1
0
uboot-1.1.4-kirkwood/cpu/mpc8xx/fec.c
2024-01-07 23:57:24 +01:00

1021 lines
26 KiB
C

/*
* (C) Copyright 2000
* Wolfgang Denk, DENX Software Engineering, wd@denx.de.
*
* 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
*/
#include <common.h>
#include <malloc.h>
#include <commproc.h>
#include <net.h>
#include <command.h>
#undef ET_DEBUG
#if (CONFIG_COMMANDS & CFG_CMD_NET) && \
(defined(FEC_ENET) || defined(CONFIG_ETHER_ON_FEC1) || defined(CONFIG_ETHER_ON_FEC2))
/* compatibility test, if only FEC_ENET defined assume ETHER on FEC1 */
#if defined(FEC_ENET) && !defined(CONFIG_ETHER_ON_FEC1) && !defined(CONFIG_ETHER_ON_FEC2)
#define CONFIG_ETHER_ON_FEC1 1
#endif
/* define WANT_MII when MII support is required */
#if defined(CFG_DISCOVER_PHY) || defined(CONFIG_FEC1_PHY) || defined(CONFIG_FEC2_PHY)
#define WANT_MII
#else
#undef WANT_MII
#endif
#if defined(WANT_MII)
#include <miiphy.h>
#if !(defined(CONFIG_MII) || (CONFIG_COMMANDS & CFG_CMD_MII))
#error "CONFIG_MII has to be defined!"
#endif
#endif
#if defined(CONFIG_RMII) && !defined(WANT_MII)
#error RMII support is unusable without a working PHY.
#endif
#ifdef CFG_DISCOVER_PHY
static int mii_discover_phy(struct eth_device *dev);
#endif
int fec8xx_miiphy_read(char *devname, unsigned char addr,
unsigned char reg, unsigned short *value);
int fec8xx_miiphy_write(char *devname, unsigned char addr,
unsigned char reg, unsigned short value);
static struct ether_fcc_info_s
{
int ether_index;
int fecp_offset;
int phy_addr;
int actual_phy_addr;
int initialized;
}
ether_fcc_info[] = {
#if defined(CONFIG_ETHER_ON_FEC1)
{
0,
offsetof(immap_t, im_cpm.cp_fec1),
#if defined(CONFIG_FEC1_PHY)
CONFIG_FEC1_PHY,
#else
-1, /* discover */
#endif
-1,
0,
},
#endif
#if defined(CONFIG_ETHER_ON_FEC2)
{
1,
offsetof(immap_t, im_cpm.cp_fec2),
#if defined(CONFIG_FEC2_PHY)
CONFIG_FEC2_PHY,
#else
-1,
#endif
-1,
0,
},
#endif
};
/* Ethernet Transmit and Receive Buffers */
#define DBUF_LENGTH 1520
#define TX_BUF_CNT 2
#define TOUT_LOOP 100
#define PKT_MAXBUF_SIZE 1518
#define PKT_MINBUF_SIZE 64
#define PKT_MAXBLR_SIZE 1520
#ifdef __GNUC__
static char txbuf[DBUF_LENGTH] __attribute__ ((aligned(8)));
#else
#error txbuf must be aligned.
#endif
static uint rxIdx; /* index of the current RX buffer */
static uint txIdx; /* index of the current TX buffer */
/*
* FEC Ethernet Tx and Rx buffer descriptors allocated at the
* immr->udata_bd address on Dual-Port RAM
* Provide for Double Buffering
*/
typedef volatile struct CommonBufferDescriptor {
cbd_t rxbd[PKTBUFSRX]; /* Rx BD */
cbd_t txbd[TX_BUF_CNT]; /* Tx BD */
} RTXBD;
static RTXBD *rtx = NULL;
static int fec_send(struct eth_device* dev, volatile void *packet, int length);
static int fec_recv(struct eth_device* dev);
static int fec_init(struct eth_device* dev, bd_t * bd);
static void fec_halt(struct eth_device* dev);
int fec_initialize(bd_t *bis)
{
struct eth_device* dev;
struct ether_fcc_info_s *efis;
int i;
for (i = 0; i < sizeof(ether_fcc_info) / sizeof(ether_fcc_info[0]); i++) {
dev = malloc(sizeof(*dev));
if (dev == NULL)
hang();
memset(dev, 0, sizeof(*dev));
/* for FEC1 make sure that the name of the interface is the same
as the old one for compatibility reasons */
if (i == 0) {
sprintf (dev->name, "FEC ETHERNET");
} else {
sprintf (dev->name, "FEC%d ETHERNET",
ether_fcc_info[i].ether_index + 1);
}
efis = &ether_fcc_info[i];
/*
* reset actual phy addr
*/
efis->actual_phy_addr = -1;
dev->priv = efis;
dev->init = fec_init;
dev->halt = fec_halt;
dev->send = fec_send;
dev->recv = fec_recv;
eth_register(dev);
#if defined(CONFIG_MII) || (CONFIG_COMMANDS & CFG_CMD_MII)
miiphy_register(dev->name,
fec8xx_miiphy_read, fec8xx_miiphy_write);
#endif
}
return 1;
}
static int fec_send(struct eth_device* dev, volatile void *packet, int length)
{
int j, rc;
struct ether_fcc_info_s *efis = dev->priv;
volatile fec_t *fecp = (volatile fec_t *)(CFG_IMMR + efis->fecp_offset);
/* section 16.9.23.3
* Wait for ready
*/
j = 0;
while ((rtx->txbd[txIdx].cbd_sc & BD_ENET_TX_READY) && (j<TOUT_LOOP)) {
udelay(1);
j++;
}
if (j>=TOUT_LOOP) {
printf("TX not ready\n");
}
rtx->txbd[txIdx].cbd_bufaddr = (uint)packet;
rtx->txbd[txIdx].cbd_datlen = length;
rtx->txbd[txIdx].cbd_sc |= BD_ENET_TX_READY | BD_ENET_TX_LAST;
__asm__ ("eieio");
/* Activate transmit Buffer Descriptor polling */
fecp->fec_x_des_active = 0x01000000; /* Descriptor polling active */
j = 0;
while ((rtx->txbd[txIdx].cbd_sc & BD_ENET_TX_READY) && (j<TOUT_LOOP)) {
#if defined(CONFIG_ICU862)
udelay(10);
#else
udelay(1);
#endif
j++;
}
if (j>=TOUT_LOOP) {
printf("TX timeout\n");
}
#ifdef ET_DEBUG
printf("%s[%d] %s: cycles: %d status: %x retry cnt: %d\n",
__FILE__,__LINE__,__FUNCTION__,j,rtx->txbd[txIdx].cbd_sc,
(rtx->txbd[txIdx].cbd_sc & 0x003C)>>2);
#endif
/* return only status bits */;
rc = (rtx->txbd[txIdx].cbd_sc & BD_ENET_TX_STATS);
txIdx = (txIdx + 1) % TX_BUF_CNT;
return rc;
}
static int fec_recv (struct eth_device *dev)
{
struct ether_fcc_info_s *efis = dev->priv;
volatile fec_t *fecp =
(volatile fec_t *) (CFG_IMMR + efis->fecp_offset);
int length;
for (;;) {
/* section 16.9.23.2 */
if (rtx->rxbd[rxIdx].cbd_sc & BD_ENET_RX_EMPTY) {
length = -1;
break; /* nothing received - leave for() loop */
}
length = rtx->rxbd[rxIdx].cbd_datlen;
if (rtx->rxbd[rxIdx].cbd_sc & 0x003f) {
#ifdef ET_DEBUG
printf ("%s[%d] err: %x\n",
__FUNCTION__, __LINE__,
rtx->rxbd[rxIdx].cbd_sc);
#endif
} else {
volatile uchar *rx = NetRxPackets[rxIdx];
length -= 4;
#if (CONFIG_COMMANDS & CFG_CMD_CDP)
if ((rx[0] & 1) != 0
&& memcmp ((uchar *) rx, NetBcastAddr, 6) != 0
&& memcmp ((uchar *) rx, NetCDPAddr, 6) != 0)
rx = NULL;
#endif
/*
* Pass the packet up to the protocol layers.
*/
if (rx != NULL)
NetReceive (rx, length);
}
/* Give the buffer back to the FEC. */
rtx->rxbd[rxIdx].cbd_datlen = 0;
/* wrap around buffer index when necessary */
if ((rxIdx + 1) >= PKTBUFSRX) {
rtx->rxbd[PKTBUFSRX - 1].cbd_sc =
(BD_ENET_RX_WRAP | BD_ENET_RX_EMPTY);
rxIdx = 0;
} else {
rtx->rxbd[rxIdx].cbd_sc = BD_ENET_RX_EMPTY;
rxIdx++;
}
__asm__ ("eieio");
/* Try to fill Buffer Descriptors */
fecp->fec_r_des_active = 0x01000000; /* Descriptor polling active */
}
return length;
}
/**************************************************************
*
* FEC Ethernet Initialization Routine
*
*************************************************************/
#define FEC_ECNTRL_PINMUX 0x00000004
#define FEC_ECNTRL_ETHER_EN 0x00000002
#define FEC_ECNTRL_RESET 0x00000001
#define FEC_RCNTRL_BC_REJ 0x00000010
#define FEC_RCNTRL_PROM 0x00000008
#define FEC_RCNTRL_MII_MODE 0x00000004
#define FEC_RCNTRL_DRT 0x00000002
#define FEC_RCNTRL_LOOP 0x00000001
#define FEC_TCNTRL_FDEN 0x00000004
#define FEC_TCNTRL_HBC 0x00000002
#define FEC_TCNTRL_GTS 0x00000001
#define FEC_RESET_DELAY 50
#if defined(CONFIG_RMII)
static inline void fec_10Mbps(struct eth_device *dev)
{
struct ether_fcc_info_s *efis = dev->priv;
int fecidx = efis->ether_index;
uint mask = (fecidx == 0) ? 0x0000010 : 0x0000008;
if ((unsigned int)fecidx >= 2)
hang();
((volatile immap_t *)CFG_IMMR)->im_cpm.cp_cptr |= mask;
}
static inline void fec_100Mbps(struct eth_device *dev)
{
struct ether_fcc_info_s *efis = dev->priv;
int fecidx = efis->ether_index;
uint mask = (fecidx == 0) ? 0x0000010 : 0x0000008;
if ((unsigned int)fecidx >= 2)
hang();
((volatile immap_t *)CFG_IMMR)->im_cpm.cp_cptr &= ~mask;
}
#endif
static inline void fec_full_duplex(struct eth_device *dev)
{
struct ether_fcc_info_s *efis = dev->priv;
volatile fec_t *fecp = (volatile fec_t *)(CFG_IMMR + efis->fecp_offset);
fecp->fec_r_cntrl &= ~FEC_RCNTRL_DRT;
fecp->fec_x_cntrl |= FEC_TCNTRL_FDEN; /* FD enable */
}
static inline void fec_half_duplex(struct eth_device *dev)
{
struct ether_fcc_info_s *efis = dev->priv;
volatile fec_t *fecp = (volatile fec_t *)(CFG_IMMR + efis->fecp_offset);
fecp->fec_r_cntrl |= FEC_RCNTRL_DRT;
fecp->fec_x_cntrl &= ~FEC_TCNTRL_FDEN; /* FD disable */
}
static void fec_pin_init(int fecidx)
{
DECLARE_GLOBAL_DATA_PTR;
bd_t *bd = gd->bd;
volatile immap_t *immr = (immap_t *) CFG_IMMR;
volatile fec_t *fecp;
/*
* only two FECs please
*/
if ((unsigned int)fecidx >= 2)
hang();
if (fecidx == 0)
fecp = &immr->im_cpm.cp_fec1;
else
fecp = &immr->im_cpm.cp_fec2;
/*
* Set MII speed to 2.5 MHz or slightly below.
* * According to the MPC860T (Rev. D) Fast ethernet controller user
* * manual (6.2.14),
* * the MII management interface clock must be less than or equal
* * to 2.5 MHz.
* * This MDC frequency is equal to system clock / (2 * MII_SPEED).
* * Then MII_SPEED = system_clock / 2 * 2,5 Mhz.
*/
fecp->fec_mii_speed = ((bd->bi_intfreq + 4999999) / 5000000) << 1;
#if defined(CONFIG_NETTA) || defined(CONFIG_NETPHONE) || defined(CONFIG_NETTA2)
/* our PHYs are the limit at 2.5 MHz */
fecp->fec_mii_speed <<= 1;
#endif
#if defined(CONFIG_MPC885_FAMILY) && defined(WANT_MII)
/* use MDC for MII */
immr->im_ioport.iop_pdpar |= 0x0080;
immr->im_ioport.iop_pddir &= ~0x0080;
#endif
if (fecidx == 0) {
#if defined(CONFIG_ETHER_ON_FEC1)
#if defined(CONFIG_MPC885_FAMILY) /* MPC87x/88x have got 2 FECs and different pinout */
#if !defined(CONFIG_RMII)
immr->im_ioport.iop_papar |= 0xf830;
immr->im_ioport.iop_padir |= 0x0830;
immr->im_ioport.iop_padir &= ~0xf000;
immr->im_cpm.cp_pbpar |= 0x00001001;
immr->im_cpm.cp_pbdir &= ~0x00001001;
immr->im_ioport.iop_pcpar |= 0x000c;
immr->im_ioport.iop_pcdir &= ~0x000c;
immr->im_cpm.cp_pepar |= 0x00000003;
immr->im_cpm.cp_pedir |= 0x00000003;
immr->im_cpm.cp_peso &= ~0x00000003;
immr->im_cpm.cp_cptr &= ~0x00000100;
#else
#if !defined(CONFIG_FEC1_PHY_NORXERR)
immr->im_ioport.iop_papar |= 0x1000;
immr->im_ioport.iop_padir &= ~0x1000;
#endif
immr->im_ioport.iop_papar |= 0xe810;
immr->im_ioport.iop_padir |= 0x0810;
immr->im_ioport.iop_padir &= ~0xe000;
immr->im_cpm.cp_pbpar |= 0x00000001;
immr->im_cpm.cp_pbdir &= ~0x00000001;
immr->im_cpm.cp_cptr |= 0x00000100;
immr->im_cpm.cp_cptr &= ~0x00000050;
#endif /* !CONFIG_RMII */
#elif !defined(CONFIG_ICU862) && !defined(CONFIG_IAD210)
/*
* Configure all of port D for MII.
*/
immr->im_ioport.iop_pdpar = 0x1fff;
/*
* Bits moved from Rev. D onward
*/
if ((get_immr(0) & 0xffff) < 0x0501)
immr->im_ioport.iop_pddir = 0x1c58; /* Pre rev. D */
else
immr->im_ioport.iop_pddir = 0x1fff; /* Rev. D and later */
#else
/*
* Configure port A for MII.
*/
#if defined(CONFIG_ICU862) && defined(CFG_DISCOVER_PHY)
/*
* On the ICU862 board the MII-MDC pin is routed to PD8 pin
* * of CPU, so for this board we need to configure Utopia and
* * enable PD8 to MII-MDC function
*/
immr->im_ioport.iop_pdpar |= 0x4080;
#endif
/*
* Has Utopia been configured?
*/
if (immr->im_ioport.iop_pdpar & (0x8000 >> 1)) {
/*
* YES - Use MUXED mode for UTOPIA bus.
* This frees Port A for use by MII (see 862UM table 41-6).
*/
immr->im_ioport.utmode &= ~0x80;
} else {
/*
* NO - set SPLIT mode for UTOPIA bus.
*
* This doesn't really effect UTOPIA (which isn't
* enabled anyway) but just tells the 862
* to use port A for MII (see 862UM table 41-6).
*/
immr->im_ioport.utmode |= 0x80;
}
#endif /* !defined(CONFIG_ICU862) */
#endif /* CONFIG_ETHER_ON_FEC1 */
} else if (fecidx == 1) {
#if defined(CONFIG_ETHER_ON_FEC2)
#if defined(CONFIG_MPC885_FAMILY) /* MPC87x/88x have got 2 FECs and different pinout */
#if !defined(CONFIG_RMII)
#warning this configuration is not tested; please report if it works
immr->im_cpm.cp_pepar |= 0x0003fffc;
immr->im_cpm.cp_pedir |= 0x0003fffc;
immr->im_cpm.cp_peso &= ~0x000087fc;
immr->im_cpm.cp_peso |= 0x00037800;
immr->im_cpm.cp_cptr &= ~0x00000080;
#else
#if !defined(CONFIG_FEC2_PHY_NORXERR)
immr->im_cpm.cp_pepar |= 0x00000010;
immr->im_cpm.cp_pedir |= 0x00000010;
immr->im_cpm.cp_peso &= ~0x00000010;
#endif
immr->im_cpm.cp_pepar |= 0x00039620;
immr->im_cpm.cp_pedir |= 0x00039620;
immr->im_cpm.cp_peso |= 0x00031000;
immr->im_cpm.cp_peso &= ~0x00008620;
immr->im_cpm.cp_cptr |= 0x00000080;
immr->im_cpm.cp_cptr &= ~0x00000028;
#endif /* CONFIG_RMII */
#endif /* CONFIG_MPC885_FAMILY */
#endif /* CONFIG_ETHER_ON_FEC2 */
}
}
static int fec_init (struct eth_device *dev, bd_t * bd)
{
struct ether_fcc_info_s *efis = dev->priv;
volatile immap_t *immr = (immap_t *) CFG_IMMR;
volatile fec_t *fecp =
(volatile fec_t *) (CFG_IMMR + efis->fecp_offset);
int i;
if (efis->ether_index == 0) {
#if defined(CONFIG_FADS) /* FADS family uses FPGA (BCSR) to control PHYs */
#if defined(CONFIG_MPC885ADS)
*(vu_char *) BCSR5 &= ~(BCSR5_MII1_EN | BCSR5_MII1_RST);
#else
/* configure FADS for fast (FEC) ethernet, half-duplex */
/* The LXT970 needs about 50ms to recover from reset, so
* wait for it by discovering the PHY before leaving eth_init().
*/
{
volatile uint *bcsr4 = (volatile uint *) BCSR4;
*bcsr4 = (*bcsr4 & ~(BCSR4_FETH_EN | BCSR4_FETHCFG1))
| (BCSR4_FETHCFG0 | BCSR4_FETHFDE |
BCSR4_FETHRST);
/* reset the LXT970 PHY */
*bcsr4 &= ~BCSR4_FETHRST;
udelay (10);
*bcsr4 |= BCSR4_FETHRST;
udelay (10);
}
#endif /* CONFIG_MPC885ADS */
#endif /* CONFIG_FADS */
}
/* Whack a reset.
* A delay is required between a reset of the FEC block and
* initialization of other FEC registers because the reset takes
* some time to complete. If you don't delay, subsequent writes
* to FEC registers might get killed by the reset routine which is
* still in progress.
*/
fecp->fec_ecntrl = FEC_ECNTRL_PINMUX | FEC_ECNTRL_RESET;
for (i = 0;
(fecp->fec_ecntrl & FEC_ECNTRL_RESET) && (i < FEC_RESET_DELAY);
++i) {
udelay (1);
}
if (i == FEC_RESET_DELAY) {
printf ("FEC_RESET_DELAY timeout\n");
return 0;
}
/* We use strictly polling mode only
*/
fecp->fec_imask = 0;
/* Clear any pending interrupt
*/
fecp->fec_ievent = 0xffc0;
/* No need to set the IVEC register */
/* Set station address
*/
#define ea eth_get_dev()->enetaddr
fecp->fec_addr_low = (ea[0] << 24) | (ea[1] << 16) | (ea[2] << 8) | (ea[3]);
fecp->fec_addr_high = (ea[4] << 8) | (ea[5]);
#undef ea
#if (CONFIG_COMMANDS & CFG_CMD_CDP)
/*
* Turn on multicast address hash table
*/
fecp->fec_hash_table_high = 0xffffffff;
fecp->fec_hash_table_low = 0xffffffff;
#else
/* Clear multicast address hash table
*/
fecp->fec_hash_table_high = 0;
fecp->fec_hash_table_low = 0;
#endif
/* Set maximum receive buffer size.
*/
fecp->fec_r_buff_size = PKT_MAXBLR_SIZE;
/* Set maximum frame length
*/
fecp->fec_r_hash = PKT_MAXBUF_SIZE;
/*
* Setup Buffers and Buffer Desriptors
*/
rxIdx = 0;
txIdx = 0;
if (!rtx) {
#ifdef CFG_ALLOC_DPRAM
rtx = (RTXBD *) (immr->im_cpm.cp_dpmem +
dpram_alloc_align (sizeof (RTXBD), 8));
#else
rtx = (RTXBD *) (immr->im_cpm.cp_dpmem + CPM_FEC_BASE);
#endif
}
/*
* Setup Receiver Buffer Descriptors (13.14.24.18)
* Settings:
* Empty, Wrap
*/
for (i = 0; i < PKTBUFSRX; i++) {
rtx->rxbd[i].cbd_sc = BD_ENET_RX_EMPTY;
rtx->rxbd[i].cbd_datlen = 0; /* Reset */
rtx->rxbd[i].cbd_bufaddr = (uint) NetRxPackets[i];
}
rtx->rxbd[PKTBUFSRX - 1].cbd_sc |= BD_ENET_RX_WRAP;
/*
* Setup Ethernet Transmitter Buffer Descriptors (13.14.24.19)
* Settings:
* Last, Tx CRC
*/
for (i = 0; i < TX_BUF_CNT; i++) {
rtx->txbd[i].cbd_sc = BD_ENET_TX_LAST | BD_ENET_TX_TC;
rtx->txbd[i].cbd_datlen = 0; /* Reset */
rtx->txbd[i].cbd_bufaddr = (uint) (&txbuf[0]);
}
rtx->txbd[TX_BUF_CNT - 1].cbd_sc |= BD_ENET_TX_WRAP;
/* Set receive and transmit descriptor base
*/
fecp->fec_r_des_start = (unsigned int) (&rtx->rxbd[0]);
fecp->fec_x_des_start = (unsigned int) (&rtx->txbd[0]);
/* Enable MII mode
*/
#if 0 /* Full duplex mode */
fecp->fec_r_cntrl = FEC_RCNTRL_MII_MODE;
fecp->fec_x_cntrl = FEC_TCNTRL_FDEN;
#else /* Half duplex mode */
fecp->fec_r_cntrl = FEC_RCNTRL_MII_MODE | FEC_RCNTRL_DRT;
fecp->fec_x_cntrl = 0;
#endif
/* Enable big endian and don't care about SDMA FC.
*/
fecp->fec_fun_code = 0x78000000;
/*
* Setup the pin configuration of the FEC
*/
fec_pin_init (efis->ether_index);
rxIdx = 0;
txIdx = 0;
/*
* Now enable the transmit and receive processing
*/
fecp->fec_ecntrl = FEC_ECNTRL_PINMUX | FEC_ECNTRL_ETHER_EN;
if (efis->phy_addr == -1) {
#ifdef CFG_DISCOVER_PHY
/*
* wait for the PHY to wake up after reset
*/
efis->actual_phy_addr = mii_discover_phy (dev);
if (efis->actual_phy_addr == -1) {
printf ("Unable to discover phy!\n");
return 0;
}
#else
efis->actual_phy_addr = -1;
#endif
} else {
efis->actual_phy_addr = efis->phy_addr;
}
#if defined(CONFIG_MII) && defined(CONFIG_RMII)
/* the MII interface is connected to FEC1
* so for the miiphy_xxx function to work we must
* call mii_init since fec_halt messes the thing up
*/
if (efis->ether_index != 0)
mii_init();
/*
* adapt the RMII speed to the speed of the phy
*/
if (miiphy_speed (dev->name, efis->actual_phy_addr) == _100BASET) {
fec_100Mbps (dev);
} else {
fec_10Mbps (dev);
}
#endif
#if defined(CONFIG_MII)
/*
* adapt to the half/full speed settings
*/
if (miiphy_duplex (dev->name, efis->actual_phy_addr) == FULL) {
fec_full_duplex (dev);
} else {
fec_half_duplex (dev);
}
#endif
/* And last, try to fill Rx Buffer Descriptors */
fecp->fec_r_des_active = 0x01000000; /* Descriptor polling active */
efis->initialized = 1;
return 1;
}
static void fec_halt(struct eth_device* dev)
{
struct ether_fcc_info_s *efis = dev->priv;
volatile fec_t *fecp = (volatile fec_t *)(CFG_IMMR + efis->fecp_offset);
int i;
/* avoid halt if initialized; mii gets stuck otherwise */
if (!efis->initialized)
return;
/* Whack a reset.
* A delay is required between a reset of the FEC block and
* initialization of other FEC registers because the reset takes
* some time to complete. If you don't delay, subsequent writes
* to FEC registers might get killed by the reset routine which is
* still in progress.
*/
fecp->fec_ecntrl = FEC_ECNTRL_PINMUX | FEC_ECNTRL_RESET;
for (i = 0;
(fecp->fec_ecntrl & FEC_ECNTRL_RESET) && (i < FEC_RESET_DELAY);
++i) {
udelay (1);
}
if (i == FEC_RESET_DELAY) {
printf ("FEC_RESET_DELAY timeout\n");
return;
}
efis->initialized = 0;
}
#if defined(CFG_DISCOVER_PHY) || defined(CONFIG_MII) || (CONFIG_COMMANDS & CFG_CMD_MII)
/* Make MII read/write commands for the FEC.
*/
#define mk_mii_read(ADDR, REG) (0x60020000 | ((ADDR << 23) | \
(REG & 0x1f) << 18))
#define mk_mii_write(ADDR, REG, VAL) (0x50020000 | ((ADDR << 23) | \
(REG & 0x1f) << 18) | \
(VAL & 0xffff))
/* Interrupt events/masks.
*/
#define FEC_ENET_HBERR ((uint)0x80000000) /* Heartbeat error */
#define FEC_ENET_BABR ((uint)0x40000000) /* Babbling receiver */
#define FEC_ENET_BABT ((uint)0x20000000) /* Babbling transmitter */
#define FEC_ENET_GRA ((uint)0x10000000) /* Graceful stop complete */
#define FEC_ENET_TXF ((uint)0x08000000) /* Full frame transmitted */
#define FEC_ENET_TXB ((uint)0x04000000) /* A buffer was transmitted */
#define FEC_ENET_RXF ((uint)0x02000000) /* Full frame received */
#define FEC_ENET_RXB ((uint)0x01000000) /* A buffer was received */
#define FEC_ENET_MII ((uint)0x00800000) /* MII interrupt */
#define FEC_ENET_EBERR ((uint)0x00400000) /* SDMA bus error */
/* PHY identification
*/
#define PHY_ID_LXT970 0x78100000 /* LXT970 */
#define PHY_ID_LXT971 0x001378e0 /* LXT971 and 972 */
#define PHY_ID_82555 0x02a80150 /* Intel 82555 */
#define PHY_ID_QS6612 0x01814400 /* QS6612 */
#define PHY_ID_AMD79C784 0x00225610 /* AMD 79C784 */
#define PHY_ID_LSI80225 0x0016f870 /* LSI 80225 */
#define PHY_ID_LSI80225B 0x0016f880 /* LSI 80225/B */
#define PHY_ID_DM9161 0x0181B880 /* Davicom DM9161 */
/* send command to phy using mii, wait for result */
static uint
mii_send(uint mii_cmd)
{
uint mii_reply;
volatile fec_t *ep;
int cnt;
ep = &(((immap_t *)CFG_IMMR)->im_cpm.cp_fec);
ep->fec_mii_data = mii_cmd; /* command to phy */
/* wait for mii complete */
cnt = 0;
while (!(ep->fec_ievent & FEC_ENET_MII)) {
if (++cnt > 1000) {
printf("mii_send STUCK!\n");
break;
}
}
mii_reply = ep->fec_mii_data; /* result from phy */
ep->fec_ievent = FEC_ENET_MII; /* clear MII complete */
#if 0
printf("%s[%d] %s: sent=0x%8.8x, reply=0x%8.8x\n",
__FILE__,__LINE__,__FUNCTION__,mii_cmd,mii_reply);
#endif
return (mii_reply & 0xffff); /* data read from phy */
}
#endif /* CFG_DISCOVER_PHY || (CONFIG_COMMANDS & CFG_CMD_MII) */
#if defined(CFG_DISCOVER_PHY)
static int mii_discover_phy(struct eth_device *dev)
{
#define MAX_PHY_PASSES 11
uint phyno;
int pass;
uint phytype;
int phyaddr;
phyaddr = -1; /* didn't find a PHY yet */
for (pass = 1; pass <= MAX_PHY_PASSES && phyaddr < 0; ++pass) {
if (pass > 1) {
/* PHY may need more time to recover from reset.
* The LXT970 needs 50ms typical, no maximum is
* specified, so wait 10ms before try again.
* With 11 passes this gives it 100ms to wake up.
*/
udelay(10000); /* wait 10ms */
}
for (phyno = 0; phyno < 32 && phyaddr < 0; ++phyno) {
phytype = mii_send(mk_mii_read(phyno, PHY_PHYIDR1));
#ifdef ET_DEBUG
printf("PHY type 0x%x pass %d type ", phytype, pass);
#endif
if (phytype != 0xffff) {
phyaddr = phyno;
phytype <<= 16;
phytype |= mii_send(mk_mii_read(phyno,
PHY_PHYIDR2));
#ifdef ET_DEBUG
printf("PHY @ 0x%x pass %d type ",phyno,pass);
switch (phytype & 0xfffffff0) {
case PHY_ID_LXT970:
printf("LXT970\n");
break;
case PHY_ID_LXT971:
printf("LXT971\n");
break;
case PHY_ID_82555:
printf("82555\n");
break;
case PHY_ID_QS6612:
printf("QS6612\n");
break;
case PHY_ID_AMD79C784:
printf("AMD79C784\n");
break;
case PHY_ID_LSI80225B:
printf("LSI L80225/B\n");
break;
case PHY_ID_DM9161:
printf("Davicom DM9161\n");
break;
default:
printf("0x%08x\n", phytype);
break;
}
#endif
}
}
}
if (phyaddr < 0) {
printf("No PHY device found.\n");
}
return phyaddr;
}
#endif /* CFG_DISCOVER_PHY */
#if (defined(CONFIG_MII) || (CONFIG_COMMANDS & CFG_CMD_MII)) && !defined(CONFIG_BITBANGMII)
/****************************************************************************
* mii_init -- Initialize the MII for MII command without ethernet
* This function is a subset of eth_init
****************************************************************************
*/
void mii_init (void)
{
volatile immap_t *immr = (immap_t *) CFG_IMMR;
volatile fec_t *fecp = &(immr->im_cpm.cp_fec);
int i, j;
for (j = 0; j < sizeof(ether_fcc_info) / sizeof(ether_fcc_info[0]); j++) {
/* Whack a reset.
* A delay is required between a reset of the FEC block and
* initialization of other FEC registers because the reset takes
* some time to complete. If you don't delay, subsequent writes
* to FEC registers might get killed by the reset routine which is
* still in progress.
*/
fecp->fec_ecntrl = FEC_ECNTRL_PINMUX | FEC_ECNTRL_RESET;
for (i = 0;
(fecp->fec_ecntrl & FEC_ECNTRL_RESET) && (i < FEC_RESET_DELAY);
++i) {
udelay (1);
}
if (i == FEC_RESET_DELAY) {
printf ("FEC_RESET_DELAY timeout\n");
return;
}
/* We use strictly polling mode only
*/
fecp->fec_imask = 0;
/* Clear any pending interrupt
*/
fecp->fec_ievent = 0xffc0;
/* Setup the pin configuration of the FEC(s)
*/
fec_pin_init(ether_fcc_info[i].ether_index);
/* Now enable the transmit and receive processing
*/
fecp->fec_ecntrl = FEC_ECNTRL_PINMUX | FEC_ECNTRL_ETHER_EN;
}
}
/*****************************************************************************
* Read and write a MII PHY register, routines used by MII Utilities
*
* FIXME: These routines are expected to return 0 on success, but mii_send
* does _not_ return an error code. Maybe 0xFFFF means error, i.e.
* no PHY connected...
* For now always return 0.
* FIXME: These routines only work after calling eth_init() at least once!
* Otherwise they hang in mii_send() !!! Sorry!
*****************************************************************************/
int fec8xx_miiphy_read(char *devname, unsigned char addr,
unsigned char reg, unsigned short *value)
{
short rdreg; /* register working value */
#ifdef MII_DEBUG
printf ("miiphy_read(0x%x) @ 0x%x = ", reg, addr);
#endif
rdreg = mii_send(mk_mii_read(addr, reg));
*value = rdreg;
#ifdef MII_DEBUG
printf ("0x%04x\n", *value);
#endif
return 0;
}
int fec8xx_miiphy_write(char *devname, unsigned char addr,
unsigned char reg, unsigned short value)
{
short rdreg; /* register working value */
#ifdef MII_DEBUG
printf ("miiphy_write(0x%x) @ 0x%x = ", reg, addr);
#endif
rdreg = mii_send(mk_mii_write(addr, reg, value));
#ifdef MII_DEBUG
printf ("0x%04x\n", value);
#endif
return 0;
}
#endif /* (CONFIG_COMMANDS & CFG_CMD_MII) && !defined(CONFIG_BITBANGMII)*/
#endif /* CFG_CMD_NET, FEC_ENET */