From: Timur Tabi Date: Wed, 20 Oct 2010 20:44:17 +0000 (-0500) Subject: powerpc/hydra: Add ethernet support on P5020/P3041 DS boards X-Git-Url: http://git.dujemihanovic.xyz/?a=commitdiff_plain;h=cba4614862c15b2199f412520de1994da33c013c;p=u-boot.git powerpc/hydra: Add ethernet support on P5020/P3041 DS boards Add support for RGMII, SGMII, and XAUI (10Gb) Ethernet on P3041DS & P5020DS ("Hydra"). The lane_to_slot[] array is initialized dynamically, since board switches can be used to control the muxing of SERDES lanes to slots. The BRDCFG1 PIXIS register is used to route the MII bus to the appropriate slot. The SERDES configuration is queried to help determine the routing between MACs and slot/phy combination. If a XAUI card is inserted, muxing for that card is enabled and never turned off. The PHY address for the 10G XAUI card depends on the slot in which it's inserted. If it's in slot 1, the address is 4. If it's in slot 2, the address is 0. Update the MDIO routing in the P3041DS and P5020DS device trees based on the board-level muxing. The SERDES configuration determines which SGMII/XGMII boards are located in which slots, and so the MDIO bus needs to be muxed correctly whenever talking to a PHY connected to any Fman MAC. The Fman Ethernet nodes in the device tree also need to be routed to the correct PHYs. Removed CONFIG_SYS_FMAN_FW as its not used anywhere. Signed-off-by: Ioana Radulescu Signed-off-by: Timur Tabi Signed-off-by: Andy Fleming Signed-off-by: Kumar Gala --- diff --git a/board/freescale/corenet_ds/Makefile b/board/freescale/corenet_ds/Makefile index 8b8d01a12e..c377c39f4b 100644 --- a/board/freescale/corenet_ds/Makefile +++ b/board/freescale/corenet_ds/Makefile @@ -28,7 +28,9 @@ LIB = $(obj)lib$(BOARD).o COBJS-y += $(BOARD).o COBJS-y += ddr.o +COBJS-$(CONFIG_P3041DS) += eth_hydra.o COBJS-$(CONFIG_P4080DS) += eth_p4080.o +COBJS-$(CONFIG_P5020DS) += eth_hydra.o COBJS-$(CONFIG_P3041DS) += p3041ds_ddr.o COBJS-$(CONFIG_P4080DS) += p4080ds_ddr.o COBJS-$(CONFIG_P5020DS) += p5020ds_ddr.o diff --git a/board/freescale/corenet_ds/eth_hydra.c b/board/freescale/corenet_ds/eth_hydra.c new file mode 100644 index 0000000000..91b340826d --- /dev/null +++ b/board/freescale/corenet_ds/eth_hydra.c @@ -0,0 +1,553 @@ +/* + * Copyright 2009-2011 Freescale Semiconductor, Inc. + * Author: Timur Tabi + * + * See file CREDITS for list of people who contributed to this + * project. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License as + * published by the Free Software Foundation; either version 2 of + * the License, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, + * MA 02111-1307 USA + */ + +/* + * This file handles the board muxing between the Fman Ethernet MACs and + * the RGMII/SGMII/XGMII PHYs on a Freescale P3041/P5020 "Hydra" reference + * board. The RGMII PHYs are the two on-board 1Gb ports. The SGMII PHYs are + * provided by the standard Freescale four-port SGMII riser card. The 10Gb + * XGMII PHY is provided via the XAUI riser card. Since there is only one + * Fman device on a P3041 and P5020, we only support one SGMII card and one + * RGMII card. + * + * Muxing is handled via the PIXIS BRDCFG1 register. The EMI1 bits control + * muxing among the RGMII PHYs and the SGMII PHYs. The value for RGMII is + * always the same (0). The value for SGMII depends on which slot the riser is + * inserted in. The EMI2 bits control muxing for the the XGMII. Like SGMII, + * the value is based on which slot the XAUI is inserted in. + * + * The SERDES configuration is used to determine where the SGMII and XAUI cards + * exist, and also which Fman MACs are routed to which PHYs. So for a given + * Fman MAC, there is one and only PHY it connects to. MACs cannot be routed + * to PHYs dynamically. + * + * + * This file also updates the device tree in three ways: + * + * 1) The status of each virtual MDIO node that is referenced by an Ethernet + * node is set to "okay". + * + * 2) The phy-handle property of each active Ethernet MAC node is set to the + * appropriate PHY node. + * + * 3) The "mux value" for each virtual MDIO node is set to the correct value, + * if necessary. Some virtual MDIO nodes do not have configurable mux + * values, so those values are hard-coded in the DTS. On the HYDRA board, + * the virtual MDIO node for the SGMII card needs to be updated. + * + * For all this to work, the device tree needs to have the following: + * + * 1) An alias for each PHY node that an Ethernet node could be routed to. + * + * 2) An alias for each real and virtual MDIO node that is disabled by default + * and might need to be enabled, and also might need to have its mux-value + * updated. + */ + +#include +#include +#include +#include +#include +#include +#include + +#include "../common/ngpixis.h" +#include "../common/fman.h" + +#ifdef CONFIG_FMAN_ENET + +#define BRDCFG1_EMI1_SEL_MASK 0x70 +#define BRDCFG1_EMI1_SEL_SLOT1 0x10 +#define BRDCFG1_EMI1_SEL_SLOT2 0x20 +#define BRDCFG1_EMI1_SEL_SLOT5 0x30 +#define BRDCFG1_EMI1_SEL_SLOT6 0x40 +#define BRDCFG1_EMI1_SEL_SLOT7 0x50 +#define BRDCFG1_EMI1_SEL_RGMII 0x00 +#define BRDCFG1_EMI1_EN 0x08 +#define BRDCFG1_EMI2_SEL_MASK 0x06 +#define BRDCFG1_EMI2_SEL_SLOT1 0x00 +#define BRDCFG1_EMI2_SEL_SLOT2 0x02 + +#define BRDCFG2_REG_GPIO_SEL 0x20 + +/* + * BRDCFG1 mask and value for each MAC + * + * This array contains the BRDCFG1 values (in mask/val format) that route the + * MDIO bus to a particular RGMII or SGMII PHY. + */ +struct { + u8 mask; + u8 val; +} mdio_mux[NUM_FM_PORTS]; + +/* + * Mapping of all 18 SERDES lanes to board slots. A value of '0' here means + * that the mapping must be determined dynamically, or that the lane maps to + * something other than a board slot + */ +static u8 lane_to_slot[] = { + 7, 7, 0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 1, 1, 0, 0 +}; + +/* + * Set the board muxing for a given MAC + * + * The MDIO layer calls this function every time it wants to talk to a PHY. + */ +void hydra_mux_mdio(u8 mask, u8 val) +{ + clrsetbits_8(&pixis->brdcfg1, mask, val); +} + +struct hydra_mdio { + u8 mask; + u8 val; + struct mii_dev *realbus; +}; + +static int hydra_mdio_read(struct mii_dev *bus, int addr, int devad, + int regnum) +{ + struct hydra_mdio *priv = bus->priv; + + hydra_mux_mdio(priv->mask, priv->val); + + return priv->realbus->read(priv->realbus, addr, devad, regnum); +} + +static int hydra_mdio_write(struct mii_dev *bus, int addr, int devad, + int regnum, u16 value) +{ + struct hydra_mdio *priv = bus->priv; + + hydra_mux_mdio(priv->mask, priv->val); + + return priv->realbus->write(priv->realbus, addr, devad, regnum, value); +} + +static int hydra_mdio_reset(struct mii_dev *bus) +{ + struct hydra_mdio *priv = bus->priv; + + return priv->realbus->reset(priv->realbus); +} + +static void hydra_mdio_set_mux(char *name, u8 mask, u8 val) +{ + struct mii_dev *bus = miiphy_get_dev_by_name(name); + struct hydra_mdio *priv = bus->priv; + + priv->mask = mask; + priv->val = val; +} + +static int hydra_mdio_init(char *realbusname, char *fakebusname) +{ + struct hydra_mdio *hmdio; + struct mii_dev *bus = mdio_alloc(); + + if (!bus) { + printf("Failed to allocate Hydra MDIO bus\n"); + return -1; + } + + hmdio = malloc(sizeof(*hmdio)); + if (!hmdio) { + printf("Failed to allocate Hydra private data\n"); + free(bus); + return -1; + } + + bus->read = hydra_mdio_read; + bus->write = hydra_mdio_write; + bus->reset = hydra_mdio_reset; + sprintf(bus->name, fakebusname); + + hmdio->realbus = miiphy_get_dev_by_name(realbusname); + + if (!hmdio->realbus) { + printf("No bus with name %s\n", realbusname); + free(bus); + free(hmdio); + return -1; + } + + bus->priv = hmdio; + + return mdio_register(bus); +} + +/* + * Given an alias or a path for a node, set the status of that node. + * + * If 'alias' is not a valid alias, then it is treated as a full path to the + * node. No error checking is performed. + * + * This function is normally called to set the status for a virtual MDIO node. + */ +static void fdt_set_node_status(void *fdt, const char *alias, + const char *status) +{ + const char *path = fdt_get_alias(fdt, alias); + + if (!path) + path = alias; + + do_fixup_by_path(fdt, path, "status", status, strlen(status) + 1, 1); +} + +/* + * Given an alias or a path for a node, set the mux value of that node. + * + * If 'alias' is not a valid alias, then it is treated as a full path to the + * node. No error checking is performed. + * + * This function is normally called to set the fsl,hydra-mdio-muxval property + * of a virtual MDIO node. + */ +static void fdt_set_mdio_mux(void *fdt, const char *alias, u32 mux) +{ + const char *path = fdt_get_alias(fdt, alias); + + if (!path) + path = alias; + + do_fixup_by_path(fdt, path, "fsl,hydra-mdio-muxval", + &mux, sizeof(mux), 1); +} + +/* + * Given the following ... + * + * 1) A pointer to an Fman Ethernet node (as identified by the 'compat' + * compatible string and 'addr' physical address) + * + * 2) An Fman port + * + * ... update the phy-handle property of the Ethernet node to point to the + * right PHY. This assumes that we already know the PHY for each port. That + * information is stored in mdio_mux[]. + * + * The offset of the Fman Ethernet node is also passed in for convenience, but + * it is not used, and we recalculate the offset anyway. + * + * Note that what we call "Fman ports" (enum fm_port) is really an Fman MAC. + * Inside the Fman, "ports" are things that connect to MACs. We only call them + * ports in U-Boot because on previous Ethernet devices (e.g. Gianfar), MACs + * and ports are the same thing. + * + * Note that this code would be cleaner if had a function called + * fm_info_get_phy_address(), which returns a value from the fm1_dtsec_info[] + * array. That's because all we're doing is figuring out the PHY address for + * a given Fman MAC and writing it to the device tree. Well, we already did + * the hard work to figure that out in board_eth_init(), so it's silly to + * repeat that here. + */ +void board_ft_fman_fixup_port(void *fdt, char *compat, phys_addr_t addr, + enum fm_port port, int offset) +{ + unsigned int mux = mdio_mux[port].val & mdio_mux[port].mask; + char phy[16]; + + if (port == FM1_10GEC1) { + /* XAUI */ + int lane = serdes_get_first_lane(XAUI_FM1); + if (lane >= 0) { + /* The XAUI PHY is identified by the slot */ + sprintf(phy, "phy_xgmii_%u", lane_to_slot[lane]); + fdt_set_phy_handle(fdt, compat, addr, phy); + } + return; + } + + if (mux == BRDCFG1_EMI1_SEL_RGMII) { + /* RGMII */ + /* The RGMII PHY is identified by the MAC connected to it */ + sprintf(phy, "phy_rgmii_%u", port == FM1_DTSEC4 ? 0 : 1); + fdt_set_phy_handle(fdt, compat, addr, phy); + } + + /* If it's not RGMII or XGMII, it must be SGMII */ + if (mux) { + /* The SGMII PHY is identified by the MAC connected to it */ + sprintf(phy, "phy_sgmii_%x", + CONFIG_SYS_FM1_DTSEC1_PHY_ADDR + (port - FM1_DTSEC1)); + fdt_set_phy_handle(fdt, compat, addr, phy); + } +} + +#define PIXIS_SW2_LANE_23_SEL 0x80 +#define PIXIS_SW2_LANE_45_SEL 0x40 +#define PIXIS_SW2_LANE_67_SEL_MASK 0x30 +#define PIXIS_SW2_LANE_67_SEL_5 0x00 +#define PIXIS_SW2_LANE_67_SEL_6 0x20 +#define PIXIS_SW2_LANE_67_SEL_7 0x10 +#define PIXIS_SW2_LANE_8_SEL 0x08 +#define PIXIS_SW2_LANE_1617_SEL 0x04 + +/* + * Initialize the lane_to_slot[] array. + * + * On the P4080DS "Expedition" board, the mapping of SERDES lanes to board + * slots is hard-coded. On the Hydra board, however, the mapping is controlled + * by board switch SW2, so the lane_to_slot[] array needs to be dynamically + * initialized. + */ +static void initialize_lane_to_slot(void) +{ + u8 sw2 = in_8(&PIXIS_SW(2)); + + lane_to_slot[2] = (sw2 & PIXIS_SW2_LANE_23_SEL) ? 7 : 4; + lane_to_slot[3] = lane_to_slot[2]; + + lane_to_slot[4] = (sw2 & PIXIS_SW2_LANE_45_SEL) ? 7 : 6; + lane_to_slot[5] = lane_to_slot[4]; + + switch (sw2 & PIXIS_SW2_LANE_67_SEL_MASK) { + case PIXIS_SW2_LANE_67_SEL_5: + lane_to_slot[6] = 5; + break; + case PIXIS_SW2_LANE_67_SEL_6: + lane_to_slot[6] = 6; + break; + case PIXIS_SW2_LANE_67_SEL_7: + lane_to_slot[6] = 7; + break; + } + lane_to_slot[7] = lane_to_slot[6]; + + lane_to_slot[8] = (sw2 & PIXIS_SW2_LANE_8_SEL) ? 3 : 0; + + lane_to_slot[16] = (sw2 & PIXIS_SW2_LANE_1617_SEL) ? 1 : 0; + lane_to_slot[17] = lane_to_slot[16]; +} + +#endif /* #ifdef CONFIG_FMAN_ENET */ + +/* + * Configure the status for the virtual MDIO nodes + * + * Rather than create the virtual MDIO nodes from scratch for each active + * virtual MDIO, we expect the DTS to have the nodes defined already, and we + * only enable the ones that are actually active. + * + * We assume that the DTS already hard-codes the status for all the + * virtual MDIO nodes to "disabled", so all we need to do is enable the + * active ones. + * + * For SGMII, we also need to set the mux value in the node. + */ +void fdt_fixup_board_enet(void *fdt) +{ +#ifdef CONFIG_FMAN_ENET + unsigned int i; + int lane; + + for (i = FM1_DTSEC1; i < FM1_DTSEC1 + CONFIG_SYS_NUM_FM1_DTSEC; i++) { + int idx = i - FM1_DTSEC1; + + switch (fm_info_get_enet_if(i)) { + case PHY_INTERFACE_MODE_SGMII: + lane = serdes_get_first_lane(SGMII_FM1_DTSEC1 + idx); + if (lane >= 0) { + fdt_set_node_status(fdt, "emi1_sgmii", "okay"); + /* Also set the MUX value */ + fdt_set_mdio_mux(fdt, "emi1_sgmii", + mdio_mux[i].val); + } + break; + case PHY_INTERFACE_MODE_RGMII: + fdt_set_node_status(fdt, "emi1_rgmii", "okay"); + break; + default: + break; + } + } + + lane = serdes_get_first_lane(XAUI_FM1); + if (lane >= 0) + fdt_set_node_status(fdt, "emi2_xgmii", "okay"); +#endif +} + +int board_eth_init(bd_t *bis) +{ +#ifdef CONFIG_FMAN_ENET + struct dtsec *tsec = (void *)CONFIG_SYS_FSL_FM1_DTSEC1_ADDR; + struct fsl_pq_mdio_info dtsec_mdio_info; + struct tgec_mdio_info tgec_mdio_info; + unsigned int i, slot; + int lane; + + printf("Initializing Fman\n"); + + initialize_lane_to_slot(); + + /* + * Set TBIPA on FM1@DTSEC1. This is needed for configurations + * where FM1@DTSEC1 isn't used directly, since it provides + * MDIO for other ports. + */ + out_be32(&tsec->tbipa, CONFIG_SYS_TBIPA_VALUE); + + /* We want to use the PIXIS to configure MUX routing, not GPIOs. */ + setbits_8(&pixis->brdcfg2, BRDCFG2_REG_GPIO_SEL); + + memset(mdio_mux, 0, sizeof(mdio_mux)); + + dtsec_mdio_info.regs = + (struct tsec_mii_mng *)CONFIG_SYS_FM1_DTSEC1_MDIO_ADDR; + dtsec_mdio_info.name = DEFAULT_FM_MDIO_NAME; + + /* Register the real 1G MDIO bus */ + fsl_pq_mdio_init(bis, &dtsec_mdio_info); + + tgec_mdio_info.regs = + (struct tgec_mdio_controller *)CONFIG_SYS_FM1_TGEC_MDIO_ADDR; + tgec_mdio_info.name = DEFAULT_FM_TGEC_MDIO_NAME; + + /* Register the real 10G MDIO bus */ + fm_tgec_mdio_init(bis, &tgec_mdio_info); + + /* Register the three virtual MDIO front-ends */ + hydra_mdio_init(DEFAULT_FM_MDIO_NAME, "HYDRA_RGMII_MDIO"); + hydra_mdio_init(DEFAULT_FM_MDIO_NAME, "HYDRA_SGMII_MDIO"); + + /* + * Program the DTSEC PHY addresses assuming that they are all SGMII. + * For any DTSEC that's RGMII, we'll override its PHY address later. + * We assume that DTSEC5 is only used for RGMII. + */ + fm_info_set_phy_address(FM1_DTSEC1, CONFIG_SYS_FM1_DTSEC1_PHY_ADDR); + fm_info_set_phy_address(FM1_DTSEC2, CONFIG_SYS_FM1_DTSEC2_PHY_ADDR); + fm_info_set_phy_address(FM1_DTSEC3, CONFIG_SYS_FM1_DTSEC3_PHY_ADDR); + fm_info_set_phy_address(FM1_DTSEC4, CONFIG_SYS_FM1_DTSEC4_PHY_ADDR); + + for (i = FM1_DTSEC1; i < FM1_DTSEC1 + CONFIG_SYS_NUM_FM1_DTSEC; i++) { + int idx = i - FM1_DTSEC1; + + switch (fm_info_get_enet_if(i)) { + case PHY_INTERFACE_MODE_SGMII: + lane = serdes_get_first_lane(SGMII_FM1_DTSEC1 + idx); + if (lane < 0) + break; + slot = lane_to_slot[lane]; + mdio_mux[i].mask = BRDCFG1_EMI1_SEL_MASK; + switch (slot) { + case 1: + /* Always DTSEC5 on Bank 3 */ + mdio_mux[i].val = BRDCFG1_EMI1_SEL_SLOT1 | + BRDCFG1_EMI1_EN; + break; + case 2: + mdio_mux[i].val = BRDCFG1_EMI1_SEL_SLOT2 | + BRDCFG1_EMI1_EN; + break; + case 5: + mdio_mux[i].val = BRDCFG1_EMI1_SEL_SLOT5 | + BRDCFG1_EMI1_EN; + break; + case 6: + mdio_mux[i].val = BRDCFG1_EMI1_SEL_SLOT6 | + BRDCFG1_EMI1_EN; + break; + case 7: + mdio_mux[i].val = BRDCFG1_EMI1_SEL_SLOT7 | + BRDCFG1_EMI1_EN; + break; + }; + + hydra_mdio_set_mux("HYDRA_SGMII_MDIO", + mdio_mux[i].mask, mdio_mux[i].val); + fm_info_set_mdio(i, + miiphy_get_dev_by_name("HYDRA_SGMII_MDIO")); + break; + case PHY_INTERFACE_MODE_RGMII: + /* + * If DTSEC4 is RGMII, then it's routed via via EC1 to + * the first on-board RGMII port. If DTSEC5 is RGMII, + * then it's routed via via EC2 to the second on-board + * RGMII port. The other DTSECs cannot be routed to + * RGMII. + */ + fm_info_set_phy_address(i, i == FM1_DTSEC4 ? 0 : 1); + mdio_mux[i].mask = BRDCFG1_EMI1_SEL_MASK; + mdio_mux[i].val = BRDCFG1_EMI1_SEL_RGMII | + BRDCFG1_EMI1_EN; + hydra_mdio_set_mux("HYDRA_RGMII_MDIO", + mdio_mux[i].mask, mdio_mux[i].val); + fm_info_set_mdio(i, + miiphy_get_dev_by_name("HYDRA_RGMII_MDIO")); + break; + case PHY_INTERFACE_MODE_NONE: + fm_info_set_phy_address(i, 0); + break; + default: + printf("Fman1: DTSEC%u set to unknown interface %i\n", + idx + 1, fm_info_get_enet_if(i)); + fm_info_set_phy_address(i, 0); + break; + } + } + + /* + * For 10G, we only support one XAUI card per Fman. If present, then we + * force its routing and never touch those bits again, which removes the + * need for Linux to do any muxing. This works because of the way + * BRDCFG1 is defined, but it's a bit hackish. + * + * The PHY address for the XAUI card depends on which slot it's in. The + * macros we use imply that the PHY address is based on which FM, but + * that's not true. On the P4080DS, FM1 could only use XAUI in slot 5, + * and FM2 could only use a XAUI in slot 4. On the Hydra board, we + * check the actual slot and just use the macros as-is, even though + * the P3041 and P5020 only have one Fman. + */ + lane = serdes_get_first_lane(XAUI_FM1); + if (lane >= 0) { + slot = lane_to_slot[lane]; + if (slot == 1) { + /* XAUI card is in slot 1 */ + clrsetbits_8(&pixis->brdcfg1, BRDCFG1_EMI2_SEL_MASK, + BRDCFG1_EMI2_SEL_SLOT1); + fm_info_set_phy_address(FM1_10GEC1, + CONFIG_SYS_FM1_10GEC1_PHY_ADDR); + } else { + /* XAUI card is in slot 2 */ + clrsetbits_8(&pixis->brdcfg1, BRDCFG1_EMI2_SEL_MASK, + BRDCFG1_EMI2_SEL_SLOT2); + fm_info_set_phy_address(FM1_10GEC1, + CONFIG_SYS_FM2_10GEC1_PHY_ADDR); + } + } + + fm_info_set_mdio(FM1_10GEC1, + miiphy_get_dev_by_name(DEFAULT_FM_TGEC_MDIO_NAME)); + + cpu_eth_init(bis); +#endif + + return pci_eth_init(bis); +}