Re: linux 2.6.33 on movidis x16

[David Daney <david.s.daney@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1642 bytes --]

On 03/18/2010 11:26 AM, David Daney wrote:
> On 03/18/2010 11:17 AM, Peter 'p2' De Schrijver wrote:
>> On 2010-03-15 18:27:41 (-0400), Jan Rovins<janr@adax.com>  wrote:
>>> Peter 'p2' De Schrijver wrote:
>>>> Hi,
>>>>
>>>> We are trying to get linux 2.6.33 to work on the movidis x16 board. 
>>>> Main thing
>>>> missing is the addresses of the PHYs. Does anyone know those ?
>>>> Unfortunately I don't have physical access to the board so I can't 
>>>> just look
>>>> for the PHY components being used.
>>>>
>>>> Thanks,
>>>>
>>>> Peter.
>>>>
>>>>
>>> Do you mean the Movidis x19 ?
>>
>> Maybe. AFAIK ours is called x16, and uses a cavium octeon CN3860 chip.
>>
>>> We have 2 of these in our lab. They are running the old 2.6.21.7  from
>>> the CnUsers 1.8.1  tool chain.
>>> They are currently being used by other developers for some application
>>> testing, but I can grab the serial console boot-up messages and send
>>> them to you, if the PHY info is in there.
>>
>> That would be helpful. Otherwise you could also try running ethtool 
>> on all the
>> interfaces. That should give you the PHY address as well.
>>
>
> I just found one of these.  Current theory is that the PHY addresses 
> are 0-7
>
> I don't think the PHY numbers are printed anywhere.

Can you try the attached patch?


from the back of the box we have:

---------------------------------------
| eth5 | eth7 |  | eth1 | eth3 |
---------------------------------------
| eth4 | eth6 |  | eth0 | eth2 |
---------------------------------------

I was able to boot 2.6.34-rc1 to a Debian root nfs mounted via eth1.

I didn't verify that eth4 - eth7 worked.

David Daney


[-- Attachment #2: movidis.diff --]
[-- Type: text/plain, Size: 1853 bytes --]

diff --git a/drivers/net/phy/mdio-octeon.c b/drivers/net/phy/mdio-octeon.c
index 61a4461..dfaaf30 100644
--- a/drivers/net/phy/mdio-octeon.c
+++ b/drivers/net/phy/mdio-octeon.c
@@ -87,6 +87,7 @@ static int octeon_mdiobus_write(struct mii_bus *bus, int phy_id,
 static int __init octeon_mdiobus_probe(struct platform_device *pdev)
 {
 	struct octeon_mdiobus *bus;
+	union cvmx_smix_en smi_en;
 	int i;
 	int err = -ENOENT;
 
@@ -102,6 +103,10 @@ static int __init octeon_mdiobus_probe(struct platform_device *pdev)
 	if (!bus->mii_bus)
 		goto err;
 
+	smi_en.u64 = 0;
+	smi_en.s.en = 1;
+	cvmx_write_csr(CVMX_SMIX_EN(bus->unit), smi_en.u64);
+
 	/*
 	 * Standard Octeon evaluation boards don't support phy
 	 * interrupts, we need to poll.
@@ -132,17 +137,22 @@ err_register:
 
 err:
 	devm_kfree(&pdev->dev, bus);
+	smi_en.u64 = 0;
+	cvmx_write_csr(CVMX_SMIX_EN(bus->unit), smi_en.u64);
 	return err;
 }
 
 static int __exit octeon_mdiobus_remove(struct platform_device *pdev)
 {
 	struct octeon_mdiobus *bus;
+	union cvmx_smix_en smi_en;
 
 	bus = dev_get_drvdata(&pdev->dev);
 
 	mdiobus_unregister(bus->mii_bus);
 	mdiobus_free(bus->mii_bus);
+	smi_en.u64 = 0;
+	cvmx_write_csr(CVMX_SMIX_EN(bus->unit), smi_en.u64);
 	return 0;
 }
 
diff --git a/drivers/staging/octeon/cvmx-helper-board.c b/drivers/staging/octeon/cvmx-helper-board.c
index 3085e38..fb8d3ee 100644
--- a/drivers/staging/octeon/cvmx-helper-board.c
+++ b/drivers/staging/octeon/cvmx-helper-board.c
@@ -153,6 +153,13 @@ int cvmx_helper_board_get_mii_address(int ipd_port)
 		 * through switch.
 		 */
 		return -1;
+	case CVMX_BOARD_TYPE_CUST_WSX16:
+		if (ipd_port >= 0 && ipd_port <= 3)
+			return ipd_port;
+		else if (ipd_port >= 16 && ipd_port <= 19)
+			return ipd_port - 16 + 4;
+		else
+			return -1;
 	}
 
 	/* Some unknown board. Somebody forgot to update this function... */