keymile add the common parts of the HDLC driver
PATCH 1/1 keymile add the common parts of the HDLC driver
This implements the ICN protocol used across the backplane and is needed by all the keymile boards.
Signed-off-by Gary Jennejohn <[email protected]>
---
checkpatch.pl reports an ERROR which is false.
board/keymile/common/keymile_hdlc_enet.c | 620 ++++++++++++++++++++++++++++++ board/keymile/common/keymile_hdlc_enet.h | 129 ++++++ 2 files changed, 749 insertions(+), 0 deletions(-) create mode 100644 board/keymile/common/keymile_hdlc_enet.c create mode 100644 board/keymile/common/keymile_hdlc_enet.h
board/keymile/common/keymile_hdlc_enet.c new file +620 -0
@@ -0,0 +1,620 @@ +/* +* (C) Copyright 2008 +* Gary Jennejohn, DENX Software Engineering GmbH, garyj at 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 */
board/keymile/common/keymile_hdlc_enet.h new file +129 -0
129 lines not on this page · raw message
Signature · 10 lines
1.5.4.3 --- Gary Jennejohn ********************************************************************* DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-0 Fax: +49-8142-66989-80 Email: office at denx.de *********************************************************************
Wolfgang Denk
AppliedPATCH 1/1 keymile add the common parts of the HDLC driver
Dear Gary Jennejohn,
Quoted 6 lines · In message <[email protected]> you wrote:
This implements the ICN protocol used across the backplane and is needed by all the keymile boards.
Signed-off-by Gary Jennejohn <[email protected]>---
Quoted 7 lines
checkpatch.pl reports an ERROR which is false.
board/keymile/common/keymile_hdlc_enet.c | 620 ++++++++++++++++++++++++++++++ board/keymile/common/keymile_hdlc_enet.h | 129 ++++++ 2 files changed, 749 insertions(+), 0 deletions(-) create mode 100644 board/keymile/common/keymile_hdlc_enet.c create mode 100644 board/keymile/common/keymile_hdlc_enet.h
Applied, thanks.
Best regards,
Wolfgang Denk
Signature · 5 lines
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
Never underestimate the bandwidth of a station wagon full of tapes.
-- Dr. Warren Jackson, Director, UTCS