Dear Florian:
For RDC R6040 FAST ETHERNET DRIVER.
There's an issue in r6040 driver and for some RDC platform. It only occurs when there's more than one Ethernet port.
When initializing the MAC in boot up, r6040 driver set the same phy address to different ports.
It also display in boot log as below.
mii_bus:phy_addr=0:01
mii_bus:phy_addr=1:01
It will let the eth0 and eth1 work not in independent.
That is, if there is a connection on eth1 but not on eth0. The eth1 can not work only if link up the eth0.
To fix this problem is that giving the different phy address to different ports.
I already tested it on r6040.c version 0.27 and version 0.28 by the change and is worked.
Below is the change
--- linux-3.7.2/drivers/net/ethernet/rdc/r6040.c.orig 2013-06-07 03:00:05.670250988 -0700
+++ linux-3.7.2/drivers/net/ethernet/rdc/r6040.c 2013-06-07 03:05:01.666407765 -0700
@@ -1037,8 +1037,18 @@ static int r6040_mii_probe(struct net_de
{
struct r6040_private *lp = netdev_priv(dev);
struct phy_device *phydev = NULL;
+ static int first_find_addr = 0;
+ int addr;
+ for ( addr = first_find_addr; addr < PHY_MAX_ADDR; addr++ ) {
+
+ if ( lp->mii_bus->phy_map[addr] )
+ phydev = lp->mii_bus->phy_map[addr];
+
+ if(phydev)
+ break;
+ }
+ first_find_addr++ ;
- phydev = phy_find_first(lp->mii_bus);
if (!phydev) {
dev_err(&lp->pdev->dev, "no PHY found\n");
return -ENODEV;
@@ -1194,6 +1204,10 @@ static int __devinit r6040_init_one(stru
lp->mii_bus->write = r6040_mdiobus_write;
lp->mii_bus->reset = r6040_mdiobus_reset;
lp->mii_bus->name = "r6040_eth_mii";
+ if ( card_idx == 0 )
+ lp->mii_bus->phy_mask = 0x1;
+ else
+ lp->mii_bus->phy_mask = 0x3;
snprintf(lp->mii_bus->id, MII_BUS_ID_SIZE, "%s-%x",
dev_name(&pdev->dev), card_idx);
lp->mii_bus->irq = kmalloc(sizeof(int)*PHY_MAX_ADDR, GFP_KERNEL);
Signed-off-by: Bruce Ye Developer <[email protected]>
Best Regards
Bruce
RDC Semiconductor
Dear Bruce,
2013/6/7 Bruce Ye (葉佳欣) <[email protected]>:
> Dear Florian:
> For RDC R6040 FAST ETHERNET DRIVER.
> There's an issue in r6040 driver and for some RDC platform. It only occurs when there's more than one Ethernet port.
> When initializing the MAC in boot up, r6040 driver set the same phy address to different ports.
> It also display in boot log as below.
> mii_bus:phy_addr=0:01
> mii_bus:phy_addr=1:01
>
> It will let the eth0 and eth1 work not in independent.
> That is, if there is a connection on eth1 but not on eth0. The eth1 can not work only if link up the eth0.
> To fix this problem is that giving the different phy address to different ports.
> I already tested it on r6040.c version 0.27 and version 0.28 by the change and is worked.
I see, thanks for the change. Have you verified that simply changing
mii_bus->phy_mask is sufficient? This should prevent the second MII
bus from probing the PHY at address 1. I will add this patch to my
patchqueue for r6040, thank you!
>
> Below is the change
>
> --- linux-3.7.2/drivers/net/ethernet/rdc/r6040.c.orig 2013-06-07 03:00:05.670250988 -0700
> +++ linux-3.7.2/drivers/net/ethernet/rdc/r6040.c 2013-06-07 03:05:01.666407765 -0700
> @@ -1037,8 +1037,18 @@ static int r6040_mii_probe(struct net_de
> {
> struct r6040_private *lp = netdev_priv(dev);
> struct phy_device *phydev = NULL;
> + static int first_find_addr = 0;
> + int addr;
> + for ( addr = first_find_addr; addr < PHY_MAX_ADDR; addr++ ) {
> +
> + if ( lp->mii_bus->phy_map[addr] )
> + phydev = lp->mii_bus->phy_map[addr];
> +
> + if(phydev)
> + break;
> + }
> + first_find_addr++ ;
>
> - phydev = phy_find_first(lp->mii_bus);
> if (!phydev) {
> dev_err(&lp->pdev->dev, "no PHY found\n");
> return -ENODEV;
> @@ -1194,6 +1204,10 @@ static int __devinit r6040_init_one(stru
> lp->mii_bus->write = r6040_mdiobus_write;
> lp->mii_bus->reset = r6040_mdiobus_reset;
> lp->mii_bus->name = "r6040_eth_mii";
> + if ( card_idx == 0 )
> + lp->mii_bus->phy_mask = 0x1;
> + else
> + lp->mii_bus->phy_mask = 0x3;
> snprintf(lp->mii_bus->id, MII_BUS_ID_SIZE, "%s-%x",
> dev_name(&pdev->dev), card_idx);
> lp->mii_bus->irq = kmalloc(sizeof(int)*PHY_MAX_ADDR, GFP_KERNEL);
>
>
>
> Signed-off-by: Bruce Ye Developer <[email protected]>
>
> Best Regards
> Bruce
> RDC Semiconductor