Blackfin ethernet driver: on chip ethernet MAC controller driver
Bryan Wu [Tue, 17 Jul 2007 06:43:44 +0000 (14:43 +0800)]
This patch implements the driver necessary use the Analog Devices
Blackfin processor's on-chip ethernet MAC controller.

[try#2]
 - add timeout control
 - kill dma_config_reg bitfields
 - some trivial cleanup

[try#3]
 - add endianess check
 - add DRV_NAME, DRV_VERSION... driver information string
 - add some comments for silicon anomaly and dma API confusion
 - some code trivial cleanup

[try#4]
 - add Blackfin latest GPIO pin mux opertion with Michael Hennerich's
  help and Dan's review
 - rewrite the DMA descriptor list operation in a more readable way
  by Joe's review

[try#5]
 - cleanup some coding style by Joe's review.

[try#6]
 - 1.1 version fix a bug when set up multicast list pointed by Mr. yoshfuji
 - rearrange the desc_list_free function.

Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Cc: Michael Buesch <mb@bu3sch.de>
Cc: Mike Frysinger <vapier.adi@gmail.com>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dan Williams <dcbw@redhat.com>
Cc: Joe Perches <joe@perches.com>
Cc: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

MAINTAINERS
drivers/net/Kconfig
drivers/net/Makefile
drivers/net/bfin_mac.c [new file with mode: 0644]
drivers/net/bfin_mac.h [new file with mode: 0644]

index e78f62f..f6b2665 100644 (file)
@@ -732,6 +732,13 @@ L: uclinux-dist-devel@blackfin.uclinux.org (subscribers-only)
 W:     http://blackfin.uclinux.org
 S:     Supported
 
+BLACKFIN EMAC DRIVER
+P:     Bryan Wu
+M:     bryan.wu@analog.com
+L:     uclinux-dist-devel@blackfin.uclinux.org (subscribers-only)
+W:     http://blackfin.uclinux.org
+S:     Supported
+
 BLACKFIN RTC DRIVER
 P:     Mike Frysinger
 M:     michael.frysinger@analog.com
index 5fb659f..3073f67 100644 (file)
@@ -838,6 +838,50 @@ config ULTRA32
          <file:Documentation/networking/net-modules.txt>. The module
          will be called smc-ultra32.
 
+config BFIN_MAC
+       tristate "Blackfin 536/537 on-chip mac support"
+       depends on NET_ETHERNET && (BF537 || BF536) && (!BF537_PORT_H)
+       select CRC32
+       select BFIN_MAC_USE_L1 if DMA_UNCACHED_NONE
+       help
+         This is the driver for blackfin on-chip mac device. Say Y if you want it
+         compiled into the kernel. This driver is also available as a module
+         ( = code which can be inserted in and removed from the running kernel
+         whenever you want). The module will be called bfin_mac.
+
+config BFIN_MAC_USE_L1
+       bool "Use L1 memory for rx/tx packets"
+       depends on BFIN_MAC && BF537
+       default y
+       help
+         To get maximum network performace, you should use L1 memory as rx/tx buffers.
+         Say N here if you want to reserve L1 memory for other uses.
+
+config BFIN_TX_DESC_NUM
+       int "Number of transmit buffer packets"
+       depends on BFIN_MAC
+       range 6 10 if BFIN_MAC_USE_L1
+       range 10 100
+       default "10"
+       help
+         Set the number of buffer packets used in driver.
+
+config BFIN_RX_DESC_NUM
+       int "Number of receive buffer packets"
+       depends on BFIN_MAC
+       range 20 100 if BFIN_MAC_USE_L1
+       range 20 800
+       default "20"
+       help
+         Set the number of buffer packets used in driver.
+
+config BFIN_MAC_RMII
+       bool "RMII PHY Interface (EXPERIMENTAL)"
+       depends on BFIN_MAC && EXPERIMENTAL
+       default n
+       help
+         Use Reduced PHY MII Interface
+
 config SMC9194
        tristate "SMC 9194 support"
        depends on NET_VENDOR_SMC && (ISA || MAC && BROKEN)
index 0e286ab..aaaa0a0 100644 (file)
@@ -200,6 +200,7 @@ obj-$(CONFIG_S2IO) += s2io.o
 obj-$(CONFIG_MYRI10GE) += myri10ge/
 obj-$(CONFIG_SMC91X) += smc91x.o
 obj-$(CONFIG_SMC911X) += smc911x.o
+obj-$(CONFIG_BFIN_MAC) += bfin_mac.o
 obj-$(CONFIG_DM9000) += dm9000.o
 obj-$(CONFIG_FEC_8XX) += fec_8xx/
 obj-$(CONFIG_PASEMI_MAC) += pasemi_mac.o
diff --git a/drivers/net/bfin_mac.c b/drivers/net/bfin_mac.c
new file mode 100644 (file)
index 0000000..9a08d65
--- /dev/null
@@ -0,0 +1,1009 @@
+/*
+ * File:       drivers/net/bfin_mac.c
+ * Based on:
+ * Maintainer:
+ *             Bryan Wu <bryan.wu@analog.com>
+ *
+ * Original author:
+ *             Luke Yang <luke.yang@analog.com>
+ *
+ * Created:
+ * Description:
+ *
+ * Modified:
+ *             Copyright 2004-2006 Analog Devices Inc.
+ *
+ * Bugs:       Enter bugs at http://blackfin.uclinux.org/
+ *
+ * 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, 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 ;  see the file COPYING.
+ * If not, write to the Free Software Foundation,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/timer.h>
+#include <linux/errno.h>
+#include <linux/irq.h>
+#include <linux/io.h>
+#include <linux/ioport.h>
+#include <linux/crc32.h>
+#include <linux/device.h>
+#include <linux/spinlock.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+
+#include <linux/platform_device.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+
+#include <asm/dma.h>
+#include <linux/dma-mapping.h>
+
+#include <asm/blackfin.h>
+#include <asm/cacheflush.h>
+#include <asm/portmux.h>
+
+#include "bfin_mac.h"
+
+#define DRV_NAME       "bfin_mac"
+#define DRV_VERSION    "1.1"
+#define DRV_AUTHOR     "Bryan Wu, Luke Yang"
+#define DRV_DESC       "Blackfin BF53[67] on-chip Ethernet MAC driver"
+
+MODULE_AUTHOR(DRV_AUTHOR);
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION(DRV_DESC);
+
+#if defined(CONFIG_BFIN_MAC_USE_L1)
+# define bfin_mac_alloc(dma_handle, size)  l1_data_sram_zalloc(size)
+# define bfin_mac_free(dma_handle, ptr)    l1_data_sram_free(ptr)
+#else
+# define bfin_mac_alloc(dma_handle, size) \
+       dma_alloc_coherent(NULL, size, dma_handle, GFP_KERNEL)
+# define bfin_mac_free(dma_handle, ptr) \
+       dma_free_coherent(NULL, sizeof(*ptr), ptr, dma_handle)
+#endif
+
+#define PKT_BUF_SZ 1580
+
+#define MAX_TIMEOUT_CNT        500
+
+/* pointers to maintain transmit list */
+static struct net_dma_desc_tx *tx_list_head;
+static struct net_dma_desc_tx *tx_list_tail;
+static struct net_dma_desc_rx *rx_list_head;
+static struct net_dma_desc_rx *rx_list_tail;
+static struct net_dma_desc_rx *current_rx_ptr;
+static struct net_dma_desc_tx *current_tx_ptr;
+static struct net_dma_desc_tx *tx_desc;
+static struct net_dma_desc_rx *rx_desc;
+
+static void desc_list_free(void)
+{
+       struct net_dma_desc_rx *r;
+       struct net_dma_desc_tx *t;
+       int i;
+#if !defined(CONFIG_BFIN_MAC_USE_L1)
+       dma_addr_t dma_handle = 0;
+#endif
+
+       if (tx_desc) {
+               t = tx_list_head;
+               for (i = 0; i < CONFIG_BFIN_TX_DESC_NUM; i++) {
+                       if (t) {
+                               if (t->skb) {
+                                       dev_kfree_skb(t->skb);
+                                       t->skb = NULL;
+                               }
+                               t = t->next;
+                       }
+               }
+               bfin_mac_free(dma_handle, tx_desc);
+       }
+
+       if (rx_desc) {
+               r = rx_list_head;
+               for (i = 0; i < CONFIG_BFIN_RX_DESC_NUM; i++) {
+                       if (r) {
+                               if (r->skb) {
+                                       dev_kfree_skb(r->skb);
+                                       r->skb = NULL;
+                               }
+                               r = r->next;
+                       }
+               }
+               bfin_mac_free(dma_handle, rx_desc);
+       }
+}
+
+static int desc_list_init(void)
+{
+       int i;
+       struct sk_buff *new_skb;
+#if !defined(CONFIG_BFIN_MAC_USE_L1)
+       /*
+        * This dma_handle is useless in Blackfin dma_alloc_coherent().
+        * The real dma handler is the return value of dma_alloc_coherent().
+        */
+       dma_addr_t dma_handle;
+#endif
+
+       tx_desc = bfin_mac_alloc(&dma_handle,
+                               sizeof(struct net_dma_desc_tx) *
+                               CONFIG_BFIN_TX_DESC_NUM);
+       if (tx_desc == NULL)
+               goto init_error;
+
+       rx_desc = bfin_mac_alloc(&dma_handle,
+                               sizeof(struct net_dma_desc_rx) *
+                               CONFIG_BFIN_RX_DESC_NUM);
+       if (rx_desc == NULL)
+               goto init_error;
+
+       /* init tx_list */
+       tx_list_head = tx_list_tail = tx_desc;
+
+       for (i = 0; i < CONFIG_BFIN_TX_DESC_NUM; i++) {
+               struct net_dma_desc_tx *t = tx_desc + i;
+               struct dma_descriptor *a = &(t->desc_a);
+               struct dma_descriptor *b = &(t->desc_b);
+
+               /*
+                * disable DMA
+                * read from memory WNR = 0
+                * wordsize is 32 bits
+                * 6 half words is desc size
+                * large desc flow
+                */
+               a->config = WDSIZE_32 | NDSIZE_6 | DMAFLOW_LARGE;
+               a->start_addr = (unsigned long)t->packet;
+               a->x_count = 0;
+               a->next_dma_desc = b;
+
+               /*
+                * enabled DMA
+                * write to memory WNR = 1
+                * wordsize is 32 bits
+                * disable interrupt
+                * 6 half words is desc size
+                * large desc flow
+                */
+               b->config = DMAEN | WNR | WDSIZE_32 | NDSIZE_6 | DMAFLOW_LARGE;
+               b->start_addr = (unsigned long)(&(t->status));
+               b->x_count = 0;
+
+               t->skb = NULL;
+               tx_list_tail->desc_b.next_dma_desc = a;
+               tx_list_tail->next = t;
+               tx_list_tail = t;
+       }
+       tx_list_tail->next = tx_list_head;      /* tx_list is a circle */
+       tx_list_tail->desc_b.next_dma_desc = &(tx_list_head->desc_a);
+       current_tx_ptr = tx_list_head;
+
+       /* init rx_list */
+       rx_list_head = rx_list_tail = rx_desc;
+
+       for (i = 0; i < CONFIG_BFIN_RX_DESC_NUM; i++) {
+               struct net_dma_desc_rx *r = rx_desc + i;
+               struct dma_descriptor *a = &(r->desc_a);
+               struct dma_descriptor *b = &(r->desc_b);
+
+               /* allocate a new skb for next time receive */
+               new_skb = dev_alloc_skb(PKT_BUF_SZ + 2);
+               if (!new_skb) {
+                       printk(KERN_NOTICE DRV_NAME
+                              ": init: low on mem - packet dropped\n");
+                       goto init_error;
+               }
+               skb_reserve(new_skb, 2);
+               r->skb = new_skb;
+
+               /*
+                * enabled DMA
+                * write to memory WNR = 1
+                * wordsize is 32 bits
+                * disable interrupt
+                * 6 half words is desc size
+                * large desc flow
+                */
+               a->config = DMAEN | WNR | WDSIZE_32 | NDSIZE_6 | DMAFLOW_LARGE;
+               /* since RXDWA is enabled */
+               a->start_addr = (unsigned long)new_skb->data - 2;
+               a->x_count = 0;
+               a->next_dma_desc = b;
+
+               /*
+                * enabled DMA
+                * write to memory WNR = 1
+                * wordsize is 32 bits
+                * enable interrupt
+                * 6 half words is desc size
+                * large desc flow
+                */
+               b->config = DMAEN | WNR | WDSIZE_32 | DI_EN |
+                               NDSIZE_6 | DMAFLOW_LARGE;
+               b->start_addr = (unsigned long)(&(r->status));
+               b->x_count = 0;
+
+               rx_list_tail->desc_b.next_dma_desc = a;
+               rx_list_tail->next = r;
+               rx_list_tail = r;
+       }
+       rx_list_tail->next = rx_list_head;      /* rx_list is a circle */
+       rx_list_tail->desc_b.next_dma_desc = &(rx_list_head->desc_a);
+       current_rx_ptr = rx_list_head;
+
+       return 0;
+
+init_error:
+       desc_list_free();
+       printk(KERN_ERR DRV_NAME ": kmalloc failed\n");
+       return -ENOMEM;
+}
+
+
+/*---PHY CONTROL AND CONFIGURATION-----------------------------------------*/
+
+/* Set FER regs to MUX in Ethernet pins */
+static int setup_pin_mux(int action)
+{
+#if defined(CONFIG_BFIN_MAC_RMII)
+       u16 pin_req[] = P_RMII0;
+#else
+       u16 pin_req[] = P_MII0;
+#endif
+
+       if (action) {
+               if (peripheral_request_list(pin_req, DRV_NAME)) {
+                       printk(KERN_ERR DRV_NAME
+                       ": Requesting Peripherals failed\n");
+                       return -EFAULT;
+               }
+       } else
+               peripheral_free_list(pin_req);
+
+       return 0;
+}
+
+/* Wait until the previous MDC/MDIO transaction has completed */
+static void poll_mdc_done(void)
+{
+       int timeout_cnt = MAX_TIMEOUT_CNT;
+
+       /* poll the STABUSY bit */
+       while ((bfin_read_EMAC_STAADD()) & STABUSY) {
+               mdelay(10);
+               if (timeout_cnt-- < 0) {
+                       printk(KERN_ERR DRV_NAME
+                       ": wait MDC/MDIO transaction to complete timeout\n");
+                       break;
+               }
+       }
+}
+
+/* Read an off-chip register in a PHY through the MDC/MDIO port */
+static u16 read_phy_reg(u16 PHYAddr, u16 RegAddr)
+{
+       poll_mdc_done();
+       /* read mode */
+       bfin_write_EMAC_STAADD(SET_PHYAD(PHYAddr) |
+                               SET_REGAD(RegAddr) |
+                               STABUSY);
+       poll_mdc_done();
+
+       return (u16) bfin_read_EMAC_STADAT();
+}
+
+/* Write an off-chip register in a PHY through the MDC/MDIO port */
+static void raw_write_phy_reg(u16 PHYAddr, u16 RegAddr, u32 Data)
+{
+       bfin_write_EMAC_STADAT(Data);
+
+       /* write mode */
+       bfin_write_EMAC_STAADD(SET_PHYAD(PHYAddr) |
+                               SET_REGAD(RegAddr) |
+                               STAOP |
+                               STABUSY);
+
+       poll_mdc_done();
+}
+
+static void write_phy_reg(u16 PHYAddr, u16 RegAddr, u32 Data)
+{
+       poll_mdc_done();
+       raw_write_phy_reg(PHYAddr, RegAddr, Data);
+}
+
+/* set up the phy */
+static void bf537mac_setphy(struct net_device *dev)
+{
+       u16 phydat;
+       struct bf537mac_local *lp = netdev_priv(dev);
+
+       /* Program PHY registers */
+       pr_debug("start setting up phy\n");
+
+       /* issue a reset */
+       raw_write_phy_reg(lp->PhyAddr, PHYREG_MODECTL, 0x8000);
+
+       /* wait half a second */
+       msleep(500);
+
+       phydat = read_phy_reg(lp->PhyAddr, PHYREG_MODECTL);
+
+       /* advertise flow control supported */
+       phydat = read_phy_reg(lp->PhyAddr, PHYREG_ANAR);
+       phydat |= (1 << 10);
+       write_phy_reg(lp->PhyAddr, PHYREG_ANAR, phydat);
+
+       phydat = 0;
+       if (lp->Negotiate)
+               phydat |= 0x1000;       /* enable auto negotiation */
+       else {
+               if (lp->FullDuplex)
+                       phydat |= (1 << 8);     /* full duplex */
+               else
+                       phydat &= (~(1 << 8));  /* half duplex */
+
+               if (!lp->Port10)
+                       phydat |= (1 << 13);    /* 100 Mbps */
+               else
+                       phydat &= (~(1 << 13)); /* 10 Mbps */
+       }
+
+       if (lp->Loopback)
+               phydat |= (1 << 14);    /* enable TX->RX loopback */
+
+       write_phy_reg(lp->PhyAddr, PHYREG_MODECTL, phydat);
+       msleep(500);
+
+       phydat = read_phy_reg(lp->PhyAddr, PHYREG_MODECTL);
+       /* check for SMSC PHY */
+       if ((read_phy_reg(lp->PhyAddr, PHYREG_PHYID1) == 0x7) &&
+       ((read_phy_reg(lp->PhyAddr, PHYREG_PHYID2) & 0xfff0) == 0xC0A0)) {
+               /*
+                * we have SMSC PHY so reqest interrupt
+                * on link down condition
+                */
+
+               /* enable interrupts */
+               write_phy_reg(lp->PhyAddr, 30, 0x0ff);
+       }
+}
+
+/**************************************************************************/
+void setup_system_regs(struct net_device *dev)
+{
+       int phyaddr;
+       unsigned short sysctl, phydat;
+       u32 opmode;
+       struct bf537mac_local *lp = netdev_priv(dev);
+       int count = 0;
+
+       phyaddr = lp->PhyAddr;
+
+       /* Enable PHY output */
+       if (!(bfin_read_VR_CTL() & PHYCLKOE))
+               bfin_write_VR_CTL(bfin_read_VR_CTL() | PHYCLKOE);
+
+       /* MDC  = 2.5 MHz */
+       sysctl = SET_MDCDIV(24);
+       /* Odd word alignment for Receive Frame DMA word */
+       /* Configure checksum support and rcve frame word alignment */
+#if defined(BFIN_MAC_CSUM_OFFLOAD)
+       sysctl |= RXDWA | RXCKS;
+#else
+       sysctl |= RXDWA;
+#endif
+       bfin_write_EMAC_SYSCTL(sysctl);
+       /* auto negotiation on  */
+       /* full duplex          */
+       /* 100 Mbps             */
+       phydat = PHY_ANEG_EN | PHY_DUPLEX | PHY_SPD_SET;
+       write_phy_reg(phyaddr, PHYREG_MODECTL, phydat);
+
+       /* test if full duplex supported */
+       do {
+               msleep(100);
+               phydat = read_phy_reg(phyaddr, PHYREG_MODESTAT);
+               if (count > 30) {
+                       printk(KERN_NOTICE DRV_NAME ": Link is down\n");
+                       printk(KERN_NOTICE DRV_NAME
+                                "please check your network connection\n");
+                       break;
+               }
+               count++;
+       } while (!(phydat & 0x0004));
+
+       phydat = read_phy_reg(phyaddr, PHYREG_ANLPAR);
+
+       if ((phydat & 0x0100) || (phydat & 0x0040)) {
+               opmode = FDMODE;
+       } else {
+               opmode = 0;
+               printk(KERN_INFO DRV_NAME
+                       ": Network is set to half duplex\n");
+       }
+
+#if defined(CONFIG_BFIN_MAC_RMII)
+       opmode |= RMII; /* For Now only 100MBit are supported */
+#endif
+
+       bfin_write_EMAC_OPMODE(opmode);
+
+       bfin_write_EMAC_MMC_CTL(RSTC | CROLL);
+
+       /* Initialize the TX DMA channel registers */
+       bfin_write_DMA2_X_COUNT(0);
+       bfin_write_DMA2_X_MODIFY(4);
+       bfin_write_DMA2_Y_COUNT(0);
+       bfin_write_DMA2_Y_MODIFY(0);
+
+       /* Initialize the RX DMA channel registers */
+       bfin_write_DMA1_X_COUNT(0);
+       bfin_write_DMA1_X_MODIFY(4);
+       bfin_write_DMA1_Y_COUNT(0);
+       bfin_write_DMA1_Y_MODIFY(0);
+}
+
+void setup_mac_addr(u8 * mac_addr)
+{
+       u32 addr_low = le32_to_cpu(*(__le32 *) & mac_addr[0]);
+       u16 addr_hi = le16_to_cpu(*(__le16 *) & mac_addr[4]);
+
+       /* this depends on a little-endian machine */
+       bfin_write_EMAC_ADDRLO(addr_low);
+       bfin_write_EMAC_ADDRHI(addr_hi);
+}
+
+static void adjust_tx_list(void)
+{
+       int timeout_cnt = MAX_TIMEOUT_CNT;
+
+       if (tx_list_head->status.status_word != 0
+           && current_tx_ptr != tx_list_head) {
+               goto adjust_head;       /* released something, just return; */
+       }
+
+       /*
+        * if nothing released, check wait condition
+        * current's next can not be the head,
+        * otherwise the dma will not stop as we want
+        */
+       if (current_tx_ptr->next->next == tx_list_head) {
+               while (tx_list_head->status.status_word == 0) {
+                       mdelay(10);
+                       if (tx_list_head->status.status_word != 0
+                           || !(bfin_read_DMA2_IRQ_STATUS() & 0x08)) {
+                               goto adjust_head;
+                       }
+                       if (timeout_cnt-- < 0) {
+                               printk(KERN_ERR DRV_NAME
+                               ": wait for adjust tx list head timeout\n");
+                               break;
+                       }
+               }
+               if (tx_list_head->status.status_word != 0) {
+                       goto adjust_head;
+               }
+       }
+
+       return;
+
+adjust_head:
+       do {
+               tx_list_head->desc_a.config &= ~DMAEN;
+               tx_list_head->status.status_word = 0;
+               if (tx_list_head->skb) {
+                       dev_kfree_skb(tx_list_head->skb);
+                       tx_list_head->skb = NULL;
+               } else {
+                       printk(KERN_ERR DRV_NAME
+                              ": no sk_buff in a transmitted frame!\n");
+               }
+               tx_list_head = tx_list_head->next;
+       } while (tx_list_head->status.status_word != 0
+                && current_tx_ptr != tx_list_head);
+       return;
+
+}
+
+static int bf537mac_hard_start_xmit(struct sk_buff *skb,
+                               struct net_device *dev)
+{
+       struct bf537mac_local *lp = netdev_priv(dev);
+       unsigned int data;
+
+       current_tx_ptr->skb = skb;
+
+       /*
+        * Is skb->data always 16-bit aligned?
+        * Do we need to memcpy((char *)(tail->packet + 2), skb->data, len)?
+        */
+       if ((((unsigned int)(skb->data)) & 0x02) == 2) {
+               /* move skb->data to current_tx_ptr payload */
+               data = (unsigned int)(skb->data) - 2;
+               *((unsigned short *)data) = (unsigned short)(skb->len);
+               current_tx_ptr->desc_a.start_addr = (unsigned long)data;
+               /* this is important! */
+               blackfin_dcache_flush_range(data, (data + (skb->len)) + 2);
+
+       } else {
+               *((unsigned short *)(current_tx_ptr->packet)) =
+                   (unsigned short)(skb->len);
+               memcpy((char *)(current_tx_ptr->packet + 2), skb->data,
+                      (skb->len));
+               current_tx_ptr->desc_a.start_addr =
+                   (unsigned long)current_tx_ptr->packet;
+               if (current_tx_ptr->status.status_word != 0)
+                       current_tx_ptr->status.status_word = 0;
+               blackfin_dcache_flush_range((unsigned int)current_tx_ptr->
+                                           packet,
+                                           (unsigned int)(current_tx_ptr->
+                                                          packet + skb->len) +
+                                           2);
+       }
+
+       /* enable this packet's dma */
+       current_tx_ptr->desc_a.config |= DMAEN;
+
+       /* tx dma is running, just return */
+       if (bfin_read_DMA2_IRQ_STATUS() & 0x08)
+               goto out;
+
+       /* tx dma is not running */
+       bfin_write_DMA2_NEXT_DESC_PTR(&(current_tx_ptr->desc_a));
+       /* dma enabled, read from memory, size is 6 */
+       bfin_write_DMA2_CONFIG(current_tx_ptr->desc_a.config);
+       /* Turn on the EMAC tx */
+       bfin_write_EMAC_OPMODE(bfin_read_EMAC_OPMODE() | TE);
+
+out:
+       adjust_tx_list();
+       current_tx_ptr = current_tx_ptr->next;
+       dev->trans_start = jiffies;
+       lp->stats.tx_packets++;
+       lp->stats.tx_bytes += (skb->len);
+       return 0;
+}
+
+static void bf537mac_rx(struct net_device *dev)
+{
+       struct sk_buff *skb, *new_skb;
+       struct bf537mac_local *lp = netdev_priv(dev);
+       unsigned short len;
+
+       /* allocate a new skb for next time receive */
+       skb = current_rx_ptr->skb;
+       new_skb = dev_alloc_skb(PKT_BUF_SZ + 2);
+       if (!new_skb) {
+               printk(KERN_NOTICE DRV_NAME
+                      ": rx: low on mem - packet dropped\n");
+               lp->stats.rx_dropped++;
+               goto out;
+       }
+       /* reserve 2 bytes for RXDWA padding */
+       skb_reserve(new_skb, 2);
+       current_rx_ptr->skb = new_skb;
+       current_rx_ptr->desc_a.start_addr = (unsigned long)new_skb->data - 2;
+
+       len = (unsigned short)((current_rx_ptr->status.status_word) & RX_FRLEN);
+       skb_put(skb, len);
+       blackfin_dcache_invalidate_range((unsigned long)skb->head,
+                                        (unsigned long)skb->tail);
+
+       dev->last_rx = jiffies;
+       skb->dev = dev;
+       skb->protocol = eth_type_trans(skb, dev);
+#if defined(BFIN_MAC_CSUM_OFFLOAD)
+       skb->csum = current_rx_ptr->status.ip_payload_csum;
+       skb->ip_summed = CHECKSUM_PARTIAL;
+#endif
+
+       netif_rx(skb);
+       lp->stats.rx_packets++;
+       lp->stats.rx_bytes += len;
+       current_rx_ptr->status.status_word = 0x00000000;
+       current_rx_ptr = current_rx_ptr->next;
+
+out:
+       return;
+}
+
+/* interrupt routine to handle rx and error signal */
+static irqreturn_t bf537mac_interrupt(int irq, void *dev_id)
+{
+       struct net_device *dev = dev_id;
+       int number = 0;
+
+get_one_packet:
+       if (current_rx_ptr->status.status_word == 0) {
+               /* no more new packet received */
+               if (number == 0) {
+                       if (current_rx_ptr->next->status.status_word != 0) {
+                               current_rx_ptr = current_rx_ptr->next;
+                               goto real_rx;
+                       }
+               }
+               bfin_write_DMA1_IRQ_STATUS(bfin_read_DMA1_IRQ_STATUS() |
+                                          DMA_DONE | DMA_ERR);
+               return IRQ_HANDLED;
+       }
+
+real_rx:
+       bf537mac_rx(dev);
+       number++;
+       goto get_one_packet;
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void bf537mac_poll(struct net_device *dev)
+{
+       disable_irq(IRQ_MAC_RX);
+       bf537mac_interrupt(IRQ_MAC_RX, dev);
+       enable_irq(IRQ_MAC_RX);
+}
+#endif                         /* CONFIG_NET_POLL_CONTROLLER */
+
+static void bf537mac_reset(void)
+{
+       unsigned int opmode;
+
+       opmode = bfin_read_EMAC_OPMODE();
+       opmode &= (~RE);
+       opmode &= (~TE);
+       /* Turn off the EMAC */
+       bfin_write_EMAC_OPMODE(opmode);
+}
+
+/*
+ * Enable Interrupts, Receive, and Transmit
+ */
+static int bf537mac_enable(struct net_device *dev)
+{
+       u32 opmode;
+
+       pr_debug("%s: %s\n", dev->name, __FUNCTION__);
+
+       /* Set RX DMA */
+       bfin_write_DMA1_NEXT_DESC_PTR(&(rx_list_head->desc_a));
+       bfin_write_DMA1_CONFIG(rx_list_head->desc_a.config);
+
+       /* Wait MII done */
+       poll_mdc_done();
+
+       /* We enable only RX here */
+       /* ASTP   : Enable Automatic Pad Stripping
+          PR     : Promiscuous Mode for test
+          PSF    : Receive frames with total length less than 64 bytes.
+          FDMODE : Full Duplex Mode
+          LB     : Internal Loopback for test
+          RE     : Receiver Enable */
+       opmode = bfin_read_EMAC_OPMODE();
+       if (opmode & FDMODE)
+               opmode |= PSF;
+       else
+               opmode |= DRO | DC | PSF;
+       opmode |= RE;
+
+#if defined(CONFIG_BFIN_MAC_RMII)
+       opmode |= RMII; /* For Now only 100MBit are supported */
+#ifdef CONFIG_BF_REV_0_2
+       opmode |= TE;
+#endif
+#endif
+       /* Turn on the EMAC rx */
+       bfin_write_EMAC_OPMODE(opmode);
+
+       return 0;
+}
+
+/* Our watchdog timed out. Called by the networking layer */
+static void bf537mac_timeout(struct net_device *dev)
+{
+       pr_debug("%s: %s\n", dev->name, __FUNCTION__);
+
+       bf537mac_reset();
+
+       /* reset tx queue */
+       tx_list_tail = tx_list_head->next;
+
+       bf537mac_enable(dev);
+
+       /* We can accept TX packets again */
+       dev->trans_start = jiffies;
+       netif_wake_queue(dev);
+}
+
+/*
+ * Get the current statistics.
+ * This may be called with the card open or closed.
+ */
+static struct net_device_stats *bf537mac_query_statistics(struct net_device
+                                                         *dev)
+{
+       struct bf537mac_local *lp = netdev_priv(dev);
+
+       pr_debug("%s: %s\n", dev->name, __FUNCTION__);
+
+       return &lp->stats;
+}
+
+/*
+ * This routine will, depending on the values passed to it,
+ * either make it accept multicast packets, go into
+ * promiscuous mode (for TCPDUMP and cousins) or accept
+ * a select set of multicast packets
+ */
+static void bf537mac_set_multicast_list(struct net_device *dev)
+{
+       u32 sysctl;
+
+       if (dev->flags & IFF_PROMISC) {
+               printk(KERN_INFO "%s: set to promisc mode\n", dev->name);
+               sysctl = bfin_read_EMAC_OPMODE();
+               sysctl |= RAF;
+               bfin_write_EMAC_OPMODE(sysctl);
+       } else if (dev->flags & IFF_ALLMULTI || dev->mc_count) {
+               /* accept all multicast */
+               sysctl = bfin_read_EMAC_OPMODE();
+               sysctl |= PAM;
+               bfin_write_EMAC_OPMODE(sysctl);
+       } else {
+               /* clear promisc or multicast mode */
+               sysctl = bfin_read_EMAC_OPMODE();
+               sysctl &= ~(RAF | PAM);
+               bfin_write_EMAC_OPMODE(sysctl);
+       }
+}
+
+/*
+ * this puts the device in an inactive state
+ */
+static void bf537mac_shutdown(struct net_device *dev)
+{
+       /* Turn off the EMAC */
+       bfin_write_EMAC_OPMODE(0x00000000);
+       /* Turn off the EMAC RX DMA */
+       bfin_write_DMA1_CONFIG(0x0000);
+       bfin_write_DMA2_CONFIG(0x0000);
+}
+
+/*
+ * Open and Initialize the interface
+ *
+ * Set up everything, reset the card, etc..
+ */
+static int bf537mac_open(struct net_device *dev)
+{
+       pr_debug("%s: %s\n", dev->name, __FUNCTION__);
+
+       /*
+        * Check that the address is valid.  If its not, refuse
+        * to bring the device up.  The user must specify an
+        * address using ifconfig eth0 hw ether xx:xx:xx:xx:xx:xx
+        */
+       if (!is_valid_ether_addr(dev->dev_addr)) {
+               printk(KERN_WARNING DRV_NAME ": no valid ethernet hw addr\n");
+               return -EINVAL;
+       }
+
+       /* initial rx and tx list */
+       desc_list_init();
+
+       bf537mac_setphy(dev);
+       setup_system_regs(dev);
+       bf537mac_reset();
+       bf537mac_enable(dev);
+
+       pr_debug("hardware init finished\n");
+       netif_start_queue(dev);
+       netif_carrier_on(dev);
+
+       return 0;
+}
+
+/*
+ *
+ * this makes the board clean up everything that it can
+ * and not talk to the outside world.   Caused by
+ * an 'ifconfig ethX down'
+ */
+static int bf537mac_close(struct net_device *dev)
+{
+       pr_debug("%s: %s\n", dev->name, __FUNCTION__);
+
+       netif_stop_queue(dev);
+       netif_carrier_off(dev);
+
+       /* clear everything */
+       bf537mac_shutdown(dev);
+
+       /* free the rx/tx buffers */
+       desc_list_free();
+
+       return 0;
+}
+
+static int __init bf537mac_probe(struct net_device *dev)
+{
+       struct bf537mac_local *lp = netdev_priv(dev);
+       int retval;
+
+       /* Grab the MAC address in the MAC */
+       *(__le32 *) (&(dev->dev_addr[0])) = cpu_to_le32(bfin_read_EMAC_ADDRLO());
+       *(__le16 *) (&(dev->dev_addr[4])) = cpu_to_le16((u16) bfin_read_EMAC_ADDRHI());
+
+       /* probe mac */
+       /*todo: how to proble? which is revision_register */
+       bfin_write_EMAC_ADDRLO(0x12345678);
+       if (bfin_read_EMAC_ADDRLO() != 0x12345678) {
+               pr_debug("can't detect bf537 mac!\n");
+               retval = -ENODEV;
+               goto err_out;
+       }
+
+       /* set the GPIO pins to Ethernet mode */
+       retval = setup_pin_mux(1);
+
+       if (retval)
+               return retval;
+
+       /*Is it valid? (Did bootloader initialize it?) */
+       if (!is_valid_ether_addr(dev->dev_addr)) {
+               /* Grab the MAC from the board somehow - this is done in the
+                  arch/blackfin/mach-bf537/boards/eth_mac.c */
+               get_bf537_ether_addr(dev->dev_addr);
+       }
+
+       /* If still not valid, get a random one */
+       if (!is_valid_ether_addr(dev->dev_addr)) {
+               random_ether_addr(dev->dev_addr);
+       }
+
+       setup_mac_addr(dev->dev_addr);
+
+       /* Fill in the fields of the device structure with ethernet values. */
+       ether_setup(dev);
+
+       dev->open = bf537mac_open;
+       dev->stop = bf537mac_close;
+       dev->hard_start_xmit = bf537mac_hard_start_xmit;
+       dev->tx_timeout = bf537mac_timeout;
+       dev->get_stats = bf537mac_query_statistics;
+       dev->set_multicast_list = bf537mac_set_multicast_list;
+#ifdef CONFIG_NET_POLL_CONTROLLER
+       dev->poll_controller = bf537mac_poll;
+#endif
+
+       /* fill in some of the fields */
+       lp->version = 1;
+       lp->PhyAddr = 0x01;
+       lp->CLKIN = 25;
+       lp->FullDuplex = 0;
+       lp->Negotiate = 1;
+       lp->FlowControl = 0;
+       spin_lock_init(&lp->lock);
+
+       /* now, enable interrupts */
+       /* register irq handler */
+       if (request_irq
+           (IRQ_MAC_RX, bf537mac_interrupt, IRQF_DISABLED | IRQF_SHARED,
+            "BFIN537_MAC_RX", dev)) {
+               printk(KERN_WARNING DRV_NAME
+                      ": Unable to attach BlackFin MAC RX interrupt\n");
+               return -EBUSY;
+       }
+
+       /* Enable PHY output early */
+       if (!(bfin_read_VR_CTL() & PHYCLKOE))
+               bfin_write_VR_CTL(bfin_read_VR_CTL() | PHYCLKOE);
+
+       retval = register_netdev(dev);
+       if (retval == 0) {
+               /* now, print out the card info, in a short format.. */
+               printk(KERN_INFO "%s: Version %s, %s\n",
+                        DRV_NAME, DRV_VERSION, DRV_DESC);
+       }
+
+err_out:
+       return retval;
+}
+
+static int bfin_mac_probe(struct platform_device *pdev)
+{
+       struct net_device *ndev;
+
+       ndev = alloc_etherdev(sizeof(struct bf537mac_local));
+       if (!ndev) {
+               printk(KERN_WARNING DRV_NAME ": could not allocate device\n");
+               return -ENOMEM;
+       }
+
+       SET_MODULE_OWNER(ndev);
+       SET_NETDEV_DEV(ndev, &pdev->dev);
+
+       platform_set_drvdata(pdev, ndev);
+
+       if (bf537mac_probe(ndev) != 0) {
+               platform_set_drvdata(pdev, NULL);
+               free_netdev(ndev);
+               printk(KERN_WARNING DRV_NAME ": not found\n");
+               return -ENODEV;
+       }
+
+       return 0;
+}
+
+static int bfin_mac_remove(struct platform_device *pdev)
+{
+       struct net_device *ndev = platform_get_drvdata(pdev);
+
+       platform_set_drvdata(pdev, NULL);
+
+       unregister_netdev(ndev);
+
+       free_irq(IRQ_MAC_RX, ndev);
+
+       free_netdev(ndev);
+
+       setup_pin_mux(0);
+
+       return 0;
+}
+
+static int bfin_mac_suspend(struct platform_device *pdev, pm_message_t state)
+{
+       return 0;
+}
+
+static int bfin_mac_resume(struct platform_device *pdev)
+{
+       return 0;
+}
+
+static struct platform_driver bfin_mac_driver = {
+       .probe = bfin_mac_probe,
+       .remove = bfin_mac_remove,
+       .resume = bfin_mac_resume,
+       .suspend = bfin_mac_suspend,
+       .driver = {
+                  .name = DRV_NAME,
+                  },
+};
+
+static int __init bfin_mac_init(void)
+{
+       return platform_driver_register(&bfin_mac_driver);
+}
+
+module_init(bfin_mac_init);
+
+static void __exit bfin_mac_cleanup(void)
+{
+       platform_driver_unregister(&bfin_mac_driver);
+}
+
+module_exit(bfin_mac_cleanup);
diff --git a/drivers/net/bfin_mac.h b/drivers/net/bfin_mac.h
new file mode 100644 (file)
index 0000000..af87189
--- /dev/null
@@ -0,0 +1,132 @@
+/*
+ * File:       drivers/net/bfin_mac.c
+ * Based on:
+ * Maintainer:
+ *             Bryan Wu <bryan.wu@analog.com>
+ *
+ * Original author:
+ *             Luke Yang <luke.yang@analog.com>
+ *
+ * Created:
+ * Description:
+ *
+ * Modified:
+ *             Copyright 2004-2006 Analog Devices Inc.
+ *
+ * Bugs:       Enter bugs at http://blackfin.uclinux.org/
+ *
+ * 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, 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 ;  see the file COPYING.
+ * If not, write to the Free Software Foundation,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ */
+
+/*
+ * PHY REGISTER NAMES
+ */
+#define PHYREG_MODECTL         0x0000
+#define PHYREG_MODESTAT                0x0001
+#define PHYREG_PHYID1          0x0002
+#define PHYREG_PHYID2          0x0003
+#define PHYREG_ANAR            0x0004
+#define PHYREG_ANLPAR          0x0005
+#define PHYREG_ANER            0x0006
+#define PHYREG_NSR             0x0010
+#define PHYREG_LBREMR          0x0011
+#define PHYREG_REC             0x0012
+#define PHYREG_10CFG           0x0013
+#define PHYREG_PHY1_1          0x0014
+#define PHYREG_PHY1_2          0x0015
+#define PHYREG_PHY2            0x0016
+#define PHYREG_TW_1            0x0017
+#define PHYREG_TW_2            0x0018
+#define PHYREG_TEST            0x0019
+
+#define PHY_RESET              0x8000
+#define PHY_ANEG_EN            0x1000
+#define PHY_DUPLEX             0x0100
+#define PHY_SPD_SET            0x2000
+
+#define BFIN_MAC_CSUM_OFFLOAD
+
+struct dma_descriptor {
+       struct dma_descriptor *next_dma_desc;
+       unsigned long start_addr;
+       unsigned short config;
+       unsigned short x_count;
+};
+
+struct status_area_rx {
+#if defined(BFIN_MAC_CSUM_OFFLOAD)
+       unsigned short ip_hdr_csum;     /* ip header checksum */
+       /* ip payload(udp or tcp or others) checksum */
+       unsigned short ip_payload_csum;
+#endif
+       unsigned long status_word;      /* the frame status word */
+};
+
+struct status_area_tx {
+       unsigned long status_word;      /* the frame status word */
+};
+
+/* use two descriptors for a packet */
+struct net_dma_desc_rx {
+       struct net_dma_desc_rx *next;
+       struct sk_buff *skb;
+       struct dma_descriptor desc_a;
+       struct dma_descriptor desc_b;
+       struct status_area_rx status;
+};
+
+/* use two descriptors for a packet */
+struct net_dma_desc_tx {
+       struct net_dma_desc_tx *next;
+       struct sk_buff *skb;
+       struct dma_descriptor desc_a;
+       struct dma_descriptor desc_b;
+       unsigned char packet[1560];
+       struct status_area_tx status;
+};
+
+struct bf537mac_local {
+       /*
+        * these are things that the kernel wants me to keep, so users
+        * can find out semi-useless statistics of how well the card is
+        * performing
+        */
+       struct net_device_stats stats;
+
+       int version;
+
+       int FlowEnabled;        /* record if data flow is active */
+       int EtherIntIVG;        /* IVG for the ethernet interrupt */
+       int RXIVG;              /* IVG for the RX completion */
+       int TXIVG;              /* IVG for the TX completion */
+       int PhyAddr;            /* PHY address */
+       int OpMode;             /* set these bits n the OPMODE regs */
+       int Port10;             /* set port speed to 10 Mbit/s */
+       int GenChksums;         /* IP checksums to be calculated */
+       int NoRcveLnth;         /* dont insert recv length at start of buffer */
+       int StripPads;          /* remove trailing pad bytes */
+       int FullDuplex;         /* set full duplex mode */
+       int Negotiate;          /* enable auto negotiation */
+       int Loopback;           /* loopback at the PHY */
+       int Cache;              /* Buffers may be cached */
+       int FlowControl;        /* flow control active */
+       int CLKIN;              /* clock in value in MHZ */
+       unsigned short IntMask; /* interrupt mask */
+       unsigned char Mac[6];   /* MAC address of the board */
+       spinlock_t lock;
+};
+
+extern void get_bf537_ether_addr(char *addr);