* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
@ 2009-05-04 17:56 Prafulla Wadaskar
2009-05-12 4:03 ` Prafulla Wadaskar
0 siblings, 1 reply; 10+ messages in thread
From: Prafulla Wadaskar @ 2009-05-04 17:56 UTC (permalink / raw)
To: u-boot
Contributors:
Yotam Admon <yotam@marvell.com>
Michael Blostein <michaelbl at marvell.com
Reviewed by: Ronen Shitrit <rshitrit@marvell.com>
Signed-off-by: Prafulla Wadaskar <prafulla@marvell.com>
---
Change log:
v2: entire rewrite/restructure of v1
used small names for variable/function names
readl/writel used to access SoC registers
Soc registers accssed using pointers through net device struct
miiphy registration done for external smi read/write access
miiphy_link used to detect phy link presence
cleaned for cosmetic changes
cpu/arm926ejs/kirkwood/cpu.c | 8 +
drivers/net/Makefile | 1 +
drivers/net/kirkwood_egiga.c | 1669 ++++++++++++++++++++++++++++++++++++++++++
drivers/net/kirkwood_egiga.h | 828 +++++++++++++++++++++
include/netdev.h | 1 +
5 files changed, 2507 insertions(+), 0 deletions(-)
create mode 100644 drivers/net/kirkwood_egiga.c
create mode 100644 drivers/net/kirkwood_egiga.h
diff --git a/cpu/arm926ejs/kirkwood/cpu.c b/cpu/arm926ejs/kirkwood/cpu.c
index bfbbe05..8ed184c 100644
--- a/cpu/arm926ejs/kirkwood/cpu.c
+++ b/cpu/arm926ejs/kirkwood/cpu.c
@@ -293,3 +293,11 @@ int arch_misc_init(void)
return 0;
}
#endif /* CONFIG_ARCH_MISC_INIT */
+
+#ifdef CONFIG_KIRKWOOD_EGIGA
+int cpu_eth_init(bd_t *bis)
+{
+ kirkwood_egiga_initialize(bis);
+ return 0;
+}
+#endif
diff --git a/drivers/net/Makefile b/drivers/net/Makefile
index a360a50..f0c5654 100644
--- a/drivers/net/Makefile
+++ b/drivers/net/Makefile
@@ -40,6 +40,7 @@ COBJS-$(CONFIG_ENC28J60) += enc28j60.o
COBJS-$(CONFIG_FSLDMAFEC) += fsl_mcdmafec.o mcfmii.o
COBJS-$(CONFIG_GRETH) += greth.o
COBJS-$(CONFIG_INCA_IP_SWITCH) += inca-ip_sw.o
+COBJS-$(CONFIG_KIRKWOOD_EGIGA) += kirkwood_egiga.o
COBJS-$(CONFIG_DRIVER_KS8695ETH) += ks8695eth.o
COBJS-$(CONFIG_DRIVER_LAN91C96) += lan91c96.o
COBJS-$(CONFIG_MACB) += macb.o
diff --git a/drivers/net/kirkwood_egiga.c b/drivers/net/kirkwood_egiga.c
new file mode 100644
index 0000000..d9b295a
--- /dev/null
+++ b/drivers/net/kirkwood_egiga.c
@@ -0,0 +1,1669 @@
+/*
+ * (C) Copyright 2009
+ * Marvell Semiconductor <www.marvell.com>
+ * Prafulla Wadaskar <prafulla@marvell.com>
+ *
+ * (C) Copyright 2003
+ * Ingo Assmus <ingo.assmus@keymile.com>
+ *
+ * based on - Driver for MV64360X ethernet ports
+ * Copyright (C) 2002 rabeeh at galileo.co.il
+ *
+ * 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., 51 Franklin Street, Fifth Floor, Boston,
+ * MA 02110-1301 USA
+ */
+
+#include <common.h>
+#include <net.h>
+#include <malloc.h>
+#include <miiphy.h>
+#include <asm/errno.h>
+#include <asm/types.h>
+#include <asm/byteorder.h>
+
+/* In case SRAM is cache coherent or non-cacheable */
+#define CONFIG_NOT_COHERENT_CACHE
+#define D_CACHE_FLUSH_LINE(addr, offset) ;
+#define CPU_PIPE_FLUSH { __asm__ __volatile__ ("nop;nop;nop;nop;nop;nop;nop");}
+
+#include "kirkwood_egiga.h"
+
+/*
+ * smi_reg_read - Read from ethernet phy register.
+ *
+ * INPUT:
+ * @phy_adr - Phy address.
+ * @reg_ofs - Phy register offset.
+ *
+ * This function reads ethernet phy register.
+ * REturns 16bit phy register value, or 0xffff on error
+ */
+int smi_reg_read(char *devname, u8 phy_adr, u8 reg_ofs, u16 * data)
+{
+ struct eth_device *dev = eth_get_dev_by_name(devname);
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct kwgbe_registers *regs = dkwgbe->regs;
+ u32 smi_reg;
+ volatile u32 timeout;
+
+ /* check to read parameters */
+ if (phy_adr == 0xEE && reg_ofs == 0xEE) {
+ data = (void *)(0x00ff & KWGBEREG_RD(regs->phyadr));
+ return 0;
+ }
+ /* check parameters */
+ if ((phy_adr << KWGBE_PHY_SMI_DEV_ADDR_OFFS) &
+ ~KWGBE_PHY_SMI_DEV_ADDR_MASK) {
+ printf("Illegal PHY device address %d\n", phy_adr);
+ return -EFAULT;
+ }
+ if ((reg_ofs << KWGBE_SMI_REG_ADDR_OFFS) & ~KWGBE_SMI_REG_ADDR_MASK) {
+ printf("Illegal PHY register offset %d\n", reg_ofs);
+ return -EFAULT;
+ }
+
+ timeout = KWGBE_PHY_SMI_TIMEOUT;
+ /* wait till the SMI is not busy */
+ do {
+ /* read smi register */
+ smi_reg = KWGBEREG_RD(regs->smi);
+ if (timeout-- == 0) {
+ printf("SMI busy timeout\n");
+ return -EFAULT;
+ }
+ } while (smi_reg & KWGBE_PHY_SMI_BUSY_MASK);
+
+ /* fill the phy address and regiser offset and read opcode */
+ smi_reg =
+ (phy_adr << KWGBE_PHY_SMI_DEV_ADDR_OFFS) | (reg_ofs <<
+ KWGBE_SMI_REG_ADDR_OFFS)
+ | KWGBE_PHY_SMI_OPCODE_READ;
+
+ /* write the smi register */
+ KWGBEREG_WR(regs->smi, smi_reg);
+
+ /*wait till readed value is ready */
+ timeout = KWGBE_PHY_SMI_TIMEOUT;
+ do {
+ /* read smi register */
+ smi_reg = KWGBEREG_RD(regs->smi);
+ if (timeout-- == 0) {
+ printf("SMI read-valid timeout\n");
+ return -EFAULT;
+ }
+ } while (!(smi_reg & KWGBE_PHY_SMI_READ_VALID_MASK));
+
+ /* Wait for the data to update in the SMI register */
+ for (timeout = 0; timeout < KWGBE_PHY_SMI_TIMEOUT; timeout++) ;
+
+ *data = (u16) (KWGBEREG_RD(regs->smi) & KWGBE_PHY_SMI_DATA_MASK);
+
+ debug("Reg(phyadr %d, off %d) Phy-value = %04x\n", phy_adr,
+ reg_ofs, *data);
+
+ return 0;
+}
+
+/*
+ * smi_reg_write - Write to ethernet phy register.
+ *
+ * @phy_adr - Phy address.
+ * @reg_ofs - Phy register offset.
+ * @data - 16bit data.
+ *
+ * This function write to ethernet phy register.
+ * Returns 0 if write succeed, -EINVAL on bad parameters , MV_ERROR on error .
+ * -ETIME on timeout
+ */
+int smi_reg_write(char *devname, u8 phy_adr, u8 reg_ofs, u16 data)
+{
+ struct eth_device *dev = eth_get_dev_by_name(devname);
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct kwgbe_registers *regs = dkwgbe->regs;
+ u32 smi_reg;
+ volatile u32 timeout;
+
+ /* check parameters */
+ if ((phy_adr << KWGBE_PHY_SMI_DEV_ADDR_OFFS) &
+ ~KWGBE_PHY_SMI_DEV_ADDR_MASK) {
+ printf("Illegal phy address\n");
+ return -EINVAL;
+ }
+ if ((reg_ofs << KWGBE_SMI_REG_ADDR_OFFS) & ~KWGBE_SMI_REG_ADDR_MASK) {
+ printf("Illegal register offset\n");
+ return -EINVAL;
+ }
+
+ /* wait till the SMI is not busy */
+ timeout = KWGBE_PHY_SMI_TIMEOUT;
+ do {
+ /* read smi register */
+ smi_reg = KWGBEREG_RD(regs->smi);
+ if (timeout-- == 0) {
+ printf("SMI busy timeout\n");
+ return -ETIME;
+ }
+ } while (smi_reg & KWGBE_PHY_SMI_BUSY_MASK);
+
+ /* fill the phy address and regiser offset and write opcode and data */
+ smi_reg = (data << KWGBE_PHY_SMI_DATA_OFFS);
+ smi_reg |=
+ (phy_adr << KWGBE_PHY_SMI_DEV_ADDR_OFFS) | (reg_ofs <<
+ KWGBE_SMI_REG_ADDR_OFFS);
+ smi_reg &= ~KWGBE_PHY_SMI_OPCODE_READ;
+
+ /* write the smi register */
+ KWGBEREG_WR(regs->smi, smi_reg);
+
+ return 0;
+}
+
+static int egiga_free_tx_rings(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct kwgbe_portinfo *portinfo = &dkwgbe->portinfo;
+ struct kwgbe_registers *regs = dkwgbe->regs;
+
+ u32 queue;
+ volatile struct kwgbe_tx_dscr *p_tx_curr_desc;
+
+ /* Stop Tx Queues */
+ KWGBEREG_WR(regs->tqc, 0x0000ff00);
+
+ /* Free TX rings */
+ debug("Clearing previously allocated TX queues... \n");
+ for (queue = 0; queue < KW_TXQ_NO; queue++) {
+ /* Free on TX rings */
+ for (p_tx_curr_desc = portinfo->p_tx_desc_area_base[queue];
+ ((u32) p_tx_curr_desc <=
+ (u32) portinfo->p_tx_desc_area_base[queue] +
+ portinfo->tx_desc_area_size[queue]);
+ p_tx_curr_desc =
+ (struct kwgbe_tx_dscr *)((u32) p_tx_curr_desc +
+ TX_DESC_ALIGNED_SIZE)) {
+ /* this is inside for loop */
+ if (p_tx_curr_desc->retinfo != 0) {
+ p_tx_curr_desc->retinfo = 0;
+ debug("freed\n");
+ }
+ }
+ debug("Done\n");
+ }
+ return 0;
+}
+
+static int egiga_free_rx_rings(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct kwgbe_portinfo *portinfo = &dkwgbe->portinfo;
+ struct kwgbe_registers *regs = dkwgbe->regs;
+
+ u32 queue;
+ volatile struct kwgbe_rx_dscr *p_rx_curr_desc;
+
+ /* Stop RX Queues */
+ KWGBEREG_WR(regs->rqc, 0x0000ff00);
+
+ /* Free RX rings */
+ debug("Clearing previously allocated RX queues...\n");
+ for (queue = 0; queue < KW_RXQ_NO; queue++) {
+ /* Free preallocated skb's on RX rings */
+ for (p_rx_curr_desc = portinfo->p_rx_desc_area_base[queue];
+ (((u32) p_rx_curr_desc <
+ ((u32) portinfo->p_rx_desc_area_base[queue] +
+ portinfo->rx_desc_area_size[queue])));
+ p_rx_curr_desc =
+ (struct kwgbe_rx_dscr *)((u32) p_rx_curr_desc +
+ RX_DESC_ALIGNED_SIZE)) {
+ if (p_rx_curr_desc->retinfo != 0) {
+ p_rx_curr_desc->retinfo = 0;
+ debug("freed\n");
+ }
+ }
+ debug("Done\n");
+ }
+ return 0;
+}
+
+static void phy_set_addr(struct kwgbe_registers *regs, int phy_addr)
+{
+ KWGBEREG_WR(regs->phyadr, phy_addr);
+}
+
+#ifdef UPDATE_STATS_BY_SOFTWARE
+/*
+ * egiga_print_stat
+ *
+ * Update the statistics structure in the private data structure
+ *
+ * Input : pointer to ethernet interface network device structure
+ * Output : N/A
+ */
+static void egiga_print_stat(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct egiga_priv *portpriv = &dkwgbe->portpriv;
+ struct net_device_stats *stats = &portpriv->stats;
+
+ /* These are false updates */
+ printf("\n### Network statistics: ###\n");
+ printf("--------------------------\n");
+ printf(" Packets received: %d\n", stats->rx_packets);
+ printf(" Packets send: %d\n", stats->tx_packets);
+ printf(" Received bytes: %lld\n", stats->rx_bytes);
+ printf(" Send bytes: %lld\n", stats->tx_bytes);
+ if (stats->rx_errors != 0)
+ printf(" Rx Errors: %d\n",
+ stats->rx_errors);
+ if (stats->rx_dropped != 0)
+ printf(" Rx dropped (CRC Errors): %d\n",
+ stats->rx_dropped);
+ if (stats->multicast != 0)
+ printf(" Rx mulicast frames: %d\n",
+ stats->multicast);
+ if (stats->collisions != 0)
+ printf(" No. of collisions: %d\n",
+ stats->collisions);
+ if (stats->rx_length_errors != 0)
+ printf(" Rx length errors: %d\n",
+ stats->rx_length_errors);
+}
+#endif
+
+/*
+ * kwgbe_stop
+ *
+ * This function is used when closing the network device.
+ * It updates the hardware,
+ * release all memory that holds buffers and descriptors and release the IRQ.
+ * Input : a pointer to the device structure
+ * Output : zero if success , nonzero if fails
+ */
+int kwgbe_stop(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct kwgbe_registers *regs = dkwgbe->regs;
+
+ /* Disable all gigE address decoder */
+ KWGBEREG_WR(regs->bare, 0x3f);
+
+ egiga_free_tx_rings(dev);
+ egiga_free_rx_rings(dev);
+
+ port_reset(regs);
+ /* Disable ethernet port interrupts */
+ KWGBEREG_WR(regs->ic, 0);
+ KWGBEREG_WR(regs->ice, 0);
+ /* Mask RX buffer and TX end interrupt */
+ KWGBEREG_WR(regs->pim, 0);
+ /* Mask phy and link status changes interrupts */
+ KWGBEREG_WR(regs->peim, 0);
+
+ /* Print Network statistics */
+#ifdef UPDATE_STATS_BY_SOFTWARE
+ /*
+ * Print statistics (only if ethernet is running),
+ * then zero all the stats fields in memory
+ */
+ if (portpriv->running == MAGIC_KWGBE_RUNNING) {
+ portpriv->running = 0;
+ egiga_print_stat(dev);
+ }
+ memset(&portpriv->stats, 0, sizeof(struct net_device_stats));
+#endif
+ debug("Ethernet stopped ...\n");
+ return 0;
+}
+
+/*
+ * clear_mib_counters - Clear all MIB counters
+ *
+ * This function clears all MIB counters of a specific ethernet port.
+ * A read from the MIB counter will reset the counter.
+ *
+ * OUTPUT: After reading all MIB counters, the counters resets.
+ * RETURN: MIB counter value.
+ */
+static void clear_mib_counters(u32 mibc)
+{
+ u32 i = mibc + 0x80;
+ u32 dummy;
+
+ /* Perform dummy reads from MIB counters */
+ for (; mibc < i; mibc += 4) {
+ dummy = readl(mibc);
+ }
+}
+
+/*
+ * egiga_update_stat
+ *
+ * Update the statistics structure in the private data structure
+ *
+ * Input : pointer to ethernet interface network device structure
+ * Output : N/A
+ */
+static void egiga_update_stat(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct egiga_priv *portpriv = &dkwgbe->portpriv;
+ struct net_device_stats *stats = &portpriv->stats;
+ struct mib_counters *mibc = &dkwgbe->regs->mibc;
+
+ /* These are false updates */
+ stats->rx_packets += KWGBEREG_RD(mibc->good_frames_received);
+ stats->tx_packets += KWGBEREG_RD(mibc->good_frames_sent);
+ stats->rx_bytes += KWGBEREG_RD(mibc->good_octets_received);
+ stats->tx_bytes += KWGBEREG_RD(mibc->good_octets_sent);
+ stats->rx_errors += KWGBEREG_RD(mibc->mac_receive_error);
+ /* Rx dropped is for received packet with CRC error */
+ stats->rx_dropped += KWGBEREG_RD(mibc->bad_crc_event);
+ stats->multicast += KWGBEREG_RD(mibc->multicast_frames_received);
+ stats->collisions +=
+ KWGBEREG_RD(mibc->collision) + KWGBEREG_RD(mibc->late_collision);
+ stats->rx_length_errors +=
+ KWGBEREG_RD(mibc->undersize_received) +
+ KWGBEREG_RD(mibc->oversize_received);
+}
+
+/*
+ * egiga_get_stats
+ *
+ * Returns a pointer to the interface statistics.
+ *
+ * Input : dev - a pointer to the required interface
+ * Output : a pointer to the interface's statistics
+ */
+static struct net_device_stats *egiga_get_stats(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct egiga_priv *portpriv = &dkwgbe->portpriv;
+
+ egiga_update_stat(dev);
+ return &portpriv->stats;
+}
+
+/*
+ * set_access_control - Config address decode parameters for Ethernet unit
+ *
+ * This function configures the address decode parameters for the Gigabit
+ * Ethernet Controller according the given parameters struct.
+ *
+ * @regs Register struct pointer.
+ * @param Address decode parameter struct.
+ */
+static void set_access_control(struct kwgbe_registers *regs,
+ struct kwgbe_winparam *param)
+{
+ u32 access_prot_reg;
+
+ /* Set access control register */
+ access_prot_reg = KWGBEREG_RD(regs->epap);
+ access_prot_reg &= (~(3 << (param->win * 2))); /* clear window permission */
+ access_prot_reg |= (param->access_ctrl << (param->win * 2));
+ KWGBEREG_WR(regs->epap, access_prot_reg);
+
+ /* Set window Size reg (SR) */
+ KWGBEREG_WR(regs->barsz[param->win].size,
+ (((param->size / 0x10000) - 1) << 16));
+
+ /* Set window Base address reg (BA) */
+ KWGBEREG_WR(regs->barsz[param->win].bar,
+ (param->target | param->attrib | param->base_addr));
+ /* High address remap reg (HARR) */
+ if (param->win < 4)
+ KWGBEREG_WR(regs->ha_remap[param->win], param->high_addr);
+
+ /* Base address enable reg (BARER) */
+ if (param->enable == 1)
+ KWGBEREG_BITS_RESET(regs->bare, (1 << param->win));
+ else
+ KWGBEREG_BITS_SET(regs->bare, (1 << param->win));
+}
+
+static void set_dram_access(struct kwgbe_registers *regs)
+{
+ struct kwgbe_winparam win_param;
+ int i;
+
+ for (i = 0; i < CONFIG_NR_DRAM_BANKS; i++) {
+ /* Set access parameters for DRAM bank i */
+ win_param.win = i; /* Use Ethernet window i */
+ win_param.target = KWGBE_TARGET_DRAM; /* Window target - DDR */
+ win_param.access_ctrl = EWIN_ACCESS_FULL; /* Enable full access */
+ win_param.high_addr = 0;
+ /* Get bank base */
+ win_param.base_addr = kw_sdram_bar(i);
+ win_param.size = kw_sdram_bs(i); /* Get bank size */
+ if (win_param.size == 0)
+ win_param.enable = 0;
+ else
+ win_param.enable = 1; /* Enable the access */
+ switch (i) {
+ case 0:
+ /* Enable DRAM bank 0 */
+ win_param.attrib = EBAR_ATTR_DRAM_CS0;
+ break;
+ case 1:
+ /* Enable DRAM bank 1 */
+ win_param.attrib = EBAR_ATTR_DRAM_CS1;
+ break;
+ case 2:
+ /* Enable DRAM bank 2 */
+ win_param.attrib = EBAR_ATTR_DRAM_CS2;
+ break;
+ case 3:
+ /* Enable DRAM bank 3 */
+ win_param.attrib = EBAR_ATTR_DRAM_CS3;
+ break;
+ default:
+ /* invalide bank, disable access */
+ win_param.enable = 0;
+ win_param.attrib = 0;
+ break;
+ }
+#ifndef CONFIG_NOT_COHERENT_CACHE
+ win_param.attrib |= EBAR_ATTR_DRAM_CACHE_COHERENCY_WB;
+#endif
+ /* Set the access control for address window (EPAPR) READ & WRITE */
+ set_access_control(regs, &win_param);
+ }
+}
+
+/*
+ * port_init_mac_tables - Clear all entrance in the UC, SMC and OMC tables
+ *
+ * Go through all the DA filter tables (Unicast, Special Multicast & Other
+ * Multicast) and set each entry to 0.
+ *
+ */
+static void port_init_mac_tables(struct kwgbe_registers *regs)
+{
+ int table_index;
+
+ /* Clear DA filter unicast table (Ex_dFUT) */
+ for (table_index = 0; table_index < 4; ++table_index)
+ KWGBEREG_WR(regs->dfut[table_index], 0);
+
+ for (table_index = 0; table_index < 64; ++table_index) {
+ /* Clear DA filter special multicast table (Ex_dFSMT) */
+ KWGBEREG_WR(regs->dfsmt[table_index], 0);
+ /* Clear DA filter other multicast table (Ex_dFOMT) */
+ KWGBEREG_WR(regs->dfomt[table_index], 0);
+ }
+}
+
+/*
+ * port_init - Initialize the Ethernet port driver
+ *
+ * @p_gbe_pctrl - ptr to Ethernet port control struct
+ *
+ * This function prepares the ethernet port to start its activity:
+ * 1) Completes the ethernet port driver struct initialization toward port
+ * start routine.
+ * 2) Resets the device to a quiescent state in case of warm reboot.
+ * 3) Enable SDMA access to all four DRAM banks as well as internal SRAM.
+ * 4) Clean MAC tables. The reset status of those tables is unknown.
+ * 5) Set PHY address.
+ * Note: Call this routine prior to port_start routine and after setting
+ * user values in the user fields of Ethernet port control struct (i.e.
+ * port_phy_addr).
+ *
+ */
+static void port_init(struct kwgbe_device *dkwgbe)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ struct kwgbe_registers *regs = dkwgbe->regs;
+ int queue;
+
+ p_gbe_pctrl->port_config = PRT_CFG_VAL;
+ p_gbe_pctrl->port_config_extend = PORT_CFG_EXTEND_VALUE;
+ p_gbe_pctrl->port_sdma_config = PORT_SDMA_CFG_VALUE;
+ p_gbe_pctrl->port_serial_control = PORT_SERIAL_CONTROL_VALUE;
+
+ p_gbe_pctrl->port_rx_queue_command = 0;
+ p_gbe_pctrl->port_tx_queue_command = 0;
+
+ /* Stop RX Queues */
+ KWGBEREG_WR(regs->rqc, 0x0000ff00);
+
+ /* Clear the ethernet port interrupts */
+ KWGBEREG_WR(regs->ic, 0);
+ KWGBEREG_WR(regs->ice, 0);
+
+ /* Unmask RX buffer and TX end interrupt */
+ KWGBEREG_WR(regs->pim, INT_CAUSE_UNMASK_ALL);
+
+ /* Unmask phy and link status changes interrupts */
+ KWGBEREG_WR(regs->peim, INT_CAUSE_UNMASK_ALL_EXT);
+
+ /* Zero out SW structs */
+ for (queue = 0; queue < MAX_RXQ_NO; queue++) {
+ CURR_RFD_SET((struct kwgbe_rx_dscr *)0x00000000, queue);
+ USED_RFD_SET((struct kwgbe_rx_dscr *)0x00000000, queue);
+ p_gbe_pctrl->rx_resource_err[queue] = 0;
+ }
+
+ for (queue = 0; queue < MAX_TXQ_NO; queue++) {
+ CURR_TFD_SET((struct kwgbe_tx_dscr *)0x00000000, queue);
+ USED_TFD_SET((struct kwgbe_tx_dscr *)0x00000000, queue);
+ FIRST_TFD_SET((struct kwgbe_tx_dscr *)0x00000000, queue);
+ p_gbe_pctrl->tx_resource_err[queue] = 0;
+ }
+
+ port_reset(regs);
+ set_dram_access(regs);
+ port_init_mac_tables(regs);
+}
+
+/*
+ * port_start - Start the Ethernet port activity.
+ *
+ * This routine prepares the Ethernet port for Rx and Tx activity:
+ * 1. Initialize Tx and Rx Current Descriptor Pointer for each queue that
+ * has been initialized a descriptor's ring (using kwgbe_init_tx_desc_ring
+ * for Tx and kwgbe_init_rx_desc_ring for Rx)
+ * 2. Initialize and enable the Ethernet configuration port by writing to
+ * the port's configuration and command registers.
+ * 3. Initialize and enable the SDMA by writing to the SDMA's
+ * configuration and command registers.
+ * After completing these steps, the ethernet port SDMA can starts to
+ * perform Rx and Tx activities.
+ *
+ * Note: Each Rx and Tx queue descriptor's list must be initialized prior
+ * to calling this function (use kwgbe_init_tx_desc_ring for Tx queues and
+ * kwgbe_init_rx_desc_ring for Rx queues).
+ *
+ * OUTPUT: Ethernet port is ready to receive and transmit.
+ *
+ * RETURN: 0 if the port PHY is not up. 1 otherwise.
+ */
+static void port_start(struct kwgbe_device *dkwgbe)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ struct kwgbe_registers *regs = dkwgbe->regs;
+ int queue;
+ volatile struct kwgbe_tx_dscr *p_tx_curr_desc;
+ volatile struct kwgbe_rx_dscr *p_rx_curr_desc;
+
+ /* Assignment of Tx CTRP of given queue */
+ for (queue = 0; queue < MAX_TXQ_NO; queue++) {
+ CURR_TFD_GET(p_tx_curr_desc, queue);
+ KWGBEREG_WR(regs->tcqdp[queue], (u32) p_tx_curr_desc);
+ }
+
+ /* Assignment of Rx CRDB of given queue */
+ for (queue = 0; queue < MAX_RXQ_NO; queue++) {
+ CURR_RFD_GET(p_rx_curr_desc, queue);
+ KWGBEREG_WR(regs->crdp[queue], (u32) p_rx_curr_desc);
+
+ if (p_rx_curr_desc != NULL)
+ /* Add the assigned Ethernet address to the port's address table */
+ port_uc_addr_set(regs, p_gbe_pctrl->port_mac_addr,
+ queue);
+ }
+
+ /* Assign port configuration and command. */
+ KWGBEREG_WR(regs->pxc, p_gbe_pctrl->port_config);
+ KWGBEREG_WR(regs->pxcx, p_gbe_pctrl->port_config_extend);
+ KWGBEREG_WR(regs->psc0, p_gbe_pctrl->port_serial_control);
+ KWGBEREG_BITS_SET(regs->psc0, KWGBE_SERIAL_PORT_EN);
+
+ /* Assign port SDMA configuration */
+ KWGBEREG_WR(regs->sdc, p_gbe_pctrl->port_sdma_config);
+ KWGBEREG_WR(regs->tqx[0].qxttbc, 0x3fffffff);
+ KWGBEREG_WR(regs->tqx[0].tqxtbc, 0x03fffcff);
+ /* Turn off the port/queue bandwidth limitation */
+ KWGBEREG_WR(regs->pmtu, 0x0);
+
+ /* Enable port Rx. */
+ KWGBEREG_WR(regs->rqc, p_gbe_pctrl->port_rx_queue_command);
+
+ /* Set maximum receive buffer to 9700 bytes */
+ KWGBEREG_WR(regs->psc0,
+ (0x5 << 17) | (KWGBEREG_RD(regs->psc0) & 0xfff1ffff));
+
+ /*
+ * Set ethernet MTU for leaky bucket mechanism to 0 - this will
+ * disable the leaky bucket mechanism .
+ */
+ KWGBEREG_WR(regs->pmtu, 0);
+}
+
+/*
+ * port_uc_addr - This function Set the port unicast address table
+ *
+ * This function locates the proper entry in the Unicast table for the
+ * specified MAC nibble and sets its properties according to function
+ * parameters.
+ * This function add/removes MAC addresses from the port unicast address
+ * table.
+ *
+ * @uc_nibble Unicast MAC Address last nibble.
+ * @queue Rx queue number for this MAC address.
+ * @option 0 = Add, 1 = remove address.
+ *
+ * RETURN: 1 if output succeeded. 0 if option parameter is invalid.
+ */
+static int port_uc_addr(struct kwgbe_registers *regs, u8 uc_nibble,
+ enum kwgbe_q queue, int option)
+{
+ u32 unicast_reg;
+ u32 tbl_offset;
+ u32 reg_offset;
+
+ /* Locate the Unicast table entry */
+ uc_nibble = (0xf & uc_nibble);
+ tbl_offset = (uc_nibble / 4); /* Register offset from unicast table base */
+ reg_offset = uc_nibble % 4; /* Entry offset within the above register */
+
+ switch (option) {
+ case REJECT_MAC_ADDR:
+ /* Clear accepts frame bit at specified unicast DA table entry */
+ unicast_reg = KWGBEREG_RD(regs->dfut[tbl_offset]);
+ unicast_reg &= (0xFF << (8 * reg_offset));
+ KWGBEREG_WR(regs->dfut[tbl_offset], unicast_reg);
+ break;
+ case ACCEPT_MAC_ADDR:
+ /* Set accepts frame bit at unicast DA filter table entry */
+ unicast_reg = KWGBEREG_RD(regs->dfut[tbl_offset]);
+ unicast_reg &= (0xFF << (8 * reg_offset));
+ unicast_reg |= ((0x01 | (queue << 1)) << (8 * reg_offset));
+ KWGBEREG_WR(regs->dfut[tbl_offset], unicast_reg);
+ break;
+ default:
+ return 0;
+ }
+ return 1;
+}
+
+/*
+ * port_uc_addr_set - This function Set the port Unicast address.
+ *
+ * This function Set the port Ethernet MAC address.
+ *
+ * @p_addr Address to be set
+ * @queue Rx queue number for this MAC address.
+ *
+ * OUTPUT: Set MAC address low and high registers. also calls
+ * port_uc_addr() To set the unicast table with the proper info.
+ */
+static void port_uc_addr_set(struct kwgbe_registers *regs, u8 * p_addr,
+ enum kwgbe_q queue)
+{
+ u32 mac_h;
+ u32 mac_l;
+
+ mac_l = (p_addr[4] << 8) | (p_addr[5]);
+ mac_h =
+ (p_addr[0] << 24) | (p_addr[1] << 16) | (p_addr[2] << 8) |
+ (p_addr[3] << 0);
+
+ KWGBEREG_WR(regs->macal, mac_l);
+ KWGBEREG_WR(regs->macah, mac_h);
+
+ /* Accept frames of this address */
+ port_uc_addr(regs, p_addr[5], queue, ACCEPT_MAC_ADDR);
+}
+
+/*
+ * port_reset - Reset Ethernet port
+ *
+ * This routine resets the chip by aborting any SDMA engine activity and
+ * clearing the MIB counters. The Receiver and the Transmit unit are in
+ * idle state after this command is performed and the port is disabled.
+ *
+ */
+static void port_reset(struct kwgbe_registers *regs)
+{
+ u32 reg_data, i;
+
+ /* Stop Tx port activity. Check port Tx activity. */
+ reg_data = KWGBEREG_RD(regs->tqc);
+
+ if (reg_data & 0xFF) {
+ /* Issue stop command for active channels only */
+ KWGBEREG_WR(regs->tqc, (reg_data << 8));
+
+ /* Wait for all Tx activity to terminate. */
+ do {
+ /* Check port cause register that all Tx queues are stopped */
+ reg_data = KWGBEREG_RD(regs->tqc);
+ }
+ while (reg_data & 0xFF);
+ }
+
+ /* Stop Rx port activity. Check port Rx activity. */
+ reg_data = KWGBEREG_RD(regs->rqc);
+
+ if (reg_data & 0xFF) {
+ /* Issue stop command for active channels only */
+ KWGBEREG_WR(regs->rqc, (reg_data << 8));
+
+ /* Wait for all Rx activity to terminate. */
+ do {
+ /* Check port cause register that all Rx queues are stopped */
+ reg_data = KWGBEREG_RD(regs->rqc);
+ }
+ while (reg_data & 0xFF);
+ }
+
+ /* Clear all MIB counters */
+ clear_mib_counters((u32) & regs->mibc);
+
+ /* Enable port in the Configuration Register */
+ KWGBEREG_BITS_RESET(regs->psc0, KWGBE_SERIAL_PORT_EN);
+ /* Set port of active in the Configuration Register */
+ KWGBEREG_BITS_RESET(regs->psc1, 1 << 4);
+#ifdef CONFIG_SYS_MII_MODE
+ /* Set MMI interface up */
+ KWGBEREG_BITS_RESET(regs->psc1, 1 << 3);
+#endif
+ for (i = 0; i < 4000; i++) ;
+ return;
+}
+
+/*
+ * kwgbe_init_rx_desc_ring - Curve a Rx chain desc list and buffer in memory.
+ *
+ * This function prepares a Rx chained list of descriptors and packet
+ * buffers in a form of a ring. The routine must be called after port
+ * initialization routine and before port start routine.
+ * The Ethernet SDMA engine uses CPU bus addresses to access the various
+ * devices in the system (i.e. DRAM). This function uses the ethernet
+ * struct 'virtual to physical' routine (set by the user) to set the ring
+ * with physical addresses.
+ *
+ * @rx_queue Number of Rx queue.
+ * @rx_desc_num Number of Rx descriptors
+ * @rx_buff_size Size of Rx buffer
+ * @rx_desc_base_addr Rx descriptors memory area base addr.
+ * @rx_buff_base_addr Rx buffer memory area base addr.
+ *
+ * The routine updates the Ethernet port control struct with information
+ * regarding the Rx descriptors and buffers.
+ *
+ * RETURN:
+ * 0 if the given descriptors memory area is not aligned according to
+ * Ethernet SDMA specifications.
+ * 1 otherwise.
+ */
+static int kwgbe_init_rx_desc_ring(struct kwgbe_device *dkwgbe,
+ enum kwgbe_q rx_queue,
+ int rx_desc_num,
+ int rx_buff_size,
+ u32 rx_desc_base_addr, u32 rx_buff_base_addr)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ struct kwgbe_rx_dscr *p_rx_desc;
+ /* pointer to link with the last descriptor */
+ struct kwgbe_rx_dscr *p_rx_prev_desc;
+ u32 buffer_addr;
+ int ix; /* a counter */
+
+ p_rx_desc = (struct kwgbe_rx_dscr *)rx_desc_base_addr;
+ p_rx_prev_desc = p_rx_desc;
+ buffer_addr = rx_buff_base_addr;
+
+ /* Rx desc Must be 4LW aligned (i.e. Descriptor_Address[3:0]=0000). */
+ if (rx_buff_base_addr & 0xF)
+ return 0;
+
+ /* Rx buffers are limited to 64K bytes and Minimum size is 8 bytes */
+ if ((rx_buff_size < 8) || (rx_buff_size > RX_BUF_MXSZ))
+ return 0;
+
+ /* Rx buffers must be 64-bit aligned. */
+ if ((rx_buff_base_addr + rx_buff_size) & 0x7)
+ return 0;
+
+ /* initialize the Rx descriptors ring */
+ for (ix = 0; ix < rx_desc_num; ix++) {
+ p_rx_desc->buf_size = rx_buff_size;
+ p_rx_desc->byte_cnt = 0x0000;
+ p_rx_desc->cmd_sts =
+ KWGBE_BUFFER_OWNED_BY_DMA | KWGBE_RX_EN_INTERRUPT;
+ p_rx_desc->nxtdesc_p = ((u32) p_rx_desc) + RX_DESC_ALIGNED_SIZE;
+ p_rx_desc->buf_ptr = buffer_addr;
+ p_rx_desc->retinfo = 0x00000000;
+ D_CACHE_FLUSH_LINE(p_rx_desc, 0);
+ buffer_addr += rx_buff_size;
+ p_rx_prev_desc = p_rx_desc;
+ p_rx_desc = (struct kwgbe_rx_dscr *)
+ ((u32) p_rx_desc + RX_DESC_ALIGNED_SIZE);
+ }
+
+ /* Closing Rx descriptors ring */
+ p_rx_prev_desc->nxtdesc_p = (rx_desc_base_addr);
+ D_CACHE_FLUSH_LINE(p_rx_prev_desc, 0);
+
+ /* Save Rx desc pointer to driver struct. */
+ CURR_RFD_SET((struct kwgbe_rx_dscr *)rx_desc_base_addr, rx_queue);
+ USED_RFD_SET((struct kwgbe_rx_dscr *)rx_desc_base_addr, rx_queue);
+
+ p_gbe_pctrl->p_rx_desc_area_base[rx_queue] =
+ (struct kwgbe_rx_dscr *)rx_desc_base_addr;
+ p_gbe_pctrl->rx_desc_area_size[rx_queue] =
+ rx_desc_num * RX_DESC_ALIGNED_SIZE;
+
+ p_gbe_pctrl->port_rx_queue_command |= (1 << rx_queue);
+
+ return 1;
+}
+
+/*
+ * kwgbe_init_tx_desc_ring - Curve a Tx chain desc list and buffer in memory.
+ *
+ * This function prepares a Tx chained list of descriptors and packet
+ * buffers in a form of a ring. The routine must be called after port
+ * initialization routine and before port start routine.
+ * The Ethernet SDMA engine uses CPU bus addresses to access the various
+ * devices in the system (i.e. DRAM). This function uses the ethernet
+ * struct 'virtual to physical' routine (set by the user) to set the ring
+ * with physical addresses.
+ *
+ * @p_gbe_pctrl Ethernet Port Control srtuct ptr.
+ * @tx_queue Number of Tx queue.
+ * @tx_desc_num Number of Tx descriptors
+ * @tx_buff_size Size of Tx buffer
+ * @tx_desc_base_addr Tx descriptors memory area base addr.
+ * @tx_buff_base_addr Tx buffer memory area base addr.
+ *
+ * The routine updates the Ethernet port control struct with information
+ * regarding the Tx descriptors and buffers.
+ *
+ * RETURN:
+ * 0 if the given descriptors memory area is not aligned according to
+ * Ethernet SDMA specifications.
+ * 1 otherwise.
+ */
+static int kwgbe_init_tx_desc_ring(struct kwgbe_device *dkwgbe,
+ enum kwgbe_q tx_queue,
+ int tx_desc_num,
+ int tx_buff_size,
+ u32 tx_desc_base_addr, u32 tx_buff_base_addr)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ struct kwgbe_tx_dscr *p_tx_desc;
+ struct kwgbe_tx_dscr *p_tx_prev_desc;
+ u32 buffer_addr;
+ int ix; /* a counter */
+
+ /* save the first desc pointer to link with the last descriptor */
+ p_tx_desc = (struct kwgbe_tx_dscr *)tx_desc_base_addr;
+ p_tx_prev_desc = p_tx_desc;
+ buffer_addr = tx_buff_base_addr;
+
+ /* Tx desc Must be 4LW aligned (i.e. Descriptor_Address[3:0]=0000). */
+ if (tx_buff_base_addr & 0xF)
+ return 0;
+
+ /* Tx buffers are limited to 64K bytes and Minimum size is 8 bytes */
+ if ((tx_buff_size > TX_BUF_MXSZ)
+ || (tx_buff_size < TX_BUF_MNSZ))
+ return 0;
+
+ /* Initialize the Tx descriptors ring */
+ for (ix = 0; ix < tx_desc_num; ix++) {
+ p_tx_desc->byte_cnt = 0x0000;
+ p_tx_desc->l4i_chk = 0x0000;
+ p_tx_desc->cmd_sts = 0x00000000;
+ p_tx_desc->nxtdesc_p = ((u32) p_tx_desc) + TX_DESC_ALIGNED_SIZE;
+
+ p_tx_desc->buf_ptr = buffer_addr;
+ p_tx_desc->retinfo = 0x00000000;
+ D_CACHE_FLUSH_LINE(p_tx_desc, 0);
+ buffer_addr += tx_buff_size;
+ p_tx_prev_desc = p_tx_desc;
+ p_tx_desc = (struct kwgbe_tx_dscr *)
+ ((u32) p_tx_desc + TX_DESC_ALIGNED_SIZE);
+
+ }
+ /* Closing Tx descriptors ring */
+ p_tx_prev_desc->nxtdesc_p = tx_desc_base_addr;
+ D_CACHE_FLUSH_LINE(p_tx_prev_desc, 0);
+ /* Set Tx desc pointer in driver struct. */
+ CURR_TFD_SET((struct kwgbe_tx_dscr *)tx_desc_base_addr, tx_queue);
+ USED_TFD_SET((struct kwgbe_tx_dscr *)tx_desc_base_addr, tx_queue);
+
+ /* Init Tx ring base and size parameters */
+ p_gbe_pctrl->p_tx_desc_area_base[tx_queue] =
+ (struct kwgbe_tx_dscr *)tx_desc_base_addr;
+ p_gbe_pctrl->tx_desc_area_size[tx_queue] =
+ (tx_desc_num * TX_DESC_ALIGNED_SIZE);
+
+ /* Add the queue to the list of Tx queues of this port */
+ p_gbe_pctrl->port_tx_queue_command |= (1 << tx_queue);
+
+ return 1;
+}
+
+/*
+ * b_copy - Copy bytes from source to destination
+ *
+ * This function supports the eight bytes limitation on Tx buffer size.
+ * The routine will zero eight bytes starting from the destination address
+ * followed by copying bytes from the source address to the destination.
+ *
+ * @src_addr 32 bit source address.
+ * @dst_add 32 bit destination address.
+ * @byte_count Number of bytes to copy.
+ */
+static void b_copy(u32 src_addr, u32 dst_addr, int byte_count)
+{
+ /* Zero the dst_addr area */
+ *(u32 *) dst_addr = 0x0;
+
+ while (byte_count != 0) {
+ *(char *)dst_addr = *(char *)src_addr;
+ dst_addr++;
+ src_addr++;
+ byte_count--;
+ }
+}
+
+/*
+ * port_send - Send an Ethernet packet
+ *
+ * This routine send a given packet described by p_pktinfo parameter. It
+ * supports transmitting of a packet spaned over multiple buffers. The
+ * routine updates 'curr' and 'first' indexes according to the packet
+ * segment passed to the routine. In case the packet segment is first,
+ * the 'first' index is update. In any case, the 'curr' index is updated.
+ * If the routine get into Tx resource error it assigns 'curr' index as
+ * 'first'. This way the function can abort Tx process of multiple
+ * descriptors per packet.
+ *
+ * @tx_queue Number of Tx queue.
+ * @p_pkt_info User packet buffer ptr.
+ *
+ * OUTPUT: Tx ring 'curr' and 'first' indexes are updated.
+ *
+ * RETURN:
+ * KWGBE_QFULL in case of Tx resource error.
+ * KWGBE_ERROR in case the routine can not access Tx desc ring.
+ * KWGBE_QLAST_RESO if the routine uses the last Tx resource.
+ * KWGBE_OK otherwise.
+ */
+static enum kwgbe_fret_sts port_send(struct kwgbe_device *dkwgbe,
+ enum kwgbe_q tx_queue,
+ struct kwgbe_pktinf *p_pkt_info)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ struct kwgbe_registers *regs = dkwgbe->regs;
+ volatile struct kwgbe_tx_dscr *p_tx_desc_first;
+ volatile struct kwgbe_tx_dscr *p_tx_desc_curr;
+ volatile struct kwgbe_tx_dscr *p_tx_next_desc_curr;
+ volatile struct kwgbe_tx_dscr *p_tx_desc_used;
+ u32 command_status;
+
+#ifdef CONFIG_TX_PKT_DISPLAY
+ {
+ u16 pcnt = p_pkt_info->byte_cnt;
+ u8 *prnt = (u8 *) p_pkt_info->buf_ptr;
+ printf("cnt=%d,", pcnt);
+ while (pcnt) {
+ printf("%02x,", prnt[0]);
+ prnt++;
+ pcnt--;
+ }
+ printf(" pckend\n");
+ }
+#endif
+ /* Do not process Tx ring in case of Tx ring resource error */
+ if (p_gbe_pctrl->tx_resource_err[tx_queue] == 1)
+ return KWGBE_QFULL;
+
+ /* Get the Tx Desc ring indexes */
+ CURR_TFD_GET(p_tx_desc_curr, tx_queue);
+ USED_TFD_GET(p_tx_desc_used, tx_queue);
+
+ if (p_tx_desc_curr == NULL)
+ return KWGBE_ERROR;
+
+ /* The following parameters are used to save readings from memory */
+ p_tx_next_desc_curr = TX_NEXT_DESC_PTR(p_tx_desc_curr, tx_queue);
+ command_status =
+ p_pkt_info->cmd_sts | KWGBE_ZERO_PADDING | KWGBE_GEN_CRC;
+
+ if (command_status & (KWGBE_TX_FIRST_DESC)) {
+ /* Update first desc */
+ FIRST_TFD_SET(p_tx_desc_curr, tx_queue);
+ p_tx_desc_first = p_tx_desc_curr;
+ } else {
+ FIRST_TFD_GET(p_tx_desc_first, tx_queue);
+ command_status |= KWGBE_BUFFER_OWNED_BY_DMA;
+ }
+
+ /*
+ * Buffers with a payload smaller than 8 bytes must be aligned to 64-bit
+ * boundary. We use the memory allocated for Tx descriptor. This memory
+ * located in TX_BUF_OFFSET_IN_DESC offset within the Tx descriptor.
+ */
+ if (p_pkt_info->byte_cnt <= 8) {
+ printf("You have failed in the < 8 bytes errata - fixme\n");
+ return KWGBE_ERROR;
+
+ p_tx_desc_curr->buf_ptr =
+ (u32) p_tx_desc_curr + TX_BUF_OFFSET_IN_DESC;
+ b_copy(p_pkt_info->buf_ptr, p_tx_desc_curr->buf_ptr,
+ p_pkt_info->byte_cnt);
+ } else
+ p_tx_desc_curr->buf_ptr = p_pkt_info->buf_ptr;
+
+ p_tx_desc_curr->byte_cnt = p_pkt_info->byte_cnt;
+ p_tx_desc_curr->retinfo = p_pkt_info->retinfo;
+
+ if (p_pkt_info->cmd_sts & (KWGBE_TX_LAST_DESC)) {
+ /* Set last desc with DMA ownership and interrupt enable. */
+ p_tx_desc_curr->cmd_sts = command_status |
+ KWGBE_BUFFER_OWNED_BY_DMA | KWGBE_TX_EN_INTERRUPT;
+
+ if (p_tx_desc_curr != p_tx_desc_first)
+ p_tx_desc_first->cmd_sts |= KWGBE_BUFFER_OWNED_BY_DMA;
+
+ /* Flush CPU pipe */
+ D_CACHE_FLUSH_LINE((u32) p_tx_desc_curr, 0);
+ D_CACHE_FLUSH_LINE((u32) p_tx_desc_first, 0);
+ CPU_PIPE_FLUSH;
+
+ /* Apply send command */
+ KWGBEREG_WR(regs->tqc, (1 << tx_queue));
+ debug("packet xmitted\n");
+
+ /*
+ * Finish Tx packet.
+ * Update first desc in case of Tx resource error
+ */
+ p_tx_desc_first = p_tx_next_desc_curr;
+ FIRST_TFD_SET(p_tx_desc_first, tx_queue);
+
+ } else {
+ p_tx_desc_curr->cmd_sts = command_status;
+ D_CACHE_FLUSH_LINE((u32) p_tx_desc_curr, 0);
+ }
+
+ /* Check for ring index overlap in the Tx desc ring */
+ if (p_tx_next_desc_curr == p_tx_desc_used) {
+ /* Update the current descriptor */
+ CURR_TFD_SET(p_tx_desc_first, tx_queue);
+
+ p_gbe_pctrl->tx_resource_err[tx_queue] = 1;
+ return KWGBE_QLAST_RESO;
+ } else {
+ /* Update the current descriptor */
+ CURR_TFD_SET(p_tx_next_desc_curr, tx_queue);
+ return KWGBE_OK;
+ }
+}
+
+/*
+ * tx_return_desc - Free all used Tx descriptors
+ *
+ * This routine returns the transmitted packet information to the caller.
+ * It uses the 'first' index to support Tx desc return in case a transmit
+ * of a packet spanned over multiple buffer still in process.
+ * In case the Tx queue was in "resource error" condition, where there are
+ * no available Tx resources, the function resets the resource error flag.
+ *
+ * @p_gbe_pctrl Ethernet Port Control srtuct ptr.
+ * @tx_queue Number of Tx queue.
+ * @p_pkt_info User packet buffer ptr.
+ *
+ * OUTPUT:Tx ring 'first' and 'used' indexes are updated.
+ *
+ * RETURN:
+ * KWGBE_ERROR in case the routine can not access Tx desc ring.
+ * KWGBE_RETRY in case there is transmission in process.
+ * KWGBE_END_OF_JOB if the routine has nothing to release.
+ * KWGBE_OK otherwise.
+ */
+static enum kwgbe_fret_sts tx_return_desc(struct kwgbe_device *dkwgbe,
+ enum kwgbe_q tx_queue,
+ struct kwgbe_pktinf *p_pkt_info)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ volatile struct kwgbe_tx_dscr *p_tx_desc_used = NULL;
+ volatile struct kwgbe_tx_dscr *p_tx_desc_first = NULL;
+ u32 command_status;
+
+ /* Get the Tx Desc ring indexes */
+ USED_TFD_GET(p_tx_desc_used, tx_queue);
+ FIRST_TFD_GET(p_tx_desc_first, tx_queue);
+
+ /* Sanity check */
+ if (p_tx_desc_used == NULL)
+ return KWGBE_ERROR;
+
+ command_status = p_tx_desc_used->cmd_sts;
+
+ /* Still transmitting... */
+ if (command_status & (KWGBE_BUFFER_OWNED_BY_DMA)) {
+ D_CACHE_FLUSH_LINE((u32) p_tx_desc_used, 0);
+ return KWGBE_RETRY;
+ }
+
+ /* Stop release. About to overlap the current available Tx descriptor */
+ if ((p_tx_desc_used == p_tx_desc_first) &&
+ (p_gbe_pctrl->tx_resource_err[tx_queue] == 0)) {
+ D_CACHE_FLUSH_LINE((u32) p_tx_desc_used, 0);
+ return KWGBE_END_OF_JOB;
+ }
+
+ /* Pass the packet information to the caller */
+ p_pkt_info->cmd_sts = command_status;
+ p_pkt_info->retinfo = p_tx_desc_used->retinfo;
+ p_tx_desc_used->retinfo = 0;
+
+ /* Update the next descriptor to release. */
+ USED_TFD_SET(TX_NEXT_DESC_PTR(p_tx_desc_used, tx_queue), tx_queue);
+
+ /* Any Tx return cancels the Tx resource error status */
+ if (p_gbe_pctrl->tx_resource_err[tx_queue] == 1)
+ p_gbe_pctrl->tx_resource_err[tx_queue] = 0;
+
+ D_CACHE_FLUSH_LINE((u32) p_tx_desc_used, 0);
+
+ return KWGBE_OK;
+}
+
+/*
+ * port_receive - Get received information from Rx ring.
+ *
+ * This routine returns the received data to the caller. There is no
+ * data copying during routine operation. All information is returned
+ * using pointer to packet information struct passed from the caller.
+ * If the routine exhausts Rx ring resources then the resource error flag
+ * is set.
+ *
+ * @p_gbe_pctrl Ethernet Port Control srtuct ptr.
+ * @rx_queue Number of Rx queue.
+ * @p_pkt_info User packet buffer.
+ *
+ * OUTPUT: Rx ring current and used indexes are updated.
+ *
+ * RETURN:
+ * KWGBE_ERROR in case the routine can not access Rx desc ring.
+ * KWGBE_QFULL if Rx ring resources are exhausted.
+ * KWGBE_END_OF_JOB if there is no received data.
+ * KWGBE_OK otherwise.
+ */
+static enum kwgbe_fret_sts port_receive(struct kwgbe_device *dkwgbe,
+ enum kwgbe_q rx_queue,
+ struct kwgbe_pktinf *p_pkt_info)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ volatile struct kwgbe_rx_dscr *p_rx_curr_desc;
+ volatile struct kwgbe_rx_dscr *p_rx_next_curr_desc;
+ volatile struct kwgbe_rx_dscr *p_rx_used_desc;
+ u32 command_status;
+
+ /* Do not process Rx ring in case of Rx ring resource error */
+ if (p_gbe_pctrl->rx_resource_err[rx_queue] == 1) {
+ printf("Rx Queue is full ...\n");
+ return KWGBE_QFULL;
+ }
+
+ /* Get the Rx Desc ring 'curr and 'used' indexes */
+ CURR_RFD_GET(p_rx_curr_desc, rx_queue);
+ USED_RFD_GET(p_rx_used_desc, rx_queue);
+
+ /* Sanity check */
+ if (p_rx_curr_desc == NULL)
+ return KWGBE_ERROR;
+
+ /* The following parameters are used to save readings from memory */
+ p_rx_next_curr_desc = RX_NEXT_DESC_PTR(p_rx_curr_desc, rx_queue);
+ command_status = p_rx_curr_desc->cmd_sts;
+
+ /* Nothing to receive... */
+ if (command_status & (KWGBE_BUFFER_OWNED_BY_DMA)) {
+ /*debug("Rx: command_status: %08x\n", command_status); */
+ D_CACHE_FLUSH_LINE((u32) p_rx_curr_desc, 0);
+ /*debug("KWGBE_END_OF_JOB ...\n"); */
+ return KWGBE_END_OF_JOB;
+ }
+
+ p_pkt_info->byte_cnt = (p_rx_curr_desc->byte_cnt) - RX_BUF_OFFSET;
+ p_pkt_info->cmd_sts = command_status;
+ p_pkt_info->buf_ptr = (p_rx_curr_desc->buf_ptr) + RX_BUF_OFFSET;
+ p_pkt_info->retinfo = p_rx_curr_desc->retinfo;
+ p_pkt_info->l4i_chk = p_rx_curr_desc->buf_size; /* IP fragment indicator */
+
+ /* Clean the return info field to indicate that the packet has been */
+ /* moved to the upper layers */
+ p_rx_curr_desc->retinfo = 0;
+
+ /* Update 'curr' in data structure */
+ CURR_RFD_SET(p_rx_next_curr_desc, rx_queue);
+
+ /* Rx descriptors resource exhausted. Set the Rx ring resource error flag */
+ if (p_rx_next_curr_desc == p_rx_used_desc)
+ p_gbe_pctrl->rx_resource_err[rx_queue] = 1;
+
+ D_CACHE_FLUSH_LINE((u32) p_rx_curr_desc, 0);
+ CPU_PIPE_FLUSH;
+ return KWGBE_OK;
+}
+
+/*
+ * rx_return_buff - Returns a Rx buffer back to the Rx ring.
+ *
+ * This routine returns a Rx buffer back to the Rx ring. It retrieves the
+ * next 'used' descriptor and attached the returned buffer to it.
+ * In case the Rx ring was in "resource error" condition, where there are
+ * no available Rx resources, the function resets the resource error flag.
+ *
+ * @p_gbe_pctrl Ethernet Port Control srtuct ptr.
+ * @rx_queue Number of Rx queue.
+ * @p_pkt_info Information on the returned buffer ptr.
+ *
+ * OUTPUT: New available Rx resource in Rx descriptor ring.
+ *
+ * RETURN:
+ * KWGBE_ERROR in case the routine can not access Rx desc ring.
+ * KWGBE_OK otherwise.
+ */
+static enum kwgbe_fret_sts rx_return_buff(struct kwgbe_device *dkwgbe,
+ enum kwgbe_q rx_queue,
+ struct kwgbe_pktinf *p_pkt_info)
+{
+ struct kwgbe_portinfo *p_gbe_pctrl = &dkwgbe->portinfo;
+ volatile struct kwgbe_rx_dscr *p_used_rx_desc; /* Where to return Rx resource */
+
+ /* Get 'used' Rx descriptor */
+ USED_RFD_GET(p_used_rx_desc, rx_queue);
+
+ /* Sanity check */
+ if (p_used_rx_desc == NULL)
+ return KWGBE_ERROR;
+
+ p_used_rx_desc->buf_ptr = p_pkt_info->buf_ptr;
+ p_used_rx_desc->retinfo = p_pkt_info->retinfo;
+ p_used_rx_desc->byte_cnt = p_pkt_info->byte_cnt;
+ p_used_rx_desc->buf_size = KW_RX_BUF_SZ; /* Reset Buffer size */
+
+ /* Flush the write pipe */
+ CPU_PIPE_FLUSH;
+
+ /* Return the descriptor to DMA ownership */
+ p_used_rx_desc->cmd_sts =
+ KWGBE_BUFFER_OWNED_BY_DMA | KWGBE_RX_EN_INTERRUPT;
+
+ /* Flush descriptor and CPU pipe */
+ D_CACHE_FLUSH_LINE((u32) p_used_rx_desc, 0);
+ CPU_PIPE_FLUSH;
+
+ /* Move the used descriptor pointer to the next descriptor */
+ USED_RFD_SET(RX_NEXT_DESC_PTR(p_used_rx_desc, rx_queue), rx_queue);
+
+ /* Any Rx return cancels the Rx resource error status */
+ if (p_gbe_pctrl->rx_resource_err[rx_queue] == 1)
+ p_gbe_pctrl->rx_resource_err[rx_queue] = 0;
+
+ return KWGBE_OK;
+}
+
+/*
+ * init function
+ */
+static int kwgbe_init(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct kwgbe_portinfo *portinfo = &dkwgbe->portinfo;
+ struct egiga_priv *portpriv = &dkwgbe->portpriv;
+
+ u32 queue;
+
+ /*
+ * ronen - when we update the MAC env params we only update dev->enetaddr
+ * see ./net/eth.c set_enetaddr()
+ */
+ memcpy(portinfo->port_mac_addr, dev->enetaddr, 6);
+
+ /* Activate the DMA channels etc */
+ port_init(dkwgbe);
+
+ /* "Allocate" setup TX rings */
+ for (queue = 0; queue < KW_TXQ_NO; queue++) {
+ u32 size;
+
+ portpriv->tx_ring_size[queue] = KW_TXQ_SZ;
+ /*size = no of DESCs times DESC-size */
+ size = (portpriv->tx_ring_size[queue] * TX_DESC_ALIGNED_SIZE);
+ portinfo->tx_desc_area_size[queue] = size;
+
+ /* first clear desc area completely */
+ memset((void *)portinfo->p_tx_desc_area_base[queue],
+ 0, portinfo->tx_desc_area_size[queue]);
+
+ /* initialize tx desc ring with low level driver */
+ if (kwgbe_init_tx_desc_ring
+ (dkwgbe, KWGBE_Q0,
+ portpriv->tx_ring_size[queue],
+ KW_TX_BUF_SZ /* Each Buffer is 1600 Byte */ ,
+ (u32) portinfo->
+ p_tx_desc_area_base[queue],
+ (u32) portinfo->p_tx_buffer_base[queue]) == 0)
+ printf("### initializing TX Ring\n");
+ }
+
+ /* "Allocate" setup RX rings */
+ for (queue = 0; queue < KW_RXQ_NO; queue++) {
+ u32 size;
+
+ /* Meantime RX Ring are fixed - but must be configurable by user */
+ portpriv->rx_ring_size[queue] = KW_RXQ_SZ;
+ size = (portpriv->rx_ring_size[queue] * RX_DESC_ALIGNED_SIZE);
+ portinfo->rx_desc_area_size[queue] = size;
+
+ /* first clear desc area completely */
+ memset((void *)portinfo->p_rx_desc_area_base[queue],
+ 0, portinfo->rx_desc_area_size[queue]);
+ if ((kwgbe_init_rx_desc_ring
+ (dkwgbe, KWGBE_Q0,
+ portpriv->rx_ring_size[queue],
+ KW_RX_BUF_SZ /* Each Buffer is 1600 Byte */ ,
+ (u32) portinfo->
+ p_rx_desc_area_base[queue],
+ (u32) portinfo->p_rx_buffer_base[queue])) == 0)
+ printf("### initializing RX Ring\n");
+ }
+
+ port_start(dkwgbe);
+
+#if (defined (CONFIG_MII) || defined (CONFIG_CMD_MII)) \
+ && defined (CONFIG_SYS_FAULT_ECHO_LINK_DOWN)
+ if (!miiphy_link(dev->name, portinfo->port_phy_addr)) {
+ printf("No link on %s\n", dev->name);
+ return 0;
+ }
+#endif
+ portpriv->running = MAGIC_KWGBE_RUNNING;
+ return 1;
+}
+
+/*
+ * egiga_start_xmit
+ *
+ * This function is queues a packet in the Tx descriptor for
+ * required port.
+ *
+ * Output : zero upon success
+ */
+int kwgbe_xmit(struct eth_device *dev, volatile void *dataPtr, int dataSize)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct egiga_priv *portpriv = &dkwgbe->portpriv;
+ struct net_device_stats *stats = &portpriv->stats;
+
+ struct kwgbe_pktinf pkt_info;
+ enum kwgbe_fret_sts status;
+ enum kwgbe_fret_sts release_result;
+
+ /* Update packet info data structure */
+ pkt_info.cmd_sts = KWGBE_TX_FIRST_DESC | KWGBE_TX_LAST_DESC; /* first last */
+ pkt_info.cmd_sts |= KWGBE_BUFFER_OWNED_BY_DMA; /* DMA owned */
+
+ pkt_info.byte_cnt = dataSize;
+ pkt_info.buf_ptr = (u32) dataPtr;
+ pkt_info.retinfo = 0;
+
+ status = port_send(dkwgbe, KWGBE_Q0, &pkt_info);
+ if ((status == KWGBE_ERROR) || (status == KWGBE_QFULL)) {
+ printf("On transmitting packet ..\n");
+ if (status == KWGBE_QFULL)
+ printf("ETH Queue is full.\n");
+ if (status == KWGBE_QLAST_RESO)
+ printf("ETH Queue: using last available resource.\n");
+ goto error;
+ }
+
+ /* Update statistics and start of transmittion time */
+ stats->tx_bytes += dataSize;
+ stats->tx_packets++;
+
+ /* Check if packet(s) is(are) transmitted correctly (release everything) */
+ do {
+ release_result = tx_return_desc(dkwgbe, KWGBE_Q0, &pkt_info);
+ switch (release_result) {
+ case KWGBE_OK:
+ debug("descriptor released\n");
+ if (pkt_info.cmd_sts & 1) {
+ printf("in TX\n");
+ stats->tx_errors++;
+ }
+ break;
+ case KWGBE_RETRY:
+ debug("transmission still in process\n");
+ break;
+ case KWGBE_ERROR:
+ printf("routine can not access Tx desc ring\n");
+ break;
+ case KWGBE_END_OF_JOB:
+ debug("the routine has nothing to release\n");
+ break;
+ default: /* should not happen */
+ break;
+ }
+ } while (release_result == KWGBE_OK);
+
+ return 0; /* success */
+ error:
+ return 1; /* Failed - higher layers will free the skb */
+}
+
+/*
+ * kwgbe_receive
+ *
+ * This function is forward packets that are received from the port's
+ * queues toward kernel core or FastRoute them to another interface.
+ *
+ * @dev a pointer to the required interface
+ *
+ * Output : number of served packets
+ */
+int kwgbe_receive(struct eth_device *dev)
+{
+ struct kwgbe_device *dkwgbe = to_dkwgbe(dev);
+ struct egiga_priv *portpriv = &dkwgbe->portpriv;
+ struct net_device_stats *stats = &portpriv->stats;
+ struct kwgbe_pktinf pkt_info;
+
+ while ((port_receive(dkwgbe, KWGBE_Q0, &pkt_info) == KWGBE_OK)) {
+
+ if (pkt_info.byte_cnt != 0) {
+ debug("%s: Received %d byte Packet @ 0x%x",
+ __FUNCTION__, pkt_info.byte_cnt,
+ pkt_info.buf_ptr);
+ }
+ /* Update statistics. Note byte count includes 4 byte CRC count */
+ stats->rx_packets++;
+ stats->rx_bytes += pkt_info.byte_cnt;
+
+ /*
+ * In case received a packet without first / last bits on OR the error
+ * summary bit is on, the packets needs to be dropeed.
+ */
+ if (((pkt_info.
+ cmd_sts & (KWGBE_RX_FIRST_DESC | KWGBE_RX_LAST_DESC)) !=
+ (KWGBE_RX_FIRST_DESC | KWGBE_RX_LAST_DESC))
+ || (pkt_info.cmd_sts & KWGBE_ERROR_SUMMARY)) {
+ stats->rx_dropped++;
+
+ printf
+ ("Received packet spread on multiple descriptors\n");
+
+ /* Is this caused by an error ? */
+ if (pkt_info.cmd_sts & KWGBE_ERROR_SUMMARY) {
+ stats->rx_errors++;
+ }
+
+ /* free these descriptors again without forwarding them to the higher layers */
+ pkt_info.buf_ptr &= ~0x7; /* realign buffer again */
+ pkt_info.byte_cnt = 0x0000; /* Reset Byte count */
+
+ if (rx_return_buff(dkwgbe, KWGBE_Q0, &pkt_info) !=
+ KWGBE_OK) {
+ printf
+ ("Error while returning the RX Desc to Ring\n");
+ } else {
+ debug("RX Desc returned to Ring");
+ }
+ /* /free these descriptors again */
+ } else {
+
+ /* !!! call higher layer processing */
+ debug
+ ("\nNow send it to upper layer protocols (NetReceive) ...");
+
+ /* let the upper layer handle the packet */
+ NetReceive((uchar *) pkt_info.buf_ptr,
+ (int)pkt_info.byte_cnt);
+
+ /* free descriptor */
+ pkt_info.buf_ptr &= ~0x7; /* realign buffer again */
+ pkt_info.byte_cnt = 0x0000; /* Reset Byte count */
+ debug("RX: pkt_info.buf_ptr = %x", pkt_info.buf_ptr);
+ if (rx_return_buff(dkwgbe, KWGBE_Q0, &pkt_info) !=
+ KWGBE_OK) {
+ printf("while returning the RX Desc to Ring");
+ } else {
+ debug("RX Desc returned to Ring");
+ }
+ /* free descriptor code end */
+ }
+ }
+ egiga_get_stats(dev); /* update statistics */
+ return 1;
+}
+
+/*
+ * Called from net/eth.c
+ */
+int kirkwood_egiga_initialize(bd_t * bis)
+{
+ struct kwgbe_device *dkwgbe;
+ struct eth_device *dev;
+ struct kwgbe_portinfo *portinfo;
+ struct egiga_priv *portpriv;
+ int devnum, x, temp;
+ char *s, *e, buf[64];
+ u8 used_ports[MAX_KWGBE_DEVS] = CONFIG_KIRKWOOD_EGIGA_PORTS;
+
+ for (devnum = 0; devnum < MAX_KWGBE_DEVS; devnum++) {
+ /*skip if port is configured not to use */
+ if (used_ports[devnum] == 0)
+ continue;
+
+ dkwgbe = malloc(sizeof(struct kwgbe_device));
+ if (!dkwgbe) {
+ printf("Error: Failed to allocate memory for kwgbe%d\n",
+ devnum);
+ return -1;
+ }
+ memset(dkwgbe, 0, sizeof(struct kwgbe_device));
+
+ dev = &dkwgbe->dev;
+ portinfo = &dkwgbe->portinfo;
+ portpriv = &dkwgbe->portpriv;
+
+ /* must be less than NAMESIZE (16) */
+ sprintf(dev->name, "egiga%d", devnum);
+ debug("Initializing %s", dev->name);
+
+ /* Extract the MAC address from the environment */
+ switch (devnum) {
+ case 0:
+ dkwgbe->regs = (void *)KW_EGIGA0_BASE;
+ s = "ethaddr";
+ break;
+ case 1:
+ dkwgbe->regs = (void *)KW_EGIGA1_BASE;
+ s = "eth1addr";
+ break;
+ default: /* this should never happen */
+ printf("%s: Invalid device number %d\n",
+ __FUNCTION__, devnum);
+ return -1;
+ }
+
+ e = getenv(s);
+ if (!e) {
+ /* Generate Ramdom MAC addresses if not set */
+ sprintf(buf, "00:50:43:%02x:%02x:%02x",
+ get_random_hex(), get_random_hex(),
+ get_random_hex());
+ setenv(s, buf);
+ }
+
+ temp = getenv_r(s, buf, sizeof(buf));
+ s = (temp > 0) ? buf : NULL;
+
+ debug("Setting MAC %d to %s", devnum, s);
+ /* on POR port is pre-powered up, so assume it is up */
+
+ for (x = 0; x < 6; ++x) {
+ dev->enetaddr[x] = s ? simple_strtoul(s, &e, 16) : 0;
+ if (s)
+ s = (*e) ? e + 1 : e;
+ }
+ /* set the MAC addr in the HW */
+ memcpy(portinfo->port_mac_addr, dev->enetaddr, 6);
+ port_uc_addr_set(dkwgbe->regs, dev->enetaddr, 0);
+
+ dev->init = (void *)kwgbe_init;
+ dev->halt = (void *)kwgbe_stop;
+ dev->send = (void *)kwgbe_xmit;
+ dev->recv = (void *)kwgbe_receive;
+
+ /*
+ * Read MIB counter on the GT in order to reset them,
+ * then zero all the stats fields in memory
+ */
+ egiga_update_stat(dev);
+ memset(&portpriv->stats, 0, sizeof(struct net_device_stats));
+
+ debug("Allocating descriptor and buffer rings");
+
+ portinfo->p_rx_desc_area_base[0] =
+ (struct kwgbe_rx_dscr *)memalign(16,
+ RX_DESC_ALIGNED_SIZE *
+ KW_RXQ_SZ + 1);
+ portinfo->p_tx_desc_area_base[0] =
+ (struct kwgbe_tx_dscr *)memalign(16,
+ TX_DESC_ALIGNED_SIZE *
+ KW_TXQ_SZ + 1);
+
+ portinfo->p_rx_buffer_base[0] =
+ (char *)memalign(16, KW_RXQ_SZ * KW_TX_BUF_SZ + 1);
+ portinfo->p_tx_buffer_base[0] =
+ (char *)memalign(16, KW_RXQ_SZ * KW_TX_BUF_SZ + 1);
+
+ eth_register(dev);
+
+ /* Set phy address of the port */
+ portinfo->port_phy_addr = PHY_BASE_ADR + devnum;
+ phy_set_addr(dkwgbe->regs, portinfo->port_phy_addr);
+
+#if defined(CONFIG_MII) || defined(CONFIG_CMD_MII)
+ miiphy_register(dev->name, smi_reg_read, smi_reg_write);
+#endif
+ }
+ return 0;
+}
diff --git a/drivers/net/kirkwood_egiga.h b/drivers/net/kirkwood_egiga.h
new file mode 100644
index 0000000..3a5a306
--- /dev/null
+++ b/drivers/net/kirkwood_egiga.h
@@ -0,0 +1,828 @@
+/*
+ * (C) Copyright 2009
+ * Marvell Semiconductor <www.marvell.com>
+ * Prafulla Wadaskar <prafulla@marvell.com>
+ *
+ * based on - Driver for MV64360X ethernet ports
+ * Copyright (C) 2002 rabeeh at galileo.co.il
+ *
+ * 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., 51 Franklin Street, Fifth Floor, Boston,
+ * MA 02110-1301 USA
+ */
+
+#ifndef __EGIGA_H__
+#define __EGIGA_H__
+
+#define MAX_KWGBE_DEVS 2 /*controller has two ports */
+
+/* PHY_BASE_ADR is board specific and can be configured */
+#if defined (CONFIG_PHY_BASE_ADR)
+#define PHY_BASE_ADR CONFIG_PHY_BASE_ADR
+#else
+#define PHY_BASE_ADR 0x08 /* default phy base addr */
+#endif
+
+/* Egiga driver statistics can be displayed optionally */
+#ifdef CONFIG_EGIGA_STATS_BY_SOFTWARE
+#define UPDATE_STATS_BY_SOFTWARE
+#else
+#undef UPDATE_STATS_BY_SOFTWARE
+#endif
+
+/* Constants */
+#define MAGIC_KWGBE_RUNNING 8031971
+#define INT_CAUSE_UNMASK_ALL 0x0007ffff
+#define INT_CAUSE_UNMASK_ALL_EXT 0x0011ffff
+
+/* Port attrib */
+/* Max queue number is 8 for controller but here configured 1 */
+#define MAX_RXQ_NO 1
+#define MAX_TXQ_NO 1
+/* Use one TX queue and one RX queue */
+#define KW_TXQ_NO 1
+#define KW_RXQ_NO 1
+
+/*
+ * Number of RX / TX descriptors on RX / TX rings.
+ * Note that allocating RX descriptors is done by allocating the RX
+ * ring AND a preallocated RX buffers (skb's) for each descriptor.
+ * The TX descriptors only allocates the TX descriptors ring,
+ * with no pre allocated TX buffers (skb's are allocated by higher layers.
+ */
+
+/* TX/RX parameters can be passed externally */
+#ifdef CONFIG_KWGBE_TXQ_SZ
+#define KW_TXQ_SZ CONFIG_KWGBE_TXQ_SZ
+#else
+#define KW_TXQ_SZ 4
+#endif
+
+#ifdef CONFIG_KWGBE_RXQ_SZ
+#define KW_RXQ_SZ CONFIG_KWGBE_RXQ_SZ
+#else
+#define KW_RXQ_SZ 4
+#endif
+
+#ifdef CONFIG_KWGBE_RXBUF_SZ
+#define KW_RX_BUF_SZ CONFIG_KWGBE_RXBUF_SZ
+#else
+#define KW_RX_BUF_SZ 1600
+#endif
+
+#ifdef CONFIG_KWGBE_TXBUF_SZ
+#define KW_TX_BUF_SZ CONFIG_KWGBE_TXBUF_SZ
+#else
+#define KW_TX_BUF_SZ 1600
+#endif
+
+#define RX_BUF_MXSZ 0xFFFF
+#define TX_BUF_MXSZ 0xFFFF /* Buffers are limited to 64k */
+#define RX_BUF_MNSZ 0x8
+#define TX_BUF_MNSZ 0x8
+
+/*
+ * Network device statistics. Akin to the 2.0 ether stats but
+ * with byte counters.
+ */
+struct net_device_stats {
+ u32 rx_packets; /* total packets received */
+ u32 tx_packets; /* total packets transmitted */
+ u64 rx_bytes; /* total bytes received */
+ u64 tx_bytes; /* total bytes transmitted */
+ u32 rx_errors; /* bad packets received */
+ u32 tx_errors; /* packet transmit problems */
+ u32 rx_dropped; /* no space in linux buffers */
+ u32 tx_dropped; /* no space available in linux */
+ u32 multicast; /* multicast packets received */
+ u32 collisions;
+ /* detailed rx_errors: */
+ u32 rx_length_errors;
+ u32 rx_over_errors; /* receiver ring buff overflow */
+ u32 rx_crc_errors; /* recved pkt with crc error */
+ u32 rx_frame_errors; /* recv'd frame alignment error */
+ u32 rx_fifo_errors; /* recv'r fifo overrun */
+ u32 rx_missed_errors; /* receiver missed packet */
+ /* detailed tx_errors */
+ u32 tx_aborted_errors;
+ u32 tx_carrier_errors;
+ u32 tx_fifo_errors;
+ u32 tx_heartbeat_errors;
+ u32 tx_window_errors;
+ /* for cslip etc */
+ u32 rx_compressed;
+ u32 tx_compressed;
+};
+
+/* Private data structure used for ethernet device */
+struct egiga_priv {
+ struct net_device_stats stats;
+ /* to buffer area aligned */
+ char *p_tx_buffer[KW_TXQ_SZ + 1];
+ char *p_rx_buffer[KW_RXQ_SZ + 1];
+ /* Size of Tx Ring per queue */
+ u32 tx_ring_size[MAX_TXQ_NO];
+ /* Size of Rx Ring per queue */
+ u32 rx_ring_size[MAX_RXQ_NO];
+ /* Magic Number for Ethernet running */
+ u32 running;
+};
+
+/* Chip Registers Structures */
+struct mib_counters {
+ u64 good_octets_received;
+ u32 bad_octets_received;
+ u32 internal_mac_transmit_err;
+ u32 good_frames_received;
+ u32 bad_frames_received;
+ u32 broadcast_frames_received;
+ u32 multicast_frames_received;
+ u32 frames_64_octets;
+ u32 frames_65_to_127_octets;
+ u32 frames_128_to_255_octets;
+ u32 frames_256_to_511_octets;
+ u32 frames_512_to_1023_octets;
+ u32 frames_1024_to_max_octets;
+ u64 good_octets_sent;
+ u32 good_frames_sent;
+ u32 excessive_collision;
+ u32 multicast_frames_sent;
+ u32 broadcast_frames_sent;
+ u32 unrec_mac_control_received;
+ u32 fc_sent;
+ u32 good_fc_received;
+ u32 bad_fc_received;
+ u32 undersize_received;
+ u32 fragments_received;
+ u32 oversize_received;
+ u32 jabber_received;
+ u32 mac_receive_error;
+ u32 bad_crc_event;
+ u32 collision;
+ u32 late_collision;
+};
+
+struct kwgbe_barsz {
+ u32 bar;
+ u32 size;
+};
+
+struct kwgbe_crdp {
+ u32 crdp;
+ u32 crdp_pad[3];
+};
+
+struct kwgbe_tqx {
+ u32 qxttbc;
+ u32 tqxtbc;
+ u32 tqxac;
+ u32 tqxpad;
+};
+
+struct kwgbe_registers {
+ u32 phyadr;
+ u32 smi;
+ u32 euda;
+ u32 eudid;
+ u8 pad1[0x080 - 0x00c - 4];
+ u32 euic;
+ u32 euim;
+ u8 pad2[0x094 - 0x084 - 4];
+ u32 euea;
+ u32 euiae;
+ u8 pad3[0x0b0 - 0x098 - 4];
+ u32 euc;
+ u8 pad3a[0x200 - 0x0b0 - 4];
+ struct kwgbe_barsz barsz[6];
+ u8 pad4[0x280 - 0x22c - 4];
+ u32 ha_remap[4];
+ u32 bare;
+ u32 epap;
+ u8 pad5[0x400 - 0x294 - 4];
+ u32 pxc;
+ u32 pxcx;
+ u32 mii_ser_params;
+ u8 pad6[0x410 - 0x408 - 4];
+ u32 evlane;
+ u32 macal;
+ u32 macah;
+ u32 sdc;
+ u32 dscp[7];
+ u32 psc0;
+ u32 vpt2p;
+ u32 ps0;
+ u32 tqc;
+ u32 psc1;
+ u32 ps1;
+ u32 mrvl_header;
+ u8 pad7[0x460 - 0x454 - 4];
+ u32 ic;
+ u32 ice;
+ u32 pim;
+ u32 peim;
+ u8 pad8[0x474 - 0x46c - 4];
+ u32 pxtfut;
+ u32 pad9;
+ u32 pxmfs;
+ u32 pad10;
+ u32 pxdfc;
+ u32 pxofc;
+ u8 pad11[0x494 - 0x488 - 4];
+ u32 peuiae;
+ u8 pad12[0x4bc - 0x494 - 4];
+ u32 eth_type_prio;
+ u8 pad13[0x4dc - 0x4bc - 4];
+ u32 tqfpc;
+ u32 pttbrc;
+ u32 tqc1;
+ u32 pmtu;
+ u32 pmtbs;
+ u8 pad14[0x60c - 0x4ec - 4];
+ struct kwgbe_crdp crdp[7];
+ u32 crdp7;
+ u32 rqc;
+ u32 cur_servd_dscr_ptr;
+ u8 pad15[0x6c0 - 0x684 - 4];
+ u32 tcqdp[8];
+ u8 pad16[0x700 - 0x6dc - 4];
+ struct kwgbe_tqx tqx[8];
+ u32 pttbc;
+ u8 pad17[0x7a8 - 0x780 - 4];
+ u32 tqxipg0;
+ u32 pad18[3];
+ u32 tqxipg1;
+ u8 pad19[0x7c0 - 0x7b8 - 4];
+ u32 hitkninlopkt;
+ u32 hitkninasyncpkt;
+ u32 lotkninasyncpkt;
+ u32 pad20;
+ u32 ts;
+ u8 pad21[0x3000 - 0x27d0 - 4];
+ struct mib_counters mibc;
+ u8 pad22[0x3400 - 0x3000 - sizeof(struct mib_counters)];
+ u32 dfsmt[64];
+ u32 dfomt[64];
+ u32 dfut[4];
+ u8 pad23[0xe20c0 - 0x7360c - 4];
+ u32 pmbus_top_arbiter;
+};
+
+#define to_dkwgbe(_kd) container_of(_kd, struct kwgbe_device, dev)
+#define KWGBEREG_WR(adr, val) writel(val, (u32)&adr)
+#define KWGBEREG_RD(adr) readl((u32)&adr)
+#define KWGBEREG_BITS_RESET(adr, val) writel(readl((u32)&adr) & ~(val), (u32)&adr)
+#define KWGBEREG_BITS_SET(adr, val) writel(readl((u32)&adr) | val, (u32)&adr)
+
+/* Default port configuration value */
+#define PRT_CFG_VAL \
+ KWGBE_UCAST_MOD_NRML | \
+ KWGBE_DFLT_RX_Q0 | \
+ KWGBE_DFLT_RX_ARPQ0 | \
+ KWGBE_RX_BC_IF_NOT_IP_OR_ARP | \
+ KWGBE_RX_BC_IF_IP | \
+ KWGBE_RX_BC_IF_ARP | \
+ KWGBE_CPTR_TCP_FRMS_DIS | \
+ KWGBE_CPTR_UDP_FRMS_DIS | \
+ KWGBE_DFLT_RX_TCPQ0 | \
+ KWGBE_DFLT_RX_UDPQ0 | \
+ KWGBE_DFLT_RX_BPDU_Q_0
+
+/* Default port extend configuration value */
+#define PORT_CFG_EXTEND_VALUE \
+ KWGBE_SPAN_BPDU_PACKETS_AS_NORMAL | \
+ KWGBE_PARTITION_DIS | \
+ KWGBE_TX_CRC_GENERATION_EN
+
+/* Default sdma control value */
+#ifdef CONFIG_NOT_COHERENT_CACHE
+#define PORT_SDMA_CFG_VALUE \
+ KWGBE_RX_BURST_SIZE_16_64BIT | \
+ KWGBE_BLM_RX_NO_SWAP | \
+ KWGBE_BLM_TX_NO_SWAP | \
+ GT_KWGBE_IPG_INT_RX(0) | \
+ KWGBE_TX_BURST_SIZE_16_64BIT;
+#else
+#define PORT_SDMA_CFG_VALUE \
+ KWGBE_RX_BURST_SIZE_4_64BIT | \
+ GT_KWGBE_IPG_INT_RX(0) | \
+ KWGBE_TX_BURST_SIZE_4_64BIT;
+#endif
+
+#define GT_KWGBE_IPG_INT_RX(value) ((value & 0x3fff) << 8)
+
+/* Default port serial control value */
+#define PORT_SERIAL_CONTROL_VALUE_TMP \
+ KWGBE_FORCE_LINK_PASS | \
+ KWGBE_DIS_AUTO_NEG_FOR_DUPLX | \
+ KWGBE_DIS_AUTO_NEG_FOR_FLOW_CTRL | \
+ KWGBE_ADV_NO_FLOW_CTRL | \
+ KWGBE_FORCE_FC_MODE_NO_PAUSE_DIS_TX | \
+ KWGBE_FORCE_BP_MODE_NO_JAM | \
+ (1 << 9) | \
+ KWGBE_DO_NOT_FORCE_LINK_FAIL | \
+ KWGBE_DIS_AUTO_NEG_SPEED_GMII | \
+ KWGBE_DTE_ADV_0 | \
+ KWGBE_MIIPHY_MAC_MODE | \
+ KWGBE_AUTO_NEG_NO_CHANGE | \
+ KWGBE_MAX_RX_PACKET_1552BYTE | \
+ KWGBE_CLR_EXT_LOOPBACK | \
+ KWGBE_SET_FULL_DUPLEX_MODE | \
+ KWGBE_DIS_FLOW_CTRL_TX_RX_IN_FULL_DUPLEX | \
+ KWGBE_SET_MII_SPEED_TO_100
+
+#ifdef CONFIG_SYS_MII_MODE
+#define PORT_SERIAL_CONTROL_VALUE \
+ PORT_SERIAL_CONTROL_VALUE_TMP | \
+ KWGBE_SET_GMII_SPEED_TO_10_100
+#else
+#define PORT_SERIAL_CONTROL_VALUE \
+ PORT_SERIAL_CONTROL_VALUE_TMP | \
+ KWGBE_SET_GMII_SPEED_TO_1000
+#endif
+
+/* Tx WRR confoguration macros */
+#define PORT_MAX_TRAN_UNIT 0x24 /* MTU register (default) 9KByte */
+#define PORT_MAX_TOKEN_BUCKET_SIZE 0x_fFFF /* PMTBS reg (default) */
+#define PORT_TOKEN_RATE 1023 /* PTTBRC reg (default) */
+/* MAC accepet/reject macros */
+#define ACCEPT_MAC_ADDR 0
+#define REJECT_MAC_ADDR 1
+/* Size of a Tx/Rx descriptor used in chain list data structure */
+#define RX_DESC_ALIGNED_SIZE 0x20
+#define TX_DESC_ALIGNED_SIZE 0x20
+/* An offest in Tx descriptors to store data for buffers less than 8 Bytes */
+#define TX_BUF_OFFSET_IN_DESC 0x18
+/* Buffer offset from buffer pointer */
+#define RX_BUF_OFFSET 0x2
+
+/* Port serial status reg (PSR) */
+#define KWGBE_INTERFACE_GMII_MII 0
+#define KWGBE_INTERFACE_PCM 1
+#define KWGBE_LINK_IS_DOWN 0
+#define KWGBE_LINK_IS_UP (1 << 1)
+#define KWGBE_PORT_AT_HALF_DUPLEX 0
+#define KWGBE_PORT_AT_FULL_DUPLEX (1 << 2)
+#define KWGBE_RX_FLOW_CTRL_DISD 0
+#define KWGBE_RX_FLOW_CTRL_ENBALED (1 << 3)
+#define KWGBE_GMII_SPEED_100_10 0
+#define KWGBE_GMII_SPEED_1000 (1 << 4)
+#define KWGBE_MII_SPEED_10 0
+#define KWGBE_MII_SPEED_100 (1 << 5)
+#define KWGBE_NO_TX 0
+#define KWGBE_TX_IN_PROGRESS (1 << 7)
+#define KWGBE_BYPASS_NO_ACTIVE 0
+#define KWGBE_BYPASS_ACTIVE (1 << 8)
+#define KWGBE_PORT_NOT_AT_PARTN_STT 0
+#define KWGBE_PORT_AT_PARTN_STT (1 << 9)
+#define KWGBE_PORT_TX_FIFO_NOT_EMPTY 0
+#define KWGBE_PORT_TX_FIFO_EMPTY (1 << 10)
+
+/* These macros describes the Port configuration reg (Px_cR) bits */
+#define KWGBE_UCAST_MOD_NRML 0
+#define KWGBE_UNICAST_PROMISCUOUS_MODE 1
+#define KWGBE_DFLT_RX_Q0 0
+#define KWGBE_DFLT_RX_Q1 (1 << 1)
+#define KWGBE_DFLT_RX_Q2 (1 << 2)
+#define KWGBE_DFLT_RX_Q3 ((1 << 2) | (1 << 1))
+#define KWGBE_DFLT_RX_Q4 (1 << 3)
+#define KWGBE_DFLT_RX_Q5 ((1 << 3) | (1 << 1))
+#define KWGBE_DFLT_RX_Q6 ((1 << 3) | (1 << 2))
+#define KWGBE_DFLT_RX_Q7 ((1 << 3) | (1 << 2) | (1 << 1))
+#define KWGBE_DFLT_RX_ARPQ0 0
+#define KWGBE_DFLT_RX_ARPQ1 (1 << 4)
+#define KWGBE_DFLT_RX_ARPQ2 (1 << 5)
+#define KWGBE_DFLT_RX_ARPQ3 ((1 << 5) | (1 << 4))
+#define KWGBE_DFLT_RX_ARPQ4 (1 << 6)
+#define KWGBE_DFLT_RX_ARPQ5 ((1 << 6) | (1 << 4))
+#define KWGBE_DFLT_RX_ARPQ6 ((1 << 6) | (1 << 5))
+#define KWGBE_DFLT_RX_ARPQ7 ((1 << 6) | (1 << 5) | (1 << 4))
+#define KWGBE_RX_BC_IF_NOT_IP_OR_ARP 0
+#define KWGBE_REJECT_BC_IF_NOT_IP_OR_ARP (1 << 7)
+#define KWGBE_RX_BC_IF_IP 0
+#define KWGBE_REJECT_BC_IF_IP (1 << 8)
+#define KWGBE_RX_BC_IF_ARP 0
+#define KWGBE_REJECT_BC_IF_ARP (1 << 9)
+#define KWGBE_TX_AM_NO_UPDATE_ERR_SMRY (1 << 12)
+#define KWGBE_CPTR_TCP_FRMS_DIS 0
+#define KWGBE_CPTR_TCP_FRMS_EN (1 << 14)
+#define KWGBE_CPTR_UDP_FRMS_DIS 0
+#define KWGBE_CPTR_UDP_FRMS_EN (1 << 15)
+#define KWGBE_DFLT_RX_TCPQ0 0
+#define KWGBE_DFLT_RX_TCPQ1 (1 << 16)
+#define KWGBE_DFLT_RX_TCPQ2 (1 << 17)
+#define KWGBE_DFLT_RX_TCPQ3 ((1 << 17) | (1 << 16))
+#define KWGBE_DFLT_RX_TCPQ4 (1 << 18)
+#define KWGBE_DFLT_RX_TCPQ5 ((1 << 18) | (1 << 16))
+#define KWGBE_DFLT_RX_TCPQ6 ((1 << 18) | (1 << 17))
+#define KWGBE_DFLT_RX_TCPQ7 ((1 << 18) | (1 << 17) | (1 << 16))
+#define KWGBE_DFLT_RX_UDPQ0 0
+#define KWGBE_DFLT_RX_UDPQ1 (1 << 19)
+#define KWGBE_DFLT_RX_UDPQ2 (1 << 20)
+#define KWGBE_DFLT_RX_UDPQ3 ((1 << 20) | (1 << 19))
+#define KWGBE_DFLT_RX_UDPQ4 (1 << 21)
+#define KWGBE_DFLT_RX_UDPQ5 ((1 << 21) | (1 << 19))
+#define KWGBE_DFLT_RX_UDPQ6 ((1 << 21) | (1 << 20))
+#define KWGBE_DFLT_RX_UDPQ7 ((1 << 21) | (1 << 20) | (1 << 19))
+#define KWGBE_DFLT_RX_BPDU_Q_0 0
+#define KWGBE_DFLT_RX_BPDU_Q_1 (1 << 22)
+#define KWGBE_DFLT_RX_BPDU_Q_2 (1 << 23)
+#define KWGBE_DFLT_RX_BPDU_Q_3 ((1 << 23) | (1 << 22))
+#define KWGBE_DFLT_RX_BPDU_Q_4 (1 << 24)
+#define KWGBE_DFLT_RX_BPDU_Q_5 ((1 << 24) | (1 << 22))
+#define KWGBE_DFLT_RX_BPDU_Q_6 ((1 << 24) | (1 << 23))
+#define KWGBE_DFLT_RX_BPDU_Q_7 ((1 << 24) | (1 << 23) | (1 << 22))
+#define KWGBE_DFLT_RX_TCP_CHKSUM_MODE (1 << 25)
+
+/* These macros describes the Port configuration extend reg (Px_cXR) bits*/
+#define KWGBE_CLASSIFY_EN 1
+#define KWGBE_SPAN_BPDU_PACKETS_AS_NORMAL 0
+#define KWGBE_SPAN_BPDU_PACKETS_TO_RX_Q7 (1 << 1)
+#define KWGBE_PARTITION_DIS 0
+#define KWGBE_PARTITION_EN (1 << 2)
+#define KWGBE_TX_CRC_GENERATION_EN 0
+#define KWGBE_TX_CRC_GENERATION_DIS (1 << 3)
+
+/* Tx/Rx queue command reg (RQCR/TQCR)*/
+#define KWGBE_Q_0_EN 1
+#define KWGBE_Q_1_EN (1 << 1)
+#define KWGBE_Q_2_EN (1 << 2)
+#define KWGBE_Q_3_EN (1 << 3)
+#define KWGBE_Q_4_EN (1 << 4)
+#define KWGBE_Q_5_EN (1 << 5)
+#define KWGBE_Q_6_EN (1 << 6)
+#define KWGBE_Q_7_EN (1 << 7)
+#define KWGBE_Q_0_DIS (1 << 8)
+#define KWGBE_Q_1_DIS (1 << 9)
+#define KWGBE_Q_2_DIS (1 << 10)
+#define KWGBE_Q_3_DIS (1 << 11)
+#define KWGBE_Q_4_DIS (1 << 12)
+#define KWGBE_Q_5_DIS (1 << 13)
+#define KWGBE_Q_6_DIS (1 << 14)
+#define KWGBE_Q_7_DIS (1 << 15)
+
+/* These macros describes the Port Sdma configuration reg (SDCR) bits */
+#define KWGBE_RIFB 1
+#define KWGBE_RX_BURST_SIZE_1_64BIT 0
+#define KWGBE_RX_BURST_SIZE_2_64BIT (1 << 1)
+#define KWGBE_RX_BURST_SIZE_4_64BIT (1 << 2)
+#define KWGBE_RX_BURST_SIZE_8_64BIT ((1 << 2) | (1 << 1))
+#define KWGBE_RX_BURST_SIZE_16_64BIT (1 << 3)
+#define KWGBE_BLM_RX_NO_SWAP (1 << 4)
+#define KWGBE_BLM_RX_BYTE_SWAP 0
+#define KWGBE_BLM_TX_NO_SWAP (1 << 5)
+#define KWGBE_BLM_TX_BYTE_SWAP 0
+#define KWGBE_DESCRIPTORS_BYTE_SWAP (1 << 6)
+#define KWGBE_DESCRIPTORS_NO_SWAP 0
+#define KWGBE_TX_BURST_SIZE_1_64BIT 0
+#define KWGBE_TX_BURST_SIZE_2_64BIT (1 << 22)
+#define KWGBE_TX_BURST_SIZE_4_64BIT (1 << 23)
+#define KWGBE_TX_BURST_SIZE_8_64BIT ((1 << 23) | (1 << 22))
+#define KWGBE_TX_BURST_SIZE_16_64BIT (1 << 24)
+
+/* These macros describes the Port serial control reg (PSCR) bits */
+#define KWGBE_SERIAL_PORT_DIS 0
+#define KWGBE_SERIAL_PORT_EN 1
+#define KWGBE_FORCE_LINK_PASS (1 << 1)
+#define KWGBE_DO_NOT_FORCE_LINK_PASS 0
+#define KWGBE_EN_AUTO_NEG_FOR_DUPLX 0
+#define KWGBE_DIS_AUTO_NEG_FOR_DUPLX (1 << 2)
+#define KWGBE_EN_AUTO_NEG_FOR_FLOW_CTRL 0
+#define KWGBE_DIS_AUTO_NEG_FOR_FLOW_CTRL (1 << 3)
+#define KWGBE_ADV_NO_FLOW_CTRL 0
+#define KWGBE_ADV_SYMMETRIC_FLOW_CTRL (1 << 4)
+#define KWGBE_FORCE_FC_MODE_NO_PAUSE_DIS_TX 0
+#define KWGBE_FORCE_FC_MODE_TX_PAUSE_DIS (1 << 5)
+#define KWGBE_FORCE_BP_MODE_NO_JAM 0
+#define KWGBE_FORCE_BP_MODE_JAM_TX (1 << 7)
+#define KWGBE_FORCE_BP_MODE_JAM_TX_ON_RX_ERR (1 << 8)
+#define KWGBE_FORCE_LINK_FAIL 0
+#define KWGBE_DO_NOT_FORCE_LINK_FAIL (1 << 10)
+#define KWGBE_DIS_AUTO_NEG_SPEED_GMII (1 << 13)
+#define KWGBE_EN_AUTO_NEG_SPEED_GMII 0
+#define KWGBE_DTE_ADV_0 0
+#define KWGBE_DTE_ADV_1 (1 << 14)
+#define KWGBE_MIIPHY_MAC_MODE 0
+#define KWGBE_MIIPHY_PHY_MODE (1 << 15)
+#define KWGBE_AUTO_NEG_NO_CHANGE 0
+#define KWGBE_RESTART_AUTO_NEG (1 << 16)
+#define KWGBE_MAX_RX_PACKET_1518BYTE 0
+#define KWGBE_MAX_RX_PACKET_1522BYTE (1 << 17)
+#define KWGBE_MAX_RX_PACKET_1552BYTE (1 << 18)
+#define KWGBE_MAX_RX_PACKET_9022BYTE ((1 << 18) | (1 << 17))
+#define KWGBE_MAX_RX_PACKET_9192BYTE (1 << 19)
+#define KWGBE_MAX_RX_PACKET_9700BYTE ((1 << 19) | (1 << 17))
+#define KWGBE_SET_EXT_LOOPBACK (1 << 20)
+#define KWGBE_CLR_EXT_LOOPBACK 0
+#define KWGBE_SET_FULL_DUPLEX_MODE (1 << 21)
+#define KWGBE_SET_HALF_DUPLEX_MODE 0
+#define KWGBE_EN_FLOW_CTRL_TX_RX_IN_FULL_DUPLEX (1 << 22)
+#define KWGBE_DIS_FLOW_CTRL_TX_RX_IN_FULL_DUPLEX 0
+#define KWGBE_SET_GMII_SPEED_TO_10_100 0
+#define KWGBE_SET_GMII_SPEED_TO_1000 (1 << 23)
+#define KWGBE_SET_MII_SPEED_TO_10 0
+#define KWGBE_SET_MII_SPEED_TO_100 (1 << 24)
+
+/* SMI register fields (KWGBE_SMI_REG) */
+#define KWGBE_PHY_SMI_TIMEOUT 10000
+#define KWGBE_PHY_SMI_DATA_OFFS 0 /* Data */
+#define KWGBE_PHY_SMI_DATA_MASK (0xffff << KWGBE_PHY_SMI_DATA_OFFS)
+#define KWGBE_PHY_SMI_DEV_ADDR_OFFS 16 /* PHY device address */
+#define KWGBE_PHY_SMI_DEV_ADDR_MASK (0x1f << KWGBE_PHY_SMI_DEV_ADDR_OFFS)
+#define KWGBE_SMI_REG_ADDR_OFFS 21 /* PHY device register address */
+#define KWGBE_SMI_REG_ADDR_MASK (0x1f << KWGBE_SMI_REG_ADDR_OFFS)
+#define KWGBE_PHY_SMI_OPCODE_OFFS 26 /* Write/Read opcode */
+#define KWGBE_PHY_SMI_OPCODE_MASK (3 << KWGBE_PHY_SMI_OPCODE_OFFS)
+#define KWGBE_PHY_SMI_OPCODE_WRITE (0 << KWGBE_PHY_SMI_OPCODE_OFFS)
+#define KWGBE_PHY_SMI_OPCODE_READ (1 << KWGBE_PHY_SMI_OPCODE_OFFS)
+#define KWGBE_PHY_SMI_READ_VALID_MASK (1 << 27) /* Read Valid */
+#define KWGBE_PHY_SMI_BUSY_MASK (1 << 28) /* Busy */
+
+/* SDMA command status fields macros */
+/* Tx & Rx descriptors status */
+#define KWGBE_ERROR_SUMMARY 1
+
+/* Tx & Rx descriptors command */
+#define KWGBE_BUFFER_OWNED_BY_DMA (1 << 31)
+
+/* Tx descriptors status */
+#define KWGBE_LC_ERROR 0
+#define KWGBE_UR_ERROR (1 << 1)
+#define KWGBE_RL_ERROR (1 << 2)
+#define KWGBE_LLC_SNAP_FORMAT (1 << 9)
+
+/* Rx descriptors status */
+#define KWGBE_CRC_ERROR 0
+#define KWGBE_OVERRUN_ERROR (1 << 1)
+#define KWGBE_MAX_FRAME_LENGTH_ERROR (1 << 2)
+#define KWGBE_RESOURCE_ERROR ((1 << 2) | (1 << 1))
+#define KWGBE_VLAN_TAGGED (1 << 19)
+#define KWGBE_BPDU_FRAME (1 << 20)
+#define KWGBE_TCP_FRAME_OVER_IP_V_4 0
+#define KWGBE_UDP_FRAME_OVER_IP_V_4 (1 << 21)
+#define KWGBE_OTHER_FRAME_TYPE (1 << 22)
+#define KWGBE_LAYER_2_IS_KWGBE_V_2 (1 << 23)
+#define KWGBE_FRAME_TYPE_IP_V_4 (1 << 24)
+#define KWGBE_FRAME_HEADER_OK (1 << 25)
+#define KWGBE_RX_LAST_DESC (1 << 26)
+#define KWGBE_RX_FIRST_DESC (1 << 27)
+#define KWGBE_UNKNOWN_DESTINATION_ADDR (1 << 28)
+#define KWGBE_RX_EN_INTERRUPT (1 << 29)
+#define KWGBE_LAYER_4_CHECKSUM_OK (1 << 30)
+
+/* Rx descriptors byte count */
+#define KWGBE_FRAME_FRAGMENTED (1 << 2)
+
+/* Tx descriptors command */
+#define KWGBE_LAYER_4_CHECKSUM_FIRST_DESC (1 << 10)
+#define KWGBE_FRAME_SET_TO_VLAN (1 << 15)
+#define KWGBE_TCP_FRAME 0
+#define KWGBE_UDP_FRAME (1 << 16)
+#define KWGBE_GEN_TCP_UDP_CHECKSUM (1 << 17)
+#define KWGBE_GEN_IP_V_4_CHECKSUM (1 << 18)
+#define KWGBE_ZERO_PADDING (1 << 19)
+#define KWGBE_TX_LAST_DESC (1 << 20)
+#define KWGBE_TX_FIRST_DESC (1 << 21)
+#define KWGBE_GEN_CRC (1 << 22)
+#define KWGBE_TX_EN_INTERRUPT (1 << 23)
+#define KWGBE_AUTO_MODE (1 << 30)
+
+/* Address decode parameters */
+/* Ethernet Base Address Register bits */
+#define EBAR_TARGET_DRAM 0x00000000
+#define EBAR_TARGET_DEVICE 0x00000001
+#define EBAR_TARGET_CBS 0x00000002
+#define EBAR_TARGET_PCI0 0x00000003
+#define EBAR_TARGET_PCI1 0x00000004
+#define EBAR_TARGET_CUNIT 0x00000005
+#define EBAR_TARGET_AUNIT 0x00000006
+#define EBAR_TARGET_GUNIT 0x00000007
+
+/* Window attrib */
+#define EBAR_ATTR_DRAM_CS0 0x00000E00
+#define EBAR_ATTR_DRAM_CS1 0x00000D00
+#define EBAR_ATTR_DRAM_CS2 0x00000B00
+#define EBAR_ATTR_DRAM_CS3 0x00000700
+
+/* DRAM Target interface */
+#define EBAR_ATTR_DRAM_NO_CACHE_COHERENCY 0x00000000
+#define EBAR_ATTR_DRAM_CACHE_COHERENCY_WT 0x00001000
+#define EBAR_ATTR_DRAM_CACHE_COHERENCY_WB 0x00002000
+
+/* Device Bus Target interface */
+#define EBAR_ATTR_DEVICE_DEVCS0 0x00001E00
+#define EBAR_ATTR_DEVICE_DEVCS1 0x00001D00
+#define EBAR_ATTR_DEVICE_DEVCS2 0x00001B00
+#define EBAR_ATTR_DEVICE_DEVCS3 0x00001700
+#define EBAR_ATTR_DEVICE_BOOTCS3 0x00000F00
+
+/* PCI Target interface */
+#define EBAR_ATTR_PCI_BYTE_SWAP 0x00000000
+#define EBAR_ATTR_PCI_NO_SWAP 0x00000100
+#define EBAR_ATTR_PCI_BYTE_WORD_SWAP 0x00000200
+#define EBAR_ATTR_PCI_WORD_SWAP 0x00000300
+#define EBAR_ATTR_PCI_NO_SNOOP_NOT_ASSERT 0x00000000
+#define EBAR_ATTR_PCI_NO_SNOOP_ASSERT 0x00000400
+#define EBAR_ATTR_PCI_IO_SPACE 0x00000000
+#define EBAR_ATTR_PCI_MEMORY_SPACE 0x00000800
+#define EBAR_ATTR_PCI_REQ64_FORCE 0x00000000
+#define EBAR_ATTR_PCI_REQ64_SIZE 0x00001000
+
+/* Window access control */
+#define EWIN_ACCESS_NOT_ALLOWED 0
+#define EWIN_ACCESS_READ_ONLY 1
+#define EWIN_ACCESS_FULL ((1 << 1) | 1)
+
+/* SDMA command macros */
+#define CURR_RFD_GET(p_curr_desc, queue) \
+ ((p_curr_desc) = p_gbe_pctrl->p_rx_curr_desc_q[queue])
+
+#define CURR_RFD_SET(p_curr_desc, queue) \
+ (p_gbe_pctrl->p_rx_curr_desc_q[queue] = (p_curr_desc))
+
+#define USED_RFD_GET(p_used_desc, queue) \
+ ((p_used_desc) = p_gbe_pctrl->p_rx_used_desc_q[queue])
+
+#define USED_RFD_SET(p_used_desc, queue)\
+(p_gbe_pctrl->p_rx_used_desc_q[queue] = (p_used_desc))
+
+#define CURR_TFD_GET(p_curr_desc, queue) \
+ ((p_curr_desc) = p_gbe_pctrl->p_tx_curr_desc_q[queue])
+
+#define CURR_TFD_SET(p_curr_desc, queue) \
+ (p_gbe_pctrl->p_tx_curr_desc_q[queue] = (p_curr_desc))
+
+#define USED_TFD_GET(p_used_desc, queue) \
+ ((p_used_desc) = p_gbe_pctrl->p_tx_used_desc_q[queue])
+
+#define USED_TFD_SET(p_used_desc, queue) \
+ (p_gbe_pctrl->p_tx_used_desc_q[queue] = (p_used_desc))
+
+#define FIRST_TFD_GET(p_first_desc, queue) \
+ ((p_first_desc) = p_gbe_pctrl->p_tx_first_desc_q[queue])
+
+#define FIRST_TFD_SET(p_first_desc, queue) \
+ (p_gbe_pctrl->p_tx_first_desc_q[queue] = (p_first_desc))
+
+/* Macros that save access to desc in order to find next desc pointer */
+#define RX_NEXT_DESC_PTR(p_rx_desc, queue) ( \
+ struct kwgbe_rx_dscr*)(((((u32)p_rx_desc - \
+ (u32)p_gbe_pctrl->p_rx_desc_area_base[queue]) + \
+ RX_DESC_ALIGNED_SIZE) % p_gbe_pctrl->rx_desc_area_size[queue]) + \
+ (u32)p_gbe_pctrl->p_rx_desc_area_base[queue])
+
+#define TX_NEXT_DESC_PTR(p_tx_desc, queue) ( \
+ struct kwgbe_tx_dscr*)(((((u32)p_tx_desc - \
+ (u32)p_gbe_pctrl->p_tx_desc_area_base[queue]) + \
+ TX_DESC_ALIGNED_SIZE) % p_gbe_pctrl->tx_desc_area_size[queue]) + \
+ (u32)p_gbe_pctrl->p_tx_desc_area_base[queue])
+
+enum kwgbe_fret_sts {
+ KWGBE_OK, /* Returned as expected. */
+ KWGBE_ERROR, /* Fundamental error. */
+ KWGBE_RETRY, /* Could not process request. Try later. */
+ KWGBE_END_OF_JOB, /* Ring has nothing to process. */
+ KWGBE_QFULL, /* Ring resource error. */
+ KWGBE_QLAST_RESO /* Ring resources about to exhaust. */
+};
+
+enum kwgbe_q {
+ KWGBE_Q0 = 0,
+ KWGBE_Q1 = 1,
+ KWGBE_Q2 = 2,
+ KWGBE_Q3 = 3,
+ KWGBE_Q4 = 4,
+ KWGBE_Q5 = 5,
+ KWGBE_Q6 = 6,
+ KWGBE_Q7 = 7
+};
+
+enum kwgbe_adrwin {
+ KWGBE_WIN0,
+ KWGBE_WIN1,
+ KWGBE_WIN2,
+ KWGBE_WIN3,
+ KWGBE_WIN4,
+ KWGBE_WIN5
+};
+
+enum kwgbe_target {
+ KWGBE_TARGET_DRAM,
+ KWGBE_TARGET_DEV,
+ KWGBE_TARGET_CBS,
+ KWGBE_TARGET_PCI0,
+ KWGBE_TARGET_PCI1
+};
+
+struct kwgbe_rx_dscr {
+ u32 cmd_sts; /* Descriptor command status */
+ u16 buf_size; /* Buffer size */
+ u16 byte_cnt; /* Descriptor buffer byte count */
+ u32 buf_ptr; /* Descriptor buffer pointer */
+ u32 nxtdesc_p; /* Next descriptor pointer */
+ u32 retinfo; /* User resource return information */
+};
+
+struct kwgbe_tx_dscr {
+ u32 cmd_sts; /* Descriptor command status */
+ u16 l4i_chk; /* CPU provided TCP Checksum */
+ u16 byte_cnt; /* Descriptor buffer byte count */
+ u32 buf_ptr; /* Descriptor buffer pointer */
+ u32 nxtdesc_p; /* Next descriptor pointer */
+ u32 retinfo; /* User resource return information */
+};
+
+/*
+ * Unified struct for Rx and Tx operations. The user is not required to
+ * be familier with neither Tx nor Rx descriptors.
+ */
+struct kwgbe_pktinf {
+ u32 cmd_sts; /* Descriptor command status */
+ u16 byte_cnt; /* Descriptor buffer byte count */
+ u16 l4i_chk; /* Tx CPU provided TCP Checksum */
+ u32 buf_ptr; /* Descriptor buffer pointer */
+ u32 retinfo; /* User resource return information */
+};
+
+struct kwgbe_winparam {
+ enum kwgbe_adrwin win; /* Window number */
+ enum kwgbe_target target; /* System targets */
+ u16 attrib; /* BAR attrib. See above macros */
+ u32 base_addr; /* Window base address in u32 form */
+ u32 high_addr; /* Window high address in u32 form */
+ u32 size; /* Size in MBytes. Must be % 64Kbyte. */
+ int enable; /* Enable/disable access to the window. */
+ u16 access_ctrl; /*Access ctrl register. see above macros */
+};
+
+/* Ethernet port specific infomation */
+struct kwgbe_portinfo {
+ int port_phy_addr; /* User phy address of Ethrnet port */
+ u8 port_mac_addr[6]; /* User defined port MAC address. */
+ u32 port_config; /* User port configuration value */
+ u32 port_config_extend; /* User port config extend value */
+ u32 port_sdma_config; /* User port SDMA config value */
+ u32 port_serial_control; /* User port serial control value */
+ u32 port_tx_queue_command; /* Port active Tx queues summary */
+ u32 port_rx_queue_command; /* Port active Rx queues summary */
+
+ /* User function to cast virtual address to CPU bus address */
+ u32(*port_virt_to_phys) (u32 addr);
+ /* User scratch pad for user specific data structures */
+ void *port_private;
+
+ int rx_resource_err[MAX_RXQ_NO]; /* Rx ring resource err flag */
+ int tx_resource_err[MAX_TXQ_NO]; /* Tx ring resource err flag */
+
+ /* Tx/Rx rings managment indexes fields. For driver use */
+
+ /* Next available Rx resource */
+ volatile struct kwgbe_rx_dscr *p_rx_curr_desc_q[MAX_RXQ_NO];
+ /* Returning Rx resource */
+ volatile struct kwgbe_rx_dscr *p_rx_used_desc_q[MAX_RXQ_NO];
+
+ /* Next available Tx resource */
+ volatile struct kwgbe_tx_dscr *p_tx_curr_desc_q[MAX_TXQ_NO];
+ /* Returning Tx resource */
+ volatile struct kwgbe_tx_dscr *p_tx_used_desc_q[MAX_TXQ_NO];
+ /* An extra Tx index to support txmt of multiple buffs per packet */
+ volatile struct kwgbe_tx_dscr *p_tx_first_desc_q[MAX_TXQ_NO];
+
+ /* Tx/Rx rings size and base variables fields. For driver use */
+
+ volatile struct kwgbe_rx_dscr *p_rx_desc_area_base[MAX_RXQ_NO];
+ u32 rx_desc_area_size[MAX_RXQ_NO];
+ char *p_rx_buffer_base[MAX_RXQ_NO];
+
+ volatile struct kwgbe_tx_dscr *p_tx_desc_area_base[MAX_TXQ_NO];
+ u32 tx_desc_area_size[MAX_TXQ_NO];
+ char *p_tx_buffer_base[MAX_TXQ_NO];
+};
+
+/* port device data struct */
+struct kwgbe_device {
+ struct eth_device dev;
+ struct kwgbe_portinfo portinfo;
+ struct egiga_priv portpriv;
+ struct kwgbe_registers *regs;
+};
+
+/* functions */
+static void port_uc_addr_set(struct kwgbe_registers *regs, u8 * p_addr,
+ enum kwgbe_q queue);
+static void port_reset(struct kwgbe_registers *regs);
+
+#endif /* __EGIGA_H__ */
diff --git a/include/netdev.h b/include/netdev.h
index 68d664c..932d986 100644
--- a/include/netdev.h
+++ b/include/netdev.h
@@ -80,6 +80,7 @@ int fec_initialize (bd_t *bis);
int greth_initialize(bd_t *bis);
void gt6426x_eth_initialize(bd_t *bis);
int inca_switch_initialize(bd_t *bis);
+int kirkwood_egiga_initialize(bd_t *bis);
int macb_eth_initialize(int id, void *regs, unsigned int phy_addr);
int mcdmafec_initialize(bd_t *bis);
int mcffec_initialize(bd_t *bis);
--
1.5.3.4
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-05-04 17:56 [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs Prafulla Wadaskar
@ 2009-05-12 4:03 ` Prafulla Wadaskar
2009-05-12 7:07 ` Ben Warren
0 siblings, 1 reply; 10+ messages in thread
From: Prafulla Wadaskar @ 2009-05-12 4:03 UTC (permalink / raw)
To: u-boot
Hi All
Ping....
I didn't here any feedback from anyone, waiting for your valued comments :-(
Regards..
Prafulla . .
> -----Original Message-----
> From: Prafulla Wadaskar [mailto:prafulla at marvell.com]
> Sent: Monday, May 04, 2009 11:27 PM
> To: u-boot at lists.denx.de
> Cc: Prabhanjan Sarnaik; Ronen Shitrit; Ashish Karkare;
> Prafulla Wadaskar
> Subject: [PATCH v2] Gbe Controller driver support for kirkwood SOCs
>
> Contributors:
> Yotam Admon <yotam@marvell.com>
> Michael Blostein <michaelbl@marvell.com
>
> Reviewed by: Ronen Shitrit <rshitrit@marvell.com>
> Signed-off-by: Prafulla Wadaskar <prafulla@marvell.com>
> ---
> Change log:
> v2: entire rewrite/restructure of v1
> used small names for variable/function names
> readl/writel used to access SoC registers
> Soc registers accssed using pointers through net device struct
> miiphy registration done for external smi read/write access
> miiphy_link used to detect phy link presence
> cleaned for cosmetic changes
>
> cpu/arm926ejs/kirkwood/cpu.c | 8 +
> drivers/net/Makefile | 1 +
> drivers/net/kirkwood_egiga.c | 1669
> ++++++++++++++++++++++++++++++++++++++++++
> drivers/net/kirkwood_egiga.h | 828 +++++++++++++++++++++
> include/netdev.h | 1 +
> 5 files changed, 2507 insertions(+), 0 deletions(-)
> create mode 100644 drivers/net/kirkwood_egiga.c
> create mode 100644 drivers/net/kirkwood_egiga.h
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-05-12 4:03 ` Prafulla Wadaskar
@ 2009-05-12 7:07 ` Ben Warren
2009-05-13 7:02 ` Prafulla Wadaskar
2009-05-25 5:36 ` Prafulla Wadaskar
0 siblings, 2 replies; 10+ messages in thread
From: Ben Warren @ 2009-05-12 7:07 UTC (permalink / raw)
To: u-boot
Hi Prafulla,
Prafulla Wadaskar wrote:
> Hi All
>
> Ping....
> I didn't here any feedback from anyone, waiting for your valued comments :-(
>
>
I've started reviewing this driver - it's a lot to digest (2500
lines!?!). Please think a little about what features are really needed
here. For example, do you really need RMON in a bootloader? I'll try
to provide more specific feedback soon.
> Regards..
> Prafulla . .
>
>
regards,
Ben
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-05-12 7:07 ` Ben Warren
@ 2009-05-13 7:02 ` Prafulla Wadaskar
2009-05-25 5:36 ` Prafulla Wadaskar
1 sibling, 0 replies; 10+ messages in thread
From: Prafulla Wadaskar @ 2009-05-13 7:02 UTC (permalink / raw)
To: u-boot
> Prafulla Wadaskar wrote:
> > Hi All
> >
> > Ping....
> > I didn't here any feedback from anyone, waiting for your valued
> > comments :-(
> >
> >
> I've started reviewing this driver - it's a lot to digest
> (2500 lines!?!). Please think a little about what features
> are really needed here. For example, do you really need RMON
> in a bootloader? I'll try to provide more specific feedback soon.
I will say YES here :-)
I have used it for network debugging and very useful during bring up activity.
But for normal operation it is not required
Regards..
Prafulla . .
> >
> regards,
> Ben
>
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-05-12 7:07 ` Ben Warren
2009-05-13 7:02 ` Prafulla Wadaskar
@ 2009-05-25 5:36 ` Prafulla Wadaskar
2009-05-25 20:03 ` Ben Warren
2009-05-26 6:37 ` Ben Warren
1 sibling, 2 replies; 10+ messages in thread
From: Prafulla Wadaskar @ 2009-05-25 5:36 UTC (permalink / raw)
To: u-boot
Hi Ben,
How are you?
Seems to be very busy as usual.... :-)
Can you pls kindly take out some time to review these patches.
hopefully they are as per your requirements since we have invested considerable time to make them up to the mark.
May be-you can accept them :-)
http://lists.denx.de/pipermail/u-boot/2009-May/053111.html
http://lists.denx.de/pipermail/u-boot/2009-May/052905.html
Just to share with you,
in netdev.h- though struct and enums defined looks similar-
I have used separate struct and enums for both mv88e61xx and mv88e1116 because only basic init support is available in those drives, but in coming time there will be more features added to them.
I think since these are initial drivers in net/phy/ we should provide common/phy.h and keep netdev.h untouched
What do you think?
Also you might be ready with review comments for this
http://lists.denx.de/pipermail/u-boot/2009-May/053113.html
Nothing much changed in v4 compared to patch version v1.
few global comments that I received for this patch from you are-
1. the code is too big.
2. can you strip down features to minimum
Just to reply these:
the entire egiga driver is kept as separate driver .c and .h,
I am in progress to reuse the code from other drivers.
The features are kept minimal required but I can cut down some of features like "update network statistics" so as to make code size small and acceptable :-)
Still your comments will help me to provide better code...
Last but not least
Shall I resend above patches if you feel this is more comfortable for you?
Regards..
Prafulla. . .
> -----Original Message-----
> From: Ben Warren [mailto:biggerbadderben at gmail.com]
> Sent: Tuesday, May 12, 2009 12:38 PM
> To: Prafulla Wadaskar
> Cc: u-boot at lists.denx.de; Ashish Karkare; Prabhanjan Sarnaik;
> Ronen Shitrit
> Subject: Re: [U-Boot] [PATCH v2] Gbe Controller driver
> support for kirkwood SOCs
>
> Hi Prafulla,
>
> Prafulla Wadaskar wrote:
> > Hi All
> >
> > Ping....
> > I didn't here any feedback from anyone, waiting for your valued
> > comments :-(
> >
> >
> I've started reviewing this driver - it's a lot to digest
> (2500 lines!?!). Please think a little about what features
> are really needed here. For example, do you really need RMON
> in a bootloader? I'll try to provide more specific feedback soon.
> > Regards..
> > Prafulla . .
> >
> >
> regards,
> Ben
>
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-05-25 5:36 ` Prafulla Wadaskar
@ 2009-05-25 20:03 ` Ben Warren
2009-05-26 6:37 ` Ben Warren
1 sibling, 0 replies; 10+ messages in thread
From: Ben Warren @ 2009-05-25 20:03 UTC (permalink / raw)
To: u-boot
Hi Prafulla,
On Sun, May 24, 2009 at 10:36 PM, Prafulla Wadaskar <prafulla@marvell.com>wrote:
> Hi Ben,
>
> How are you?
Not bad, thanks for asking!
>
> Seems to be very busy as usual.... :-)
>
Yeah, very busy but that's still not a good excuse. I usually respond
quickly to small changes but not so much with the big ones because I want to
make sure they get a thorough review. I'm planning on spending a few hours
tonight going through my queue, and your patches are prominent.
regards,
Ben
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-05-25 5:36 ` Prafulla Wadaskar
2009-05-25 20:03 ` Ben Warren
@ 2009-05-26 6:37 ` Ben Warren
2009-06-05 10:12 ` Wolfgang Denk
1 sibling, 1 reply; 10+ messages in thread
From: Ben Warren @ 2009-05-26 6:37 UTC (permalink / raw)
To: u-boot
Hi Prafulla,
Prafulla Wadaskar wrote:
> Hi Ben,
>
> How are you?
> Seems to be very busy as usual.... :-)
>
> Can you pls kindly take out some time to review these patches.
> hopefully they are as per your requirements since we have invested considerable time to make them up to the mark.
> May be-you can accept them :-)
> http://lists.denx.de/pipermail/u-boot/2009-May/053111.html
>
We don't currently have a proper PHY framework in place, although over
the years there's been much talk in this area. For now I'll pass on
this one.
> http://lists.denx.de/pipermail/u-boot/2009-May/052905.html
>
I've pulled this into net/next. Thanks for spinning it so many times.
> Just to share with you,
> in netdev.h- though struct and enums defined looks similar-
> I have used separate struct and enums for both mv88e61xx and mv88e1116 because only basic init support is available in those drives, but in coming time there will be more features added to them.
> I think since these are initial drivers in net/phy/ we should provide common/phy.h and keep netdev.h untouched
> What do you think?
>
>
As I mentioned above, the existing PHY infrastructure is weak. The
switch driver is fine here, but I don't want to add any regular PHYs
without a properly designed interface. A lot of work has been done in
this area but hasn't been published yet. Stay tuned...
> Also you might be ready with review comments for this
> http://lists.denx.de/pipermail/u-boot/2009-May/053113.html
> Nothing much changed in v4 compared to patch version v1.
> few global comments that I received for this patch from you are-
> 1. the code is too big.
> 2. can you strip down features to minimum
> Just to reply these:
> the entire egiga driver is kept as separate driver .c and .h,
> I am in progress to reuse the code from other drivers.
> The features are kept minimal required but I can cut down some of features like "update network statistics" so as to make code size small and acceptable :-)
> Still your comments will help me to provide better code...
>
>
The driver is really big, which isn't necessarily a show stopper. I'd
like to see you #ifdef CONFIG_xxx out the non-essential stuff
(statistics etc.) so that users aren't forced to compile it in. Also,
please put the changes to cpu.c into a different patch.
> Last but not least
> Shall I resend above patches if you feel this is more comfortable for you?
>
> Regards..
> Prafulla. . .
>
>
Thanks for being so flexible.
regards,
Ben
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-05-26 6:37 ` Ben Warren
@ 2009-06-05 10:12 ` Wolfgang Denk
2009-06-05 16:34 ` Ben Warren
0 siblings, 1 reply; 10+ messages in thread
From: Wolfgang Denk @ 2009-06-05 10:12 UTC (permalink / raw)
To: u-boot
Dear Ben,
In message <4A1B8E3F.7030207@gmail.com> you wrote:
>
> > http://lists.denx.de/pipermail/u-boot/2009-May/052905.html
> >
> I've pulled this into net/next. Thanks for spinning it so many times.
Can you please send a pull request for next soon? Thanks in advance.
Best regards,
Wolfgang Denk
--
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
C++ is the best example of second-system effect since OS/360.
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-06-05 10:12 ` Wolfgang Denk
@ 2009-06-05 16:34 ` Ben Warren
2009-06-08 4:28 ` Prafulla Wadaskar
0 siblings, 1 reply; 10+ messages in thread
From: Ben Warren @ 2009-06-05 16:34 UTC (permalink / raw)
To: u-boot
Wolfgang Denk wrote:
> Dear Ben,
>
> In message <4A1B8E3F.7030207@gmail.com> you wrote:
>
>>> http://lists.denx.de/pipermail/u-boot/2009-May/052905.html
>>>
>>>
>> I've pulled this into net/next. Thanks for spinning it so many times.
>>
>
> Can you please send a pull request for next soon? Thanks in advance.
>
> Best regards,
>
> Wolfgang Denk
>
>
You bet.
Ben
^ permalink raw reply [flat|nested] 10+ messages in thread
* [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs
2009-06-05 16:34 ` Ben Warren
@ 2009-06-08 4:28 ` Prafulla Wadaskar
0 siblings, 0 replies; 10+ messages in thread
From: Prafulla Wadaskar @ 2009-06-08 4:28 UTC (permalink / raw)
To: u-boot
> -----Original Message-----
> From: Ben Warren [mailto:biggerbadderben at gmail.com]
> Sent: Friday, June 05, 2009 10:04 PM
> To: Wolfgang Denk
> Cc: Prafulla Wadaskar; u-boot at lists.denx.de; Ashish Karkare;
> Prabhanjan Sarnaik; Ronen Shitrit
> Subject: Re: [U-Boot] [PATCH v2] Gbe Controller driver
> support for kirkwood SOCs
>
> Wolfgang Denk wrote:
> > Dear Ben,
> >
> > In message <4A1B8E3F.7030207@gmail.com> you wrote:
> >
> >>> http://lists.denx.de/pipermail/u-boot/2009-May/052905.html
> >>>
> >>>
> >> I've pulled this into net/next. Thanks for spinning it so
> many times.
> >>
> >
> > Can you please send a pull request for next soon? Thanks in advance.
> >
> > Best regards,
> >
> > Wolfgang Denk
> >
> >
> You bet.
May be it is little bit confusing.
I think Wolfgang is talking with reference to Marvell MV88E61XX Switch Driver which is already in next branch.
Whereas subject for this mail is different.
We wish to see MV88E61XX driver in, there are some other users for this too..
Regards..
Prafulla . .
>
> Ben
>
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2009-06-08 4:28 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-05-04 17:56 [U-Boot] [PATCH v2] Gbe Controller driver support for kirkwood SOCs Prafulla Wadaskar
2009-05-12 4:03 ` Prafulla Wadaskar
2009-05-12 7:07 ` Ben Warren
2009-05-13 7:02 ` Prafulla Wadaskar
2009-05-25 5:36 ` Prafulla Wadaskar
2009-05-25 20:03 ` Ben Warren
2009-05-26 6:37 ` Ben Warren
2009-06-05 10:12 ` Wolfgang Denk
2009-06-05 16:34 ` Ben Warren
2009-06-08 4:28 ` Prafulla Wadaskar
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.