uClinux nommu does not detect PHY in SGMII mode

[Altera_Forum]

I have a 4sgx230 board and have been trying to get uClinux up on it. I have finally gotten it but I have a problem where the PHY is not detected. I am using TSE and the PHY is Marvel 88E1111. The board can only be used in SGMII mode since the other signals are not connected. I get this error at boot time: 

 .... .... console [ttyJ0] enabled ttyS0 at MMIO 0x8004c80 (irq = 10) is a Altera UART atse driver failed to detect an ether phy:/usr/src/nios2-linux/linux-2.6/drivers/net/atse.c:1455:atse_probe() TCP cubic registered NET: Registered protocol family 17 .... ....  I can go in the sopc_builder and change it to use MII/GMII and get it to detect the phy, but I know it will not work since the hardware is not present, first indication being that soft reset of the TSE never returns (SW reset bit ATSE_MAC_CC_SW_RESET_BIT gets stuck). ..... ..... eth0: atse.c: v1.1, June 3, 2008 by Joseph (Camel) Chen <joe4camel@gmail.com>  eth0: atse.c: modified version by Eintisy Chuang <eintisy.chuang@gfec.com.tw>  eth0: altera tripple speed, ether hw addr 00:07:ed:0d:09:19, marvell 88e1111 phy ..... .....  /> dhcpcd eth0 ATSE: Waiting on PHY link ...... atse driver:/usr/src/nios2-linux/linux-2.6/drivers/net/atse.c:530:mac sw reset bit never cleared! eth0 up, link speed 1000, full duplex, eth hw addr 00:07:ed:0d:09:19 dhcpcd[26]: timed out waiting for a valid DHCP seth0 down erver response   Anyone seen it before, I am about to jump into the code, but am hoping someone has already done that.

[Altera_Forum]

you can add a code to switch on the gmii mode in the driver files.

[Altera_Forum]

It seems to be already in SGMII mode, (if_mode_register == 0x03) 

 I just started looking into code, ill probably get it done soon

[Altera_Forum]

I guess the code is using the wrong set of registers for detecting the PHY, it works fine (tested with ping, dhcp, etc) if I use the second set. I changed the detection code: 

from:   for (phy_addr = 0x0; phy_addr < 0xFF; phy_addr++) {  /* set the phy address to the mac map data */  writel(phy_addr, ATSE_MAC_REG_MDIO_ADDR_0);  phy_id_1 = readl(ATSE_MAC_REG_MDIO_SPACE_0 + ATSE_PHY_ID_1_OFFSET);  phy_id_2 = readl(ATSE_MAC_REG_MDIO_SPACE_0 + ATSE_PHY_ID_2_OFFSET);  if (phy_id_1 != phy_id_2) {  found_phy = 1;  break;  }   To:   for (phy_addr = 0x0; phy_addr < 0xFF; phy_addr++) {  /* set the phy address to the mac map data */  writel(phy_addr, ATSE_MAC_REG_MDIO_ADDR_1);  phy_id_1 = readl(ATSE_MAC_REG_MDIO_SPACE_1 + ATSE_PHY_ID_1_OFFSET);  phy_id_2 = readl(ATSE_MAC_REG_MDIO_SPACE_1 + ATSE_PHY_ID_2_OFFSET);  if (phy_id_1 != phy_id_2) {  found_phy = 1;  break;  }   Ill go through the TSE docs again and create the right patch so it works in all modes.

[Altera_Forum]

The problem is that the current atse driver (besides many other assumptions) assumes that we will never use PCS mode. Here is the patch that should fix this problem. To compile it define USE_PCS_MODE. Ill try to submit it. 

  diff -uNr nios2-linux.vanilla/linux-2.6/drivers/net/atse.h nios2-linux/linux-2.6/drivers/net/atse.h --- nios2-linux.vanilla/linux-2.6/drivers/net/atse.h 2010-06-02 15:42:18.000000000 -0500 +++ nios2-linux/linux-2.6/drivers/net/atse.h 2010-06-29 14:32:24.000000000 -0500 @@ -72,10 +72,18 @@ /* number of bytes in one MDIO register */ # define ATSE_PHY_ID_1_OFFSET (2 * BYTES_IN_WORD) # define ATSE_PHY_ID_2_OFFSET (3 * BYTES_IN_WORD) + +#ifdef USE_PCS_MODE +#define ATSE_MAC_REG_MDIO_ADDR_1 (ATSE_MAC_BASE + 0x3C ) +#define ATSE_MAC_REG_MDIO_ADDR_0 (ATSE_MAC_BASE + 0x40 ) +#define ATSE_MAC_REG_PCS_0 (ATSE_MAC_BASE + 0x0200) +#define ATSE_MAC_REG_MDIO_SPACE_0 (ATSE_MAC_BASE + 0x0280) +#else # define ATSE_MAC_REG_MDIO_ADDR_0 (ATSE_MAC_BASE + 0x3C ) # define ATSE_MAC_REG_MDIO_ADDR_1 (ATSE_MAC_BASE + 0x40 ) # define ATSE_MAC_REG_MDIO_SPACE_0 (ATSE_MAC_BASE + 0x0200) # define ATSE_MAC_REG_MDIO_SPACE_1 (ATSE_MAC_BASE + 0x0280) +#endif # define ATSE_SET_PHY_MDIO_CONTROL(n) writel((n), ATSE_MAC_REG_MDIO_SPACE_0) # define ATSE_GET_PHY_MDIO_CONTROL() readl(ATSE_MAC_REG_MDIO_SPACE_0)

[Altera_Forum]

My impression is that the altera_tse (not atse) is the one that is written better. Though I haven't done a comparison, on first glance (a little over a year ago), altera_tse was the better driver...

[Altera_Forum]

I was not even able to get a peep out of that one (should have looked at it too), but now that I look at the source, you are correct. It is definitely better and worth looking. Ill take a look.

[Altera_Forum]

Hi, is there anyone get the altera_tse.c working on stratix iv GX FPGA development kit borad? Thanks.

[Altera_Forum]

Hi, 

 I'm also trying to get a 4SGX230 to talk to the 88E111 over SGMII. However, I cannot even compile the current atse.h/c driver because of:  #ifndef na_tse_mac_control_port # error "**** This build has not been configured with Alteras example design of TSE-SGDMA." # error "**** You need to do: make vendor_hwselect SYSPTF=<your_path>/NiosII_stratixII_2s60_RoHS_TSE_SGDMA_sopc.ptf" # error "**** Or, You need to unselect ATSE ethernet driver with make menuconfig" # endif  Which "make vendor_hwselect" isn't used / doesn't work anymore, so it's way out of date.   Using altera_tse.c/h doesn't seem to work for me. Negotiation fails and the kernel outputs: Trying 10/HALF  I've tried specifying the SGMII mode in the atl_tse_config struct, but nothing seems to work: static struct alt_tse_config tsemac0_config = { .mii_id = 0, .phy_addr = 0, .tse_supported_modes = PHY_GBIT_FEATURES, /* supported modes can be SUPPORTED_10baseT_Half SUPPORTED_10baseT_Full SUPPORTED_100baseT_Half SUPPORTED_100baseT_Full SUPPORTED_Autoneg SUPPORTED_TP SUPPORTED_MII ---------- Up to here is PHY_BASIC_FEATURES SUPPORTED_1000baseT_Half SUPPORTED_1000baseT_Full -- here PHY_GBIT_FEATURES */ .interface = PHY_INTERFACE_MODE_SGMII, /* Interfaces can be PHY_INTERFACE_MODE_MII PHY_INTERFACE_MODE_GMII PHY_INTERFACE_MODE_SGMII PHY_INTERFACE_MODE_TBI PHY_INTERFACE_MODE_RMII PHY_INTERFACE_MODE_RGMII PHY_INTERFACE_MODE_RGMII_ID PHY_INTERFACE_MODE_RGMII_RXID PHY_INTERFACE_MODE_RGMII_TXID PHY_INTERFACE_MODE_RTBI */ .flags = 0, /* these are apparently phy specific... */ .autoneg = AUTONEG_ENABLE, /* speed and duplex only valid if autoneg is AUTONED_DISABLE */ .speed = SPEED_100, /* SPEED_10, SPEED_100, SPEED_1000 */ .duplex = DUPLEX_HALF, /* DUPLEX_HALF, DUPLEX_FULL */ .rx_fifo_depth = ALT_TSE_TX_RX_FIFO_DEPTH, .tx_fifo_depth = ALT_TSE_TX_RX_FIFO_DEPTH, .ethaddr = {0x00 , 0x70 , 0xed , 0x11 , 0x12 , 0x12}, };   Is there an updated atse driver? I've updated my local git repository, but the log for the atse.h shows it was last touched in 2008.   Edit:  Ethtool output: root:/> ethtool eth0 Settings for eth0: Supported ports: Supported link modes: Supports auto-negotiation: Yes Advertised link modes: Not reported Advertised auto-negotiation: Yes Speed: 1000Mb/s Duplex: Full Port: MII PHYAD: 0 Transceiver: external Auto-negotiation: off Current message level: 0x00000000 (0) Link detected: yes

[Altera_Forum]

The same kind of hack for using PCS works for the altera_tse.c drivers. This is a diff against the latest checkout (linux 2.6.35): 

 --- a/drivers/net/altera_tse.c +++ b/drivers/net/altera_tse.c @@ -112,8 +112,10 @@ static int altera_tse_mdio_read(struct mii_bus *bus, int mii_id, int regnum) mac_dev = (alt_tse_mac *) bus->priv; /* set MDIO address */ - writel(mii_id, &mac_dev->mdio_phy0_addr); - mdio_regs = (unsigned int *) &mac_dev->mdio_phy0; + //writel(mii_id, &mac_dev->mdio_phy0_addr); + //mdio_regs = (unsigned int *) &mac_dev->mdio_phy0; + writel(mii_id, &mac_dev->mdio_phy1_addr); + mdio_regs = (unsigned int *) &mac_dev->mdio_phy1; /* get the data */ data = readl(&mdio_regs); @@ -130,8 +132,10 @@ static int altera_tse_mdio_write(struct mii_bus *bus, int mii_id, int regnum, u1 mac_dev = (alt_tse_mac *) bus->priv; /* set MDIO address */ - writel(mii_id, &mac_dev->mdio_phy0_addr); - mdio_regs = (unsigned int *) &mac_dev->mdio_phy0; + //writel(mii_id, &mac_dev->mdio_phy0_addr); + //mdio_regs = (unsigned int *) &mac_dev->mdio_phy0; + writel(mii_id, &mac_dev->mdio_phy1_addr); + mdio_regs = (unsigned int *) &mac_dev->mdio_phy1; /* get the data */ data = (unsigned int) value;