's driver,
ID_MODE_DISis only set when the PHY interface is RGMII. Don't enable it for the rest of setups.
Quoted 8 lines
Quoted 3 lines
This has been seen to misconfigure RPi4's PHY when booting Linux.
Signed-off-by Nicolas Saenz Julienne <[email protected]>I forgot to add:
Fixes d53e3fa385 ("net: Add support for Broadcom GENETv5 Ethernetcontroller")
Ping :)
Regards,
Nicolas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200311/81793e0e/attachment.sig>
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
In reply to Nicolas Saenz JulienneQuoted 15 lines · Nicolas Saenz Julienne
As per Linux's driver,
ID_MODE_DISis only set when the PHY interface is RGMII. Don't enable it for the rest of setups.This has been seen to misconfigure RPi4's PHY when booting Linux.
Signed-off-by Nicolas Saenz Julienne <[email protected]>--- drivers/net/bcmgenet.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/net/bcmgenet.c +0
-0
drivers/net/bcmgenet.c static int bcmgenet_adjust_link(struct bcmgenet_eth_priv *priv) · line 454
Quoted 8 lines
drivers/net/bcmgenet.c +4
-1@@ -448,7 +448,10 @@ static int bcmgenet_adjust_link(struct bcmgenet_eth_priv *priv) }clrsetbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,OOB_DISABLE , -RGMII_LINK |RGMII_MODE_EN |ID_MODE_DIS ); +RGMII_LINK |RGMII_MODE_EN ); + +if (phy_dev->interface ==PHY_INTERFACE_MODE_RGMII ) +setbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,ID_MODE_DIS );
Is this given because by different DTS? Shouldn't that be uniform on the RPi4?
BTW Joe, will you take patches for this driver through your branch? For now I delegated it to me, but I'm fine either way.
Regards,
Matthias
Quoted 4 lines
drivers/net/bcmgenet.c +0
-0writel (speed <<CMD_SPEED_SHIFT , (priv->mac_reg +UMAC_CMD ));
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
Hi Matthias,
Quoted 20 lines · Matthias Brugger
Quoted 18 lines · Nicolas Saenz Julienne
As per Linux's driver,
ID_MODE_DISis only set when the PHY interface is RGMII. Don't enable it for the rest of setups.This has been seen to misconfigure RPi4's PHY when booting Linux.
Signed-off-by Nicolas Saenz Julienne <[email protected]>--- drivers/net/bcmgenet.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/net/bcmgenet.c +0
-0@@ -448,7 +448,10 @@ static int bcmgenet_adjust_link(struct bcmgenet_eth_priv *priv) }
Quoted 8 lines
drivers/net/bcmgenet.c static int bcmgenet_adjust_link(struct bcmgenet_eth_priv *priv) · line 454
Quoted 6 lines
drivers/net/bcmgenet.c +4
-1clrsetbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,OOB_DISABLE , -RGMII_LINK |RGMII_MODE_EN |ID_MODE_DIS ); +RGMII_LINK |RGMII_MODE_EN ); + +if (phy_dev->interface ==PHY_INTERFACE_MODE_RGMII ) +setbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,ID_MODE_DIS );Is this given because by different DTS? Shouldn't that be uniform on the RPi4?
The interface type is read from DT, the 'phy-mode' property. In the case of the RPi4 it's 'rgmii-rxid'.
The downstream DT used to be configured differently ('rgmii' and using 'ethernet-phy-ieee802.3-c22'), that's why you might have seen the board working at some point with this driver. But as we updated the DT to match upstream's we switched to 'rgmii-rxid' which is being misconfigured as 'rgmii' in u-boot. So you have u-boot configuring 'rgmii' while Linux configures 'rgmii-rxid', which fails to clear the ID_MODE_DIS bit. This, I imagine, blocks the delay configuration process from the PHY (I don't have any documentation).
Regards,
Nicolas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200221/f0024b31/attachment.sig>
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
Florian,
Quoted 36 lines · Nicolas Saenz Julienne
Hi Matthias,
Quoted 19 lines · Matthias Brugger
Quoted 18 lines · Nicolas Saenz Julienne
As per Linux's driver,
ID_MODE_DISis only set when the PHY interface is RGMII. Don't enable it for the rest of setups.This has been seen to misconfigure RPi4's PHY when booting Linux.
Signed-off-by Nicolas Saenz Julienne <[email protected]>--- drivers/net/bcmgenet.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/net/bcmgenet.c +0
-0@@ -448,7 +448,10 @@ static int bcmgenet_adjust_link(struct bcmgenet_eth_priv *priv) }Quoted 9 lines
drivers/net/bcmgenet.c static int bcmgenet_adjust_link(struct · line 455
Quoted 6 lines
drivers/net/bcmgenet.c +4
-1clrsetbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,OOB_DISABLE , -RGMII_LINK |RGMII_MODE_EN |ID_MODE_DIS ); +RGMII_LINK |RGMII_MODE_EN ); + +if (phy_dev->interface ==PHY_INTERFACE_MODE_RGMII ) +setbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,ID_MODE_DIS );Is this given because by different DTS? Shouldn't that be uniform on the RPi4?
The interface type is read from DT, the 'phy-mode' property. In the case of the
RPi4 it's 'rgmii-rxid'.
Quoted 9 lines
The downstream DT used to be configured differently ('rgmii' and using 'ethernet-phy-ieee802.3-c22'), that's why you might have seen the board working
at some point with this driver. But as we updated the DT to match upstream's we
switched to 'rgmii-rxid' which is being misconfigured as 'rgmii' in u-boot. So you have u-boot configuring 'rgmii' while Linux configures 'rgmii-rxid', which fails to clear theID_MODE_DISbit. This, I imagine, blocks the delay configuration process from the PHY (I don't have any documentation).
With this in mind, would it make sens to do this in the Linux driver?
(quoted diff) +0 -0
b/drivers/net/ethernet/broadcom/genet/bcmmii.c
index 6392a2530183..10244941a7a6 100644
drivers/net/ethernet/broadcom/genet/bcmmii.c +0 -0
@@ -294,6 +294,7 @@ int bcmgenet_mii_config(struct net_device *dev, bool init)
*/
if (priv->ext_phy) {
reg = bcmgenet_ext_readl(priv, EXT_RGMII_OOB_CTRL);
+ reg &= ~ID_MODE_DIS;
reg |= id_mode_dis;
if (GENET_IS_V1(priv) || GENET_IS_V2(priv) || GENET_IS_V3(priv))
reg |= RGMII_MODE_EN_V123;It all comes down to wheter we trust bootloader's config or not.
Regards,
Nicolas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200221/b3b0b39d/attachment.sig>
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
In reply to Nicolas Saenz JulienneQuoted 6 lines · Nicolas Saenz Julienne
As per Linux's driver,
ID_MODE_DISis only set when the PHY interface is RGMII. Don't enable it for the rest of setups.This has been seen to misconfigure RPi4's PHY when booting Linux.
Signed-off-by Nicolas Saenz Julienne <[email protected]>
Does the failure look like the following: you have a driver for the Broadcom PHY used on the Pi4 in u-boot, and the phy_dev->interface value is being used to configure the Ethernet PHY chip in a certain way.
Later when you boot Linux, you do not have CONFIG_BROADCOM_PHY enabled so the Generic PHY driver gets used instead, and there is a disagreement between the AMAC and PHY as to whom should be adding the delays?
At any rate:
Quoted 21 lines
--- drivers/net/bcmgenet.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/net/bcmgenet.c +4
-1@@ -448,7 +448,10 @@ static int bcmgenet_adjust_link(struct bcmgenet_eth_priv *priv) }clrsetbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,OOB_DISABLE , -RGMII_LINK |RGMII_MODE_EN |ID_MODE_DIS ); +RGMII_LINK |RGMII_MODE_EN ); + +if (phy_dev->interface ==PHY_INTERFACE_MODE_RGMII ) +setbits_32 (priv->mac_reg +EXT_RGMII_OOB_CTRL ,ID_MODE_DIS );writel (speed <<CMD_SPEED_SHIFT , (priv->mac_reg +UMAC_CMD ));
Signature · 1 line
Florian
PATCH 1/1 net: bcmgenet: Don't set ID_MODE_DIS when not using RGMII
Hi Florian,
Quoted 8 lines · Florian Fainelli
Quoted 6 lines · Nicolas Saenz Julienne
As per Linux's driver,
ID_MODE_DISis only set when the PHY interface is RGMII. Don't enable it for the rest of setups.This has been seen to misconfigure RPi4's PHY when booting Linux.
Signed-off-by Nicolas Saenz Julienne <[email protected]>
Quoted 7 lines
Does the failure look like the following: you have a driver for the Broadcom PHY used on the Pi4 in u-boot, and the
phy_dev->interfacevalue is being used to configure the Ethernet PHY chip in a certain way.Later when you boot Linux, you do not have
CONFIG_BROADCOM_PHYenabled so the Generic PHY driver gets used instead, and there is a disagreement between the AMAC and PHY as to whom should be adding the delays?
I added an explanation to Matthias' response. I think it fits yours, modulo my limited knowledge in the area :)
Quoted 3 lines
At any rate:
Reviewed-by Florian Fainelli <[email protected]>
Thanks!
Regards,
Nicolas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200221/8431dedf/attachment.sig>