]> git.dujemihanovic.xyz Git - u-boot.git/commitdiff
keymile add the common parts of the HDLC driver
authorGary Jennejohn <garyj@denx.de>
Sun, 9 Nov 2008 11:36:15 +0000 (12:36 +0100)
committerWolfgang Denk <wd@denx.de>
Sun, 7 Dec 2008 00:33:52 +0000 (01:33 +0100)
This implements the ICN protocol used across the backplane and is
needed by all the keymile boards.

Signed-off-by: Gary Jennejohn <garyj@denx.de>
board/keymile/common/keymile_hdlc_enet.c [new file with mode: 0644]
board/keymile/common/keymile_hdlc_enet.h [new file with mode: 0644]

diff --git a/board/keymile/common/keymile_hdlc_enet.c b/board/keymile/common/keymile_hdlc_enet.c
new file mode 100644 (file)
index 0000000..141371b
--- /dev/null
@@ -0,0 +1,620 @@
+/*
+ * (C) Copyright 2008
+ * Gary Jennejohn, DENX Software Engineering GmbH, garyj@denx.de.
+ *
+ * Based in part on cpu/mpc8260/ether_scc.c.
+ *
+ * See file CREDITS for list of people who contributed to this
+ * project.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
+ * MA 02111-1307 USA
+ */
+
+#include <common.h>
+#include <malloc.h>
+#include <net.h>
+
+#ifdef CONFIG_KEYMILE_HDLC_ENET
+#ifdef TEST_IT
+#include <command.h>
+#endif
+
+#include "keymile_hdlc_enet.h"
+
+extern char keymile_slot;      /* our slot number in the backplane */
+
+/* Allow up to about 50 ms for sending */
+#define TOUT_LOOP      50000
+
+/*
+ * Since, except during initialization, ethact is always HDLC ETHERNET
+ * while we're in the driver, just use serial_printf() everywhere for
+ * output.  This avoids possible conflicts when netconsole is being
+ * used.
+ */
+#define dprintf(fmt, args...)  serial_printf(fmt, ##args)
+
+/* Cannot use the storage from net.c because we allocate larger buffers */
+static volatile uchar MyPktBuf[HDLC_PKTBUFSRX * PKT_MAXBLR_SIZE + PKTALIGN];
+static volatile uchar *MyRxPackets[HDLC_PKTBUFSRX]; /* Receive packet */
+
+static unsigned int keymile_rxIdx;     /* index of the current RX buffer */
+
+static IPaddr_t cachedNumbers[CACHEDNUMBERS]; /* 4 bytes per entry */
+void initCachedNumbers(int);
+
+/*
+  * SCC Ethernet Tx and Rx buffer descriptors allocated at the
+  *  immr->udata_bd address on Dual-Port RAM
+  * Provide for Double Buffering
+  */
+typedef volatile struct CommonBufferDescriptor {
+    cbd_t txbd;                        /* Tx BD */
+    cbd_t rxbd[HDLC_PKTBUFSRX];        /* Rx BD */
+} RTXBD;
+
+/*
+ * This must be extern because it is allocated in DPRAM using CPM-sepcific
+ * code.
+ */
+static RTXBD *rtx;
+
+static int keymile_hdlc_enet_send(struct eth_device *, volatile void *, int);
+static int keymile_hdlc_enet_recv(struct eth_device *);
+void keymile_hdlc_enet_init_bds(RTXBD *);
+extern int keymile_hdlc_enet_init(struct eth_device *, bd_t *);
+extern void keymile_hdlc_enet_halt(struct eth_device *);
+
+/* flags in the buffer descriptor not defined anywhere else */
+#define BD_SC_CT       BD_SC_CD
+#define BD_SC_CR       0x04
+#define BD_SC_DE       0x80
+#ifndef BD_SC_TC
+#define BD_SC_TC       ((ushort)0x0400)        /* Transmit CRC */
+#endif
+#define BD_SC_FIRST    BD_SC_TC
+#define BD_SC_STATS (BD_SC_BR | BD_SC_FR | BD_SC_PR | BD_SC_CR | BD_SC_CD \
+               | BD_SC_OV | BD_SC_DE)
+
+#if defined(TEST_RX) || defined(TEST_TX) || defined(TEST_IT)
+static void hexdump(unsigned char *buf, int len)
+{
+       int i;
+       const int bytesPerLine = 32;
+
+       if (len > 4 * bytesPerLine)
+               len = 4 * bytesPerLine;
+       dprintf("\t address: %08x\n", (unsigned int)buf);
+       for (i = 0; i < len; i++) {
+               if (i % bytesPerLine == 0)
+                       dprintf("%04x: ", (unsigned short)i);
+               dprintf("%02x ", buf[i]);
+               if ((i + 1) % bytesPerLine == 0) {
+                       dprintf("\n");
+                       continue;
+               }
+               if ((i + 1) % 8 == 0)
+                       printf(" ");
+       }
+       if (len % bytesPerLine)
+               dprintf("\n");
+}
+#endif
+
+int keymile_hdlc_enet_initialize(bd_t *bis)
+{
+       struct eth_device *dev;
+
+       dev = (struct eth_device *) malloc(sizeof *dev);
+       memset(dev, 0, sizeof *dev);
+#ifdef TEST_IT
+       seth = dev;
+#endif
+
+       sprintf(dev->name, "HDLC ETHERNET");
+       dev->init   = keymile_hdlc_enet_init;
+       dev->halt   = keymile_hdlc_enet_halt;
+       dev->send   = keymile_hdlc_enet_send;
+       dev->recv   = keymile_hdlc_enet_recv;
+
+       eth_register(dev);
+
+       return 1;
+}
+
+/*
+ * This is called from the board-specific driver after rtx is allocated.
+ */
+void keymile_hdlc_enet_init_bds(RTXBD *board_rtx)
+{
+       volatile cbd_t *bdp;
+       int i;
+
+       rtx = board_rtx;
+       keymile_rxIdx = 0;
+       /*
+        * Initialize the buffer descriptors.
+        */
+       bdp = &rtx->txbd;
+       bdp->cbd_sc = 0;
+       bdp->cbd_bufaddr = 0;
+       bdp->cbd_sc = BD_SC_WRAP;
+
+       /*
+        *      Setup RX packet buffers, aligned correctly.
+        *      Borrowed from net/net.c.
+        */
+       MyRxPackets[0] = &MyPktBuf[0] + (PKTALIGN - 1);
+       MyRxPackets[0] -= (ulong)MyRxPackets[0] % PKTALIGN;
+       for (i = 1; i < HDLC_PKTBUFSRX; i++)
+               MyRxPackets[i] = MyRxPackets[0] + i * PKT_MAXBLR_SIZE;
+
+       bdp = &rtx->rxbd[0];
+       for (i = 0; i < HDLC_PKTBUFSRX; i++) {
+               bdp->cbd_sc = BD_SC_EMPTY;
+               /* Leave space at the start for INET header. */
+               bdp->cbd_bufaddr = (unsigned int)(MyRxPackets[i] +
+                       INET_HDR_ALIGN);
+               bdp++;
+       }
+       bdp--;
+       bdp->cbd_sc |= BD_SC_WRAP;
+}
+
+/*
+ * This returns the current port number for NETCONSOLE.  If nc_port
+ * in netconsole.c weren't declared static we wouldn't need this.
+ */
+static short get_netcons_port(void)
+{
+       char *p;
+       short nc_port;
+
+       nc_port = 6666; /* default */
+
+       p = getenv("ncip");
+       if (p != NULL) {
+               p = strchr(p, ':');
+               if (p != NULL)
+                       nc_port = simple_strtoul(p + 1, NULL, 10);
+       }
+
+       return htons(nc_port);
+}
+
+/*
+ * Read the port numbers from the variables
+ */
+void initCachedNumbers(int verbose)
+{
+       char *str;
+       ushort port;
+
+       /* already in network order */
+       cachedNumbers[IP_ADDR] = getenv_IPaddr("ipaddr");
+       /* already in network order */
+       cachedNumbers[IP_SERVER] = getenv_IPaddr("serverip");
+       str = getenv("tftpsrcp");
+       if (str != NULL) {
+               /* avoid doing htons() again and again */
+               port = htons((ushort)simple_strtol(str, NULL, 10));
+               cachedNumbers[TFTP_SRC_PORT] = port;
+       } else
+               /* this can never be a valid port number */
+               cachedNumbers[TFTP_SRC_PORT] = (ulong)-1;
+       str = getenv("tftpdstp");
+       if (str != NULL) {
+               /* avoid doing htons() again and again */
+               port = htons((ushort)simple_strtol(str, NULL, 10));
+               cachedNumbers[TFTP_DST_PORT] = port;
+       } else
+               /* this is the default value */
+               cachedNumbers[TFTP_DST_PORT] = htons(WELL_KNOWN_PORT);
+       /* already in network order */
+       cachedNumbers[NETCONS_PORT] = get_netcons_port();
+       if (verbose) {
+               dprintf("\nIP Number Initialization:\n");
+               dprintf(" ip address          %08lx\n", cachedNumbers[IP_ADDR]);
+               dprintf(" server ip address   %08lx\n",
+                       cachedNumbers[IP_SERVER]);
+               dprintf(" tftp client port    %ld\n",
+                       cachedNumbers[TFTP_SRC_PORT]);
+               dprintf(" tftp server port    %ld\n",
+                       cachedNumbers[TFTP_DST_PORT]);
+               dprintf(" netcons port        %ld\n",
+                       cachedNumbers[NETCONS_PORT]);
+               dprintf(" slot number (hex)   %02x\n", keymile_slot);
+       }
+}
+
+static void keymile_hdlc_enet_doarp(volatile void *packet, int len)
+{
+       ARP_t *arp;
+       IPaddr_t src_ip; /* U-Boot's IP */
+       IPaddr_t dest_ip; /* the mgcoge's IP */
+       unsigned char *packet_copy = malloc(len);
+
+       /*
+        * Handling an ARP request means that a new transfer has started.
+        * Update our cached parameters now.
+        */
+       initCachedNumbers(0); /* may reinit port numbers */
+
+       /* special handling required for ARP */
+       arp = (ARP_t *)(packet + ETHER_HDR_SIZE);
+       /*
+        *      XXXX
+        * This is pretty dirty!  NetReceive only uses
+        * a few fields when handling an ARP reply, so
+        * we only modify those here.  This could
+        * result in catastrophic failure at a later
+        * time if the handler is modified!
+        */
+       arp->ar_op = htons(ARPOP_REPLY);
+       /* save his/our IP */
+       src_ip = NetReadIP(&arp->ar_data[6]);
+       dest_ip = NetReadIP(&arp->ar_data[16]);
+       /* copy target IP to source IP */
+       NetCopyIP(&arp->ar_data[6], &dest_ip);
+       /* copy our IP to the right place */
+       NetCopyIP(&arp->ar_data[16], &src_ip);
+       /* always use 0x7f as the MAC for the coge */
+       arp->ar_data[0] = HDLC_UACUA;
+       /*
+        * copy the packet
+        * if NetReceive wants to write to stdout, it may overwrite packet
+        * especially if stdout is set to nc!
+        *
+        * However, if the malloc() above fails then we can still try the
+        * original packet, rather than causing the transfer to fail.
+        */
+       if (packet_copy != NULL) {
+               memcpy(packet_copy, (char *)packet, len);
+               NetReceive(packet_copy, len);
+               free(packet_copy);
+       } else
+               NetReceive(packet, len);
+}
+
+/*
+ * NOTE all callers ignore the returned value!
+ * At the moment this only handles ARP Requests, TFTP and NETCONSOLE.
+ */
+static int keymile_hdlc_enet_send(struct eth_device *dev, volatile void *packet,
+       int len)
+{
+       int j;
+       uint data_addr;
+       int data_len;
+       struct icn_hdr header;
+       struct icn_frame *frame;
+       Ethernet_t *et;
+       ARP_t *arp;
+       IP_t *ip;
+
+       if (len > (MAX_FRAME_LENGTH - sizeof(header)))
+               return -1;
+
+       frame = NULL;
+       et = NULL;
+       arp = NULL;
+       ip = NULL;
+
+       j = 0;
+       while ((rtx->txbd.cbd_sc & BD_SC_READY) && (j < TOUT_LOOP)) {
+               /* will also trigger Wd if needed, but maybe too often  */
+               udelay(1);
+               j++;
+       }
+       if (j >= TOUT_LOOP) {
+               dprintf("TX not ready sc %x\n", rtx->txbd.cbd_sc);
+               return -1;
+       }
+
+       /*
+        * First check for an ARP Request since this requires special handling.
+        */
+       if (len >= (ARP_HDR_SIZE + ETHER_HDR_SIZE)) {
+               et = (Ethernet_t *)packet;
+               arp = (ARP_t *)(((char *)et) + ETHER_HDR_SIZE);
+               /* ARP and REQUEST? */
+               if (et->et_protlen == PROT_ARP &&
+                       arp->ar_op == htons(ARPOP_REQUEST)) {
+                       /* just short-circuit the request on the U-Boot side */
+                       keymile_hdlc_enet_doarp(packet, len);
+                       return 0;
+               }
+       }
+
+       /*
+        * GJ - I suppose the assumption here that len will always be
+        * > INET_HDR_SIZE is alright as long as the network stack
+        * isn't changed.
+        * Do not send INET header.
+        */
+       data_len = len + sizeof(header) - INET_HDR_SIZE;
+       frame = (struct icn_frame *) (((char *)packet) + INET_HDR_SIZE -
+               sizeof(header));
+
+#ifdef TEST_TX
+       printf("frame: %08x, ", frame);
+       hexdump((unsigned char *)packet, data_len + INET_HDR_SIZE);
+#endif
+
+       data_addr = (uint)frame;
+       if (len >= (IP_HDR_SIZE + ETHER_HDR_SIZE))
+               ip = (IP_t *)(packet + ETHER_HDR_SIZE);
+       /* Is it TFTP? TFTP always uses UDP and the cached dport */
+       if (ip != NULL && ip->ip_p == IPPROTO_UDP && ip->udp_dst ==
+                       (ushort)cachedNumbers[TFTP_DST_PORT]) {
+               /* just in case the port wasn't set in the environment */
+               if (cachedNumbers[TFTP_SRC_PORT] == (ulong)-1)
+                       cachedNumbers[TFTP_SRC_PORT] = ip->udp_src;
+               frame->hdr.application = MGS_TFTP;
+       }
+       /*
+        * Is it NETCONSOLE?  NETCONSOLE always uses UDP.
+        */
+       else if (ip != NULL && ip->ip_p == IPPROTO_UDP
+               && ip->udp_dst == (ushort)cachedNumbers[NETCONS_PORT]) {
+                       frame->hdr.application = MGS_NETCONS;
+       } else {
+               /* reject unknown packets */
+               /* may do some check on frame->hdr.application */
+               dprintf("Unknown packet type in %s, rejected\n",
+                       __func__);
+               return -1;
+       }
+       /*
+        * Could extract the target's slot ID from its MAC here,
+        * but u-boot only wants to talk to the active server.
+        *
+        * avoid setting new source address when moving to another slot
+        */
+       frame->hdr.src_addr = keymile_slot;
+       frame->hdr.dest_addr = HDLC_UACUA;
+#ifdef TEST_TX
+       {
+               dprintf("TX: ");
+               hexdump((unsigned char *)data_addr, data_len);
+       }
+#endif
+
+       flush_cache(data_addr, data_len);
+       rtx->txbd.cbd_bufaddr = data_addr;
+       rtx->txbd.cbd_datlen = data_len;
+       rtx->txbd.cbd_sc |= (BD_SC_READY | BD_SC_TC | BD_SC_LAST | BD_SC_WRAP);
+
+       while ((rtx->txbd.cbd_sc & BD_SC_READY) && (j < TOUT_LOOP)) {
+               /* will also trigger Wd if needed, but maybe too often  */
+               udelay(1);
+               j++;
+       }
+       if (j >= TOUT_LOOP)
+               dprintf("TX timeout\n");
+#ifdef ET_DEBUG
+       dprintf("cycles: %d    status: %x\n", j, rtx->txbd.cbd_sc);
+#endif
+       j = (rtx->txbd.cbd_sc & BD_SC_STATS); /* return only status bits */
+       return j;
+}
+
+/*
+ * During a receive, the RxIdx points to the current incoming buffer.
+ * When we update through the ring, if the next incoming buffer has
+ * not been given to the system, we just set the empty indicator,
+ * effectively tossing the packet.
+ */
+static int keymile_hdlc_enet_recv(struct eth_device *dev)
+{
+       int length;
+       unsigned char app;
+       struct icn_frame *fp;
+       Ethernet_t *ep;
+       IP_t *ip;
+
+       for (;;) {
+               if (rtx->rxbd[keymile_rxIdx].cbd_sc & BD_SC_EMPTY) {
+                       length = -1;
+                       break;  /* nothing received - leave for() loop */
+               }
+
+               length = rtx->rxbd[keymile_rxIdx].cbd_datlen;
+#ifdef TEST_RX
+               dprintf("packet %d bytes long\n", length);
+#endif
+
+               /*
+                * BD_SC_BR -> LG bit
+                * BD_SC_FR -> NO bit
+                * BD_SC_PR -> AB bit
+                * BD_SC_NAK -> CR bit
+                * 0x80 -> DE bit
+                */
+               if (rtx->rxbd[keymile_rxIdx].cbd_sc & BD_SC_STATS) {
+#ifdef ET_DEBUG
+                       dprintf("err: %x\n", rtx->rxbd[keymile_rxIdx].cbd_sc);
+#endif
+               } else if (length > MAX_FRAME_LENGTH) { /* can't happen */
+#ifdef ET_DEBUG
+                       dprintf("err: packet too big\n");
+#endif
+               } else {
+                       fp = (struct icn_frame *)(MyRxPackets[keymile_rxIdx] +
+                               INET_HDR_ALIGN - INET_HDR_SIZE);
+#ifdef TEST_RX
+                       dprintf("RX %d: ", keymile_rxIdx);
+                       hexdump((unsigned char *)MyRxPackets[keymile_rxIdx],
+                               INET_HDR_ALIGN + INET_HDR_SIZE + 4);
+#endif
+                       /* copy icn header to the beginning */
+                       memcpy(fp, ((char *)fp + INET_HDR_SIZE),
+                               sizeof(struct icn_hdr));
+                       app = fp->hdr.application;
+                       if (app == MGS_NETCONS || app == MGS_TFTP) {
+                               struct icn_hdr *ih = &fp->hdr;
+                               unsigned char icn_src_addr = ih->src_addr;
+                               unsigned char icn_dest_addr = ih->dest_addr;
+
+                               /*
+                                * expand header by INET_HDR_SIZE
+                                */
+                               length += INET_HDR_SIZE;
+                               /* initalize header */
+                               memset((char *)fp->data, 0x00, INET_HDR_SIZE);
+                               ep = (Ethernet_t *)fp->data;
+                               /* set MACs */
+                               ep->et_dest[0] = icn_dest_addr;
+                               ep->et_src[0] = icn_src_addr;
+                               ep->et_protlen = htons(PROT_IP);
+                               /* set ip stuff */
+                               ip = (IP_t *)(fp->data + ETHER_HDR_SIZE);
+                               /* set ip addresses */
+                               ip->ip_src = cachedNumbers[IP_SERVER];
+                               ip->ip_dst = cachedNumbers[IP_ADDR];
+                               /* ip length */
+                               ip->ip_len = htons(length - ETHER_HDR_SIZE -
+                                       REMOVE);
+                               /* ip proto */
+                               ip->ip_p = IPPROTO_UDP;
+                               switch (app) {
+                               case MGS_TFTP:
+                                       /* swap src/dst port numbers */
+                                       ip->udp_src = (ushort)
+                                               cachedNumbers[TFTP_DST_PORT];
+                                       ip->udp_dst = (ushort)
+                                               cachedNumbers[TFTP_SRC_PORT];
+                                       ip->udp_len = ip->ip_len -
+                                               IP_HDR_SIZE_NO_UDP;
+                                       ip->udp_xsum = 0;
+                                       break;
+                               case MGS_NETCONS:
+                                       ip->udp_src = (ushort)
+                                               cachedNumbers[NETCONS_PORT];
+                                       /*
+                                        * in drivers/net/netconsole.c src port
+                                        * equals dest port
+                                        */
+                                       ip->udp_dst = ip->udp_src;
+                                       ip->udp_len = ip->ip_len -
+                                               IP_HDR_SIZE_NO_UDP;
+                                       ip->udp_xsum = 0;
+                                       break;
+                               }
+                               /* ip version */
+                               ip->ip_hl_v = (0x40) | (0x0f &
+                                       (IP_HDR_SIZE_NO_UDP / 4));
+                               ip->ip_tos = 0;
+                               ip->ip_id = 0;
+                               /* flags, fragment offset */
+                               ip->ip_off = htons(0x4000);
+                               ip->ip_ttl = 255; /* time to live */
+                               /* have to fixup the checksum */
+                               ip->ip_sum = ~NetCksum((uchar *)ip,
+                                       IP_HDR_SIZE_NO_UDP / 2);
+                               /*
+                                * Pass the packet up to the protocol layers
+                                * but remove dest_addr, src_addr, application
+                                * and the CRC.
+                                */
+#ifdef TEST_RX
+                               hexdump((unsigned char *)fp->data,
+                                       INET_HDR_SIZE + 4);
+#endif
+                               NetReceive(fp->data, length - REMOVE);
+                       } else {
+                               /*
+                                * the other application types are not yet
+                                * supported by u-boot.
+                                */
+                               /* normally drop it */
+#ifdef TEST_NO
+                               /* send it anyway */
+                               fp = (struct icn_frame *)
+                                       (MyRxPackets[keymile_rxIdx] +
+                                               INET_HDR_ALIGN);
+                               NetReceive(fp->data, length - REMOVE);
+#endif
+                       }
+               }
+
+               /* Give the buffer back to the SCC. */
+               rtx->rxbd[keymile_rxIdx].cbd_datlen = 0;
+
+               /* wrap around buffer index when necessary */
+               if ((keymile_rxIdx + 1) >= HDLC_PKTBUFSRX) {
+                       rtx->rxbd[HDLC_PKTBUFSRX - 1].cbd_sc =
+                               (BD_SC_WRAP | BD_SC_EMPTY);
+                       keymile_rxIdx = 0;
+               } else {
+                       rtx->rxbd[keymile_rxIdx].cbd_sc = BD_SC_EMPTY;
+                       keymile_rxIdx++;
+               }
+       }
+       return length;
+}
+
+#ifdef TEST_IT
+/* simple send test routine */
+int hdlc_enet_stest(struct cmd_tbl_s *a, int b, int c, char **d)
+{
+       unsigned char pkt[2];
+       int ret;
+
+       dprintf("enter stest\n");
+       /* may have to initialize things */
+       if (seth->state != ETH_STATE_ACTIVE) {
+               /* the bd_t* is not used */
+               if (seth->init(seth, NULL) >= 0)
+                       seth->state = ETH_STATE_ACTIVE;
+       }
+       pkt[0] = 0xea;
+       pkt[1] = 0xae;
+       ret = keymile_hdlc_enet_send(seth, pkt, 2);
+       dprintf("return from send %x\n", ret);
+       dprintf("exit stest\n");
+       return ret;
+}
+U_BOOT_CMD(
+       stest, 1, 1, hdlc_enet_stest,
+       "stest  - simple send test for hdlc_enet\n",
+       "no arguments\n"
+);
+/* simple receive test routine */
+int hdlc_enet_rtest(struct cmd_tbl_s *a, int b, int c, char **d)
+{
+       int ret;
+
+       dprintf("enter rtest\n");
+       /* may have to initialize things */
+       if (seth->state != ETH_STATE_ACTIVE) {
+               /* the bd_t* is not used */
+               if (seth->init(seth, NULL) >= 0)
+                       seth->state = ETH_STATE_ACTIVE;
+       }
+       ret = keymile_hdlc_enet_recv(seth);
+       dprintf("return from recv %x\n", ret);
+       dprintf("exit rtest\n");
+       return ret;
+}
+U_BOOT_CMD(
+       rtest, 1, 1, hdlc_enet_rtest,
+       "rtest  - simple receive test for hdlc_enet\n",
+       "no arguments\n"
+);
+#endif
+
+#endif /* CONFIG_KEYMILE_HDLC_ENET */
diff --git a/board/keymile/common/keymile_hdlc_enet.h b/board/keymile/common/keymile_hdlc_enet.h
new file mode 100644 (file)
index 0000000..965bd5a
--- /dev/null
@@ -0,0 +1,129 @@
+/*
+ * (C) Copyright 2008
+ * Gary Jennejohn, DENX Software Engineering GmbH, garyj@denx.de.
+ *
+ * See file CREDITS for list of people who contributed to this
+ * project.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
+ * MA 02111-1307 USA
+ */
+
+#ifndef _KEYMILE_HDLC_ENET_H_
+#define _KEYMILE_HDLC_ENET_H_
+
+/* Unfortuantely, we have do this to get the flag defines in the cbd_t */
+#ifdef CONFIG_MGSUVD
+#include <commproc.h>
+#endif
+#ifdef CONFIG_MGCOGE
+#include <mpc8260.h>
+#include <asm/cpm_8260.h>
+#endif
+
+/*
+ * Defines for the ICN protocol used for communication over HDLC
+ * on the backplane between MGSUVDs and MGCOGEs.
+ */
+
+/*
+ * MAC which is reserved for communication (0x00 - 0xff in the last byte,
+ * which is the slot number)
+ */
+
+/*
+ * A DLL frame looks like this:
+ *  8 bit  |  8 bit     |  8 bit |  8 bit     | n * 8 bit| 16 bit|  8 bit
+ *  opening| destination| source | application|  data    |  FCS  | closing
+ *   flag  |  address   | address|            |          |       |  flag
+ *  (HW)      (APP)       (APP)      (APP)       (APP)      (HW)   (HW)
+ */
+
+/*
+ * The opening flag, the FCS and the closing flag are set by the hardware so
+ * they are not reflected in this struct.
+ */
+struct icn_hdr {
+       unsigned char dest_addr;
+       unsigned char src_addr;
+       unsigned char application;
+} __attribute__((packed));
+
+#define ICNHDR_LEN     (sizeof(struct icn_hdr))
+#define CRC_LEN                (sizeof(short))
+/* bytes to remove from packet before sending it upstream */
+#define REMOVE         (ICNHDR_LEN + CRC_LEN)
+
+struct icn_frame {
+       struct icn_hdr hdr;
+       unsigned char data[0];  /* a place holder */
+} __attribute__((packed));
+
+/* Address field */
+#define HDLC_UUA       0x00 /* Unicast Unit Address */
+#define HDLC_UUA_MASK  0x3f /* the last 6 bits contain the slot number */
+#define SET_HDLC_UUA(x) ((HDLC_UUA | ((x) & HDLC_UUA_MASK)))
+#define HDLC_UACUA     0x7f /* Unicast Active Control Unit Address */
+#define HDLC_BCAST     0xff /* broadcast */
+
+/* Application field */
+#define MGS_UUSP       0x00
+#define MGS_UREP       0x01
+#define MGS_IUP                0x02
+#define MGS_UTA                0x03
+#define MGS_MDS                0x04
+#define MGS_ITIME      0x05
+/* added by DENX */
+#define MGS_NETCONS    0x06    /* netconsole */
+#define MGS_TFTP       0x07
+
+/* Useful defines for buffer sizes, etc. */
+#define HDLC_PKTBUFSRX         32
+#define MAX_FRAME_LENGTH       1500 /* ethernet frame size */
+                        /*   14        +    28     */
+#define INET_HDR_SIZE  (ETHER_HDR_SIZE + IP_HDR_SIZE)
+#define INET_HDR_ALIGN (((INET_HDR_SIZE + PKTALIGN - 1) / PKTALIGN) * PKTALIGN)
+/* INET_HDR_SIZE is stripped off */
+#define PKT_MAXBLR_SIZE        (MAX_FRAME_LENGTH + INET_HDR_ALIGN)
+
+/*
+ * It is too slow to read always the port numbers and IP addresses from the
+ * string variables.
+ * cachedNumbers is meant to cache it.
+ * THIS IS ONLY A SPEED IMPROVEMENT!
+ */
+enum {
+       IP_ADDR = 0,    /* getenv_IPaddr("serverip"); */
+       IP_SERVER,      /* getenv_IPaddr("ipaddr"); */
+       TFTP_SRC_PORT,  /* simple_strtol(getenv("tftpsrcp"), NULL, 10); */
+       TFTP_DST_PORT,  /* simple_strtol(getenv("tftpdstp"), NULL, 10); */
+       NETCONS_PORT,   /* simple_strtol(getenv("ncip"), NULL, 10); */
+       CACHEDNUMBERS
+};
+
+#define WELL_KNOWN_PORT 69     /* Well known TFTP port # */
+
+/* define this to create a test commend (htest) */
+#undef TEST_IT
+#ifdef TEST_IT
+/* have to save a copy of the eth_device for the test command's use */
+struct eth_device *seth;
+#endif
+/* define this for outputting of received packets */
+#undef TEST_RX
+/* define this for outputting of packets being sent */
+#undef TEST_TX
+
+#endif /* _KEYMILE_HDLC_ENET_H_ */