[PATCH] net: sky2: allow mac to come from dt

The driver reads the mac address from the device registers which would
need to have been programmed by the bootloader.  This patch adds
the ability to pull the mac from devicetree via the pci device dt node.

Signed-off-by: Tim Harvey <tharvey@gateworks.com>
Cc: netdev@vger.kernel.org
Cc: devicetree@vger.kernel.org
Cc: Grant Likely <redacted>
Cc: Rob Herring <robh+dt@kernel.org>
---
 drivers/net/ethernet/marvell/sky2.c | 24 +++++++++++++++++++++++-
 1 file changed, 23 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/marvell/sky2.c b/drivers/net/ethernet/marvell/sky2.c
index 55a37ae..bff493a 100644
--- a/drivers/net/ethernet/marvell/sky2.c
+++ b/drivers/net/ethernet/marvell/sky2.c

@@ -44,6 +44,8 @@
 #include <linux/prefetch.h>
 #include <linux/debugfs.h>
 #include <linux/mii.h>
+#include <linux/of_device.h>
+#include <linux/of_net.h>
 
 #include <asm/irq.h>
 

@@ -4748,6 +4750,7 @@ static struct net_device *sky2_init_netdev(struct sky2_hw *hw, unsigned port,
 {
 	struct sky2_port *sky2;
 	struct net_device *dev = alloc_etherdev(sizeof(*sky2));
+	unsigned char *iap, tmpaddr[ETH_ALEN];
 
 	if (!dev)
 		return NULL;

@@ -4805,8 +4808,27 @@ static struct net_device *sky2_init_netdev(struct sky2_hw *hw, unsigned port,
 
 	dev->features |= dev->hw_features;
 
+	/* try to get mac address in the following order:
+	 * 1) from device tree data
+	 * 2) from internal registers set by bootloader
+	 */
+	iap = NULL;
+	if (IS_ENABLED(CONFIG_OF)) {
+		struct device_node *np = hw->pdev->dev.of_node;
+		if (np)
+			iap = (unsigned char *) of_get_mac_address(np);
+	}
+
+	/* 2) mac registers set by bootloader
+	 */
+	if (!iap || !is_valid_ether_addr(iap)) {
+		memcpy_fromio(&tmpaddr, hw->regs + B2_MAC_1 + port * 8,
+			ETH_ALEN);
+		iap = &tmpaddr[0];
+	}
+
 	/* read the mac address */
-	memcpy_fromio(dev->dev_addr, hw->regs + B2_MAC_1 + port * 8, ETH_ALEN);
+	ether_addr_copy(dev->dev_addr, iap);
 
 	return dev;
 }

-- 
1.8.3.2

On Wed, Mar 5, 2014 at 12:22 AM, Tim Harvey [off-list ref] wrote:

This shouldn't be needed.

+               struct device_node *np = hw->pdev->dev.of_node;
+               if (np)
+                       iap = (unsigned char *) of_get_mac_address(np);
+       }

This should all just be 1 line. of_get_mac_address should just return
NULL if !CONFIG_OF or np == NULL.

+
+       /* 2) mac registers set by bootloader
+        */
+       if (!iap || !is_valid_ether_addr(iap)) {

Perhaps is_valid_ether_addr check should also be moved into of_get_mac_address.

Rob

Hello.

On 05-03-2014 10:22, Tim Harvey wrote:

The driver reads the mac address from the device registers which would
need to have been programmed by the bootloader.  This patch adds
the ability to pull the mac from devicetree via the pci device dt node.

    I highly doubt that "[local-]mac-address" prop would be added to 
(autodiscovered) PCI device node.

Signed-off-by: Tim Harvey <tharvey@gateworks.com>
Cc: netdev@vger.kernel.org
Cc: devicetree@vger.kernel.org
Cc: Grant Likely <redacted>
Cc: Rob Herring <robh+dt@kernel.org>
---
  drivers/net/ethernet/marvell/sky2.c | 24 +++++++++++++++++++++++-
  1 file changed, 23 insertions(+), 1 deletion(-)

[...]

    This line should start right under & on the previous line.

+		iap = &tmpaddr[0];

    Why not just 'tmpaddr'?

+	}
+

WBR, Sergei

On Tue, Mar 4, 2014 at 10:32 PM, Rob Herring [off-list ref] wrote:
<snip>

quoted

@@ -4805,8 +4808,27 @@ static struct net_device *sky2_init_netdev(struct sky2_hw *hw, unsigned port,

        dev->features |= dev->hw_features;

+       /* try to get mac address in the following order:
+        * 1) from device tree data
+        * 2) from internal registers set by bootloader
+        */
+       iap = NULL;
+       if (IS_ENABLED(CONFIG_OF)) {

This shouldn't be needed.

quoted

+               struct device_node *np = hw->pdev->dev.of_node;
+               if (np)
+                       iap = (unsigned char *) of_get_mac_address(np);
+       }

This should all just be 1 line. of_get_mac_address should just return
NULL if !CONFIG_OF or np == NULL.

agreed - I verified this is true

quoted

+
+       /* 2) mac registers set by bootloader
+        */
+       if (!iap || !is_valid_ether_addr(iap)) {

Perhaps is_valid_ether_addr check should also be moved into of_get_mac_address.

looks like it already does that - I'll simplify the above

Rob

Thanks for the review!

Tim

On Wed, Mar 5, 2014 at 6:15 AM, Sergei Shtylyov
[off-list ref] wrote:

Hello.

Hi Sergei,

On 05-03-2014 10:22, Tim Harvey wrote:

quoted

The driver reads the mac address from the device registers which would
need to have been programmed by the bootloader.  This patch adds
the ability to pull the mac from devicetree via the pci device dt node.

   I highly doubt that "[local-]mac-address" prop would be added to
(autodiscovered) PCI device node.

I wouldn't have done this if I didn't need it ;)

The u-boot bootloader will do this for boards supporting devicetree.
Specifically, it will take the ethaddr/eth<n<addr/etc env vars and
apply a local-mac-address property to the devicetree using ethernet<n>
devicetree aliases.  This is to support MAC addresses coming from
EEPROM, eFUSE, etc.  The key here is that it doesn't need to know
where in the tree the eth devices are because it makes use of the
aliases node.

This of course requires that you do have an ethernet<n> alias to your
PCI based network device in your dt which I agree may seem strange for
an auto-probed bus, but in the case of a bootloader without PCI
support and/or sky2 support this seems the proper way to get the MAC
address from the bootloader to the driver.

[...]

quoted

@@ -4805,8 +4808,27 @@ static struct net_device *sky2_init_netdev(struct

sky2_hw *hw, unsigned port,

        dev->features |= dev->hw_features;

+       /* try to get mac address in the following order:
+        * 1) from device tree data
+        * 2) from internal registers set by bootloader
+        */
+       iap = NULL;
+       if (IS_ENABLED(CONFIG_OF)) {
+               struct device_node *np = hw->pdev->dev.of_node;
+               if (np)
+                       iap = (unsigned char *) of_get_mac_address(np);
+       }
+
+       /* 2) mac registers set by bootloader
+        */
+       if (!iap || !is_valid_ether_addr(iap)) {
+               memcpy_fromio(&tmpaddr, hw->regs + B2_MAC_1 + port * 8,
+                       ETH_ALEN);

   This line should start right under & on the previous line.

agreed - not sure why checkpatch.pl didn't catch that

quoted

+               iap = &tmpaddr[0];

agreed - I will change this as its more readable

   Why not just 'tmpaddr'?

quoted

+       }
+

WBR, Sergei

Thanks for the review!

Tim

Hello.

On 03/05/2014 06:38 PM, Tim Harvey wrote:

quoted

quoted

The driver reads the mac address from the device registers which would
need to have been programmed by the bootloader.  This patch adds
the ability to pull the mac from devicetree via the pci device dt node.

quoted

    I highly doubt that "[local-]mac-address" prop would be added to
(autodiscovered) PCI device node.

I wouldn't have done this if I didn't need it ;)

The u-boot bootloader will do this for boards supporting devicetree.
Specifically, it will take the ethaddr/eth<n<addr/etc env vars and
apply a local-mac-address property to the devicetree using ethernet<n>
devicetree aliases.  This is to support MAC addresses coming from
EEPROM, eFUSE, etc.  The key here is that it doesn't need to know
where in the tree the eth devices are because it makes use of the
aliases node.

This of course requires that you do have an ethernet<n> alias to your
PCI based network device in your dt which I agree may seem strange for
an auto-probed bus, but in the case of a bootloader without PCI
support and/or sky2 support this seems the proper way to get the MAC
address from the bootloader to the driver.

    Ah, thanks for the explanation.

quoted

[...]

quoted

quoted

@@ -4805,8 +4808,27 @@ static struct net_device *sky2_init_netdev(struct

sky2_hw *hw, unsigned port,

         dev->features |= dev->hw_features;

+       /* try to get mac address in the following order:
+        * 1) from device tree data
+        * 2) from internal registers set by bootloader
+        */
+       iap = NULL;
+       if (IS_ENABLED(CONFIG_OF)) {
+               struct device_node *np = hw->pdev->dev.of_node;
+               if (np)
+                       iap = (unsigned char *) of_get_mac_address(np);
+       }
+
+       /* 2) mac registers set by bootloader
+        */

    Why not make it one-line? And it's kind of repetitive.

quoted

quoted

+       if (!iap || !is_valid_ether_addr(iap)) {
+               memcpy_fromio(&tmpaddr, hw->regs + B2_MAC_1 + port * 8,
+                       ETH_ALEN);

quoted

    This line should start right under & on the previous line.

agreed - not sure why checkpatch.pl didn't catch that

    It catches this only with --strict option.

quoted

WBR, Sergei

Thanks for the review!

    Not at all.

Tim

WBR, Sergei