/* * Copyright (c) 1982, 1986 Regents of the University of California. * All rights reserved. The Berkeley software License Agreement * specifies the terms and conditions for redistribution. * * @(#)if_ec.c 7.2 (Berkeley) 10/23/87 */ #include "ec.h" #if NEC > 0 /* * 3Com Ethernet Controller interface */ #include "../machine/pte.h" #include "param.h" #include "systm.h" #include "mbuf.h" #include "buf.h" #include "protosw.h" #include "socket.h" #include "syslog.h" #include "vmmac.h" #include "ioctl.h" #include "errno.h" #include "../net/if.h" #include "../net/netisr.h" #include "../net/route.h" #ifdef INET #include "../netinet/in.h" #include "../netinet/in_systm.h" #include "../netinet/in_var.h" #include "../netinet/ip.h" #include "../netinet/if_ether.h" #endif #ifdef NS #include "../netns/ns.h" #include "../netns/ns_if.h" #endif #include "../vax/cpu.h" #include "../vax/mtpr.h" #include "if_ecreg.h" #include "if_uba.h" #include "../vaxuba/ubareg.h" #include "../vaxuba/ubavar.h" #if CLSIZE == 2 #define ECBUFSIZE 32 /* on-board memory, clusters */ #endif int ecubamem(), ecprobe(), ecattach(), ecrint(), ecxint(), eccollide(); struct uba_device *ecinfo[NEC]; u_short ecstd[] = { 0 }; struct uba_driver ecdriver = { ecprobe, 0, ecattach, 0, ecstd, "ec", ecinfo, 0, 0, 0, 0, ecubamem }; int ecinit(),ecioctl(),ecoutput(),ecreset(); struct mbuf *ecget(); extern struct ifnet loif; /* * Ethernet software status per interface. * * Each interface is referenced by a network interface structure, * es_if, which the routing code uses to locate the interface. * This structure contains the output queue for the interface, its address, ... * We also have, for each interface, a UBA interface structure, which * contains information about the UNIBUS resources held by the interface: * map registers, buffered data paths, etc. Information is cached in this * structure for use by the if_uba.c routines in running the interface * efficiently. */ struct ec_softc { struct arpcom es_ac; /* common Ethernet structures */ #define es_if es_ac.ac_if /* network-visible interface */ #define es_addr es_ac.ac_enaddr /* hardware Ethernet address */ struct ifuba es_ifuba; /* UNIBUS resources */ short es_mask; /* mask for current output delay */ short es_oactive; /* is output active? */ u_char *es_buf[16]; /* virtual addresses of buffers */ } ec_softc[NEC]; /* * Configure on-board memory for an interface. * Called from autoconfig and after a uba reset. * The address of the memory on the uba is supplied in the device flags. */ ecubamem(ui, uban) register struct uba_device *ui; { register caddr_t ecbuf = (caddr_t) &umem[uban][ui->ui_flags]; register struct ecdevice *addr = (struct ecdevice *)ui->ui_addr; /* * Make sure csr is there (we run before ecprobe). */ if (badaddr((caddr_t)addr, 2)) return (-1); #if VAX780 if (cpu == VAX_780 && uba_hd[uban].uh_uba->uba_sr) { uba_hd[uban].uh_uba->uba_sr = uba_hd[uban].uh_uba->uba_sr; return (-1); } #endif /* * Make sure memory is turned on */ addr->ec_rcr = EC_AROM; /* * Tell the system that the board has memory here, so it won't * attempt to allocate the addresses later. */ if (ubamem(uban, ui->ui_flags, ECBUFSIZE*CLSIZE, 1) == 0) { printf("ec%d: cannot reserve uba addresses\n", ui->ui_unit); addr->ec_rcr = EC_MDISAB; /* disable memory */ return (-1); } /* * Check for existence of buffers on Unibus. */ if (badaddr((caddr_t)ecbuf, 2)) { bad: printf("ec%d: buffer mem not found\n", ui->ui_unit); (void) ubamem(uban, ui->ui_flags, ECBUFSIZE*2, 0); addr->ec_rcr = EC_MDISAB; /* disable memory */ return (-1); } #if VAX780 if (cpu == VAX_780 && uba_hd[uban].uh_uba->uba_sr) { uba_hd[uban].uh_uba->uba_sr = uba_hd[uban].uh_uba->uba_sr; goto bad; } #endif if (ui->ui_alive == 0) /* Only printf from autoconfig */ printf("ec%d: mem %x-%x\n", ui->ui_unit, ui->ui_flags, ui->ui_flags + ECBUFSIZE*CLBYTES - 1); ui->ui_type = 1; /* Memory on, allocated */ return (0); } /* * Do output DMA to determine interface presence and * interrupt vector. DMA is too short to disturb other hosts. */ ecprobe(reg, ui) caddr_t reg; struct uba_device *ui; { register int br, cvec; /* r11, r10 value-result */ register struct ecdevice *addr = (struct ecdevice *)reg; register caddr_t ecbuf = (caddr_t) &umem[ui->ui_ubanum][ui->ui_flags]; #ifdef lint br = 0; cvec = br; br = cvec; ecrint(0); ecxint(0); eccollide(0); #endif /* * Check that buffer memory was found and enabled. */ if (ui->ui_type == 0) return(0); /* * Make a one byte packet in what should be buffer #0. * Submit it for sending. This should cause an xmit interrupt. * The xmit interrupt vector is 8 bytes after the receive vector, * so adjust for this before returning. */ *(u_short *)ecbuf = (u_short) 03777; ecbuf[03777] = '\0'; addr->ec_xcr = EC_XINTEN|EC_XWBN; DELAY(100000); addr->ec_xcr = EC_XCLR; if (cvec > 0 && cvec != 0x200) { if (cvec & 04) { /* collision interrupt */ cvec -= 04; br += 1; /* rcv is collision + 1 */ } else { /* xmit interrupt */ cvec -= 010; br += 2; /* rcv is xmit + 2 */ } } return (1); } /* * Interface exists: make available by filling in network interface * record. System will initialize the interface when it is ready * to accept packets. */ ecattach(ui) struct uba_device *ui; { struct ec_softc *es = &ec_softc[ui->ui_unit]; register struct ifnet *ifp = &es->es_if; register struct ecdevice *addr = (struct ecdevice *)ui->ui_addr; int i, j; u_char *cp; ifp->if_unit = ui->ui_unit; ifp->if_name = "ec"; ifp->if_mtu = ETHERMTU; /* * Read the ethernet address off the board, one nibble at a time. */ addr->ec_xcr = EC_UECLR; /* zero address pointer */ addr->ec_rcr = EC_AROM; cp = es->es_addr; #define NEXTBIT addr->ec_rcr = EC_AROM|EC_ASTEP; addr->ec_rcr = EC_AROM for (i=0; i < sizeof (es->es_addr); i++) { *cp = 0; for (j=0; j<=4; j+=4) { *cp |= ((addr->ec_rcr >> 8) & 0xf) << j; NEXTBIT; NEXTBIT; NEXTBIT; NEXTBIT; } cp++; } printf("ec%d: hardware address %s\n", ui->ui_unit, ether_sprintf(es->es_addr)); ifp->if_init = ecinit; ifp->if_ioctl = ecioctl; ifp->if_output = ecoutput; ifp->if_reset = ecreset; ifp->if_flags = IFF_BROADCAST; for (i=0; i<16; i++) es->es_buf[i] = (u_char *)&umem[ui->ui_ubanum][ui->ui_flags + 2048*i]; if_attach(ifp); } /* * Reset of interface after UNIBUS reset. * If interface is on specified uba, reset its state. */ ecreset(unit, uban) int unit, uban; { register struct uba_device *ui; if (unit >= NEC || (ui = ecinfo[unit]) == 0 || ui->ui_alive == 0 || ui->ui_ubanum != uban) return; printf(" ec%d", unit); ec_softc[unit].es_if.if_flags &= ~IFF_RUNNING; ecinit(unit); } /* * Initialization of interface; clear recorded pending * operations, and reinitialize UNIBUS usage. */ ecinit(unit) int unit; { struct ec_softc *es = &ec_softc[unit]; struct ecdevice *addr; register struct ifnet *ifp = &es->es_if; int i, s; /* not yet, if address still unknown */ if (ifp->if_addrlist == (struct ifaddr *)0) return; /* * Hang receive buffers and start any pending writes. * Writing into the rcr also makes sure the memory * is turned on. */ if ((ifp->if_flags & IFF_RUNNING) == 0) { addr = (struct ecdevice *)ecinfo[unit]->ui_addr; s = splimp(); /* * write our ethernet address into the address recognition ROM * so we can always use the same EC_READ bits (referencing ROM), * in case we change the address sometime. * Note that this is safe here as the receiver is NOT armed. */ ec_setaddr(es->es_addr, unit); /* * Arm the receiver */ for (i = ECRHBF; i >= ECRLBF; i--) addr->ec_rcr = EC_READ | i; es->es_oactive = 0; es->es_mask = ~0; es->es_if.if_flags |= IFF_RUNNING; if (es->es_if.if_snd.ifq_head) ecstart(unit); splx(s); } } /* * Start output on interface. Get another datagram to send * off of the interface queue, and copy it to the interface * before starting the output. */ ecstart(unit) { register struct ec_softc *es = &ec_softc[unit]; struct ecdevice *addr; struct mbuf *m; if ((es->es_if.if_flags & IFF_RUNNING) == 0) return; IF_DEQUEUE(&es->es_if.if_snd, m); if (m == 0) return; ecput(es->es_buf[ECTBF], m); addr = (struct ecdevice *)ecinfo[unit]->ui_addr; addr->ec_xcr = EC_WRITE|ECTBF; es->es_oactive = 1; } /* * Ethernet interface transmitter interrupt. * Start another output if more data to send. */ ecxint(unit) int unit; { register struct ec_softc *es = &ec_softc[unit]; register struct ecdevice *addr = (struct ecdevice *)ecinfo[unit]->ui_addr; if (es->es_oactive == 0) return; if ((addr->ec_xcr&EC_XDONE) == 0 || (addr->ec_xcr&EC_XBN) != ECTBF) { printf("ec%d: stray xmit interrupt, xcr=%b\n", unit, addr->ec_xcr, EC_XBITS); es->es_oactive = 0; addr->ec_xcr = EC_XCLR; return; } es->es_if.if_opackets++; es->es_oactive = 0; es->es_mask = ~0; addr->ec_xcr = EC_XCLR; if (es->es_if.if_snd.ifq_head) ecstart(unit); } /* * Collision on ethernet interface. Do exponential * backoff, and retransmit. If have backed off all * the way print warning diagnostic, and drop packet. */ eccollide(unit) int unit; { register struct ec_softc *es = &ec_softc[unit]; register struct ecdevice *addr = (struct ecdevice *)ecinfo[unit]->ui_addr; register i; int delay; es->es_if.if_collisions++; if (es->es_oactive == 0) return; /* * Es_mask is a 16 bit number with n low zero bits, with * n the number of backoffs. When es_mask is 0 we have * backed off 16 times, and give up. */ if (es->es_mask == 0) { es->es_if.if_oerrors++; log(LOG_ERR, "ec%d: send error\n", unit); /* * Reset interface, then requeue rcv buffers. * Some incoming packets may be lost, but that * can't be helped. */ addr->ec_xcr = EC_UECLR; for (i=ECRHBF; i>=ECRLBF; i--) addr->ec_rcr = EC_READ|i; /* * Reset and transmit next packet (if any). */ es->es_oactive = 0; es->es_mask = ~0; if (es->es_if.if_snd.ifq_head) ecstart(unit); return; } /* * Do exponential backoff. Compute delay based on low bits * of the interval timer (1 bit for each transmission attempt, * but at most 5 bits). Then delay for that number of * slot times. A slot time is 51.2 microseconds (rounded to 51). * This does not take into account the time already used to * process the interrupt. */ es->es_mask <<= 1; delay = mfpr(ICR) & 0x1f &~ es->es_mask; DELAY(delay * 51); /* * Clear the controller's collision flag, thus enabling retransmit. */ addr->ec_xcr = EC_CLEAR; } /* * Ethernet interface receiver interrupt. * If input error just drop packet. * Otherwise examine * packet to determine type. If can't determine length * from type, then have to drop packet. Othewise decapsulate * packet based on type and pass to type specific higher-level * input routine. */ ecrint(unit) int unit; { struct ecdevice *addr = (struct ecdevice *)ecinfo[unit]->ui_addr; while (addr->ec_rcr & EC_RDONE) ecread(unit); } ecread(unit) int unit; { register struct ec_softc *es = &ec_softc[unit]; struct ecdevice *addr = (struct ecdevice *)ecinfo[unit]->ui_addr; register struct ether_header *ec; struct mbuf *m; int len, off, resid, ecoff, rbuf; register struct ifqueue *inq; u_char *ecbuf; es->es_if.if_ipackets++; rbuf = addr->ec_rcr & EC_RBN; if (rbuf < ECRLBF || rbuf > ECRHBF) panic("ecrint"); ecbuf = es->es_buf[rbuf]; ecoff = *(short *)ecbuf; if (ecoff <= ECRDOFF || ecoff > 2046) { es->es_if.if_ierrors++; #ifdef notdef if (es->es_if.if_ierrors % 100 == 0) printf("ec%d: += 100 input errors\n", unit); #endif goto setup; } /* * Get input data length. * Get pointer to ethernet header (in input buffer). * Deal with trailer protocol: if type is trailer type * get true type from first 16-bit word past data. * Remember that type was trailer by setting off. */ len = ecoff - ECRDOFF - sizeof (struct ether_header); ec = (struct ether_header *)(ecbuf + ECRDOFF); ec->ether_type = ntohs((u_short)ec->ether_type); #define ecdataaddr(ec, off, type) ((type)(((caddr_t)((ec)+1)+(off)))) if (ec->ether_type >= ETHERTYPE_TRAIL && ec->ether_type < ETHERTYPE_TRAIL+ETHERTYPE_NTRAILER) { off = (ec->ether_type - ETHERTYPE_TRAIL) * 512; if (off >= ETHERMTU) goto setup; /* sanity */ ec->ether_type = ntohs(*ecdataaddr(ec, off, u_short *)); resid = ntohs(*(ecdataaddr(ec, off+2, u_short *))); if (off + resid > len) goto setup; /* sanity */ len = off + resid; } else off = 0; if (len == 0) goto setup; /* * Pull packet off interface. Off is nonzero if packet * has trailing header; ecget will then force this header * information to be at the front, but we still have to drop * the type and length which are at the front of any trailer data. */ m = ecget(ecbuf, len, off, &es->es_if); if (m == 0) goto setup; if (off) { struct ifnet *ifp; ifp = *(mtod(m, struct ifnet **)); m->m_off += 2 * sizeof (u_short); m->m_len -= 2 * sizeof (u_short); *(mtod(m, struct ifnet **)) = ifp; } switch (ec->ether_type) { #ifdef INET case ETHERTYPE_IP: schednetisr(NETISR_IP); inq = &ipintrq; break; case ETHERTYPE_ARP: arpinput(&es->es_ac, m); goto setup; #endif #ifdef NS case ETHERTYPE_NS: schednetisr(NETISR_NS); inq = &nsintrq; break; #endif default: m_freem(m); goto setup; } if (IF_QFULL(inq)) { IF_DROP(inq); m_freem(m); goto setup; } IF_ENQUEUE(inq, m); setup: /* * Reset for next packet. */ addr->ec_rcr = EC_READ|EC_RCLR|rbuf; } /* * Ethernet output routine. * Encapsulate a packet of type family for the local net. * Use trailer local net encapsulation if enough data in first * packet leaves a multiple of 512 bytes of data in remainder. * If destination is this address or broadcast, send packet to * loop device to kludge around the fact that 3com interfaces can't * talk to themselves. */ ecoutput(ifp, m0, dst) struct ifnet *ifp; struct mbuf *m0; struct sockaddr *dst; { int type, s, error; u_char edst[6]; struct in_addr idst; register struct ec_softc *es = &ec_softc[ifp->if_unit]; register struct mbuf *m = m0; register struct ether_header *ec; register int off; struct mbuf *mcopy = (struct mbuf *)0; int usetrailers; if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING)) { error = ENETDOWN; goto bad; } switch (dst->sa_family) { #ifdef INET case AF_INET: idst = ((struct sockaddr_in *)dst)->sin_addr; if (!arpresolve(&es->es_ac, m, &idst, edst, &usetrailers)) return (0); /* if not yet resolved */ if (!bcmp((caddr_t)edst, (caddr_t)etherbroadcastaddr, sizeof(edst))) mcopy = m_copy(m, 0, (int)M_COPYALL); off = ntohs((u_short)mtod(m, struct ip *)->ip_len) - m->m_len; /* need per host negotiation */ if (usetrailers && off > 0 && (off & 0x1ff) == 0 && m->m_off >= MMINOFF + 2 * sizeof (u_short)) { type = ETHERTYPE_TRAIL + (off>>9); m->m_off -= 2 * sizeof (u_short); m->m_len += 2 * sizeof (u_short); *mtod(m, u_short *) = ntohs((u_short)ETHERTYPE_IP); *(mtod(m, u_short *) + 1) = ntohs((u_short)m->m_len); goto gottrailertype; } type = ETHERTYPE_IP; off = 0; goto gottype; #endif #ifdef NS case AF_NS: bcopy((caddr_t)&(((struct sockaddr_ns *)dst)->sns_addr.x_host), (caddr_t)edst, sizeof (edst)); if (!bcmp((caddr_t)edst, (caddr_t)&ns_broadhost, sizeof(edst))) { mcopy = m_copy(m, 0, (int)M_COPYALL); } else if (!bcmp((caddr_t)edst, (caddr_t)&ns_thishost, sizeof(edst))) { return(looutput(&loif, m, dst)); } type = ETHERTYPE_NS; off = 0; goto gottype; #endif case AF_UNSPEC: ec = (struct ether_header *)dst->sa_data; bcopy((caddr_t)ec->ether_dhost, (caddr_t)edst, sizeof (edst)); type = ec->ether_type; goto gottype; default: printf("ec%d: can't handle af%d\n", ifp->if_unit, dst->sa_family); error = EAFNOSUPPORT; goto bad; } gottrailertype: /* * Packet to be sent as trailer: move first packet * (control information) to end of chain. */ while (m->m_next) m = m->m_next; m->m_next = m0; m = m0->m_next; m0->m_next = 0; m0 = m; gottype: /* * Add local net header. If no space in first mbuf, * allocate another. */ if (m->m_off > MMAXOFF || MMINOFF + sizeof (struct ether_header) > m->m_off) { m = m_get(M_DONTWAIT, MT_HEADER); if (m == 0) { error = ENOBUFS; goto bad; } m->m_next = m0; m->m_off = MMINOFF; m->m_len = sizeof (struct ether_header); } else { m->m_off -= sizeof (struct ether_header); m->m_len += sizeof (struct ether_header); } ec = mtod(m, struct ether_header *); bcopy((caddr_t)edst, (caddr_t)ec->ether_dhost, sizeof (edst)); bcopy((caddr_t)es->es_addr, (caddr_t)ec->ether_shost, sizeof(ec->ether_shost)); ec->ether_type = htons((u_short)type); /* * Queue message on interface, and start output if interface * not yet active. */ s = splimp(); if (IF_QFULL(&ifp->if_snd)) { IF_DROP(&ifp->if_snd); error = ENOBUFS; goto qfull; } IF_ENQUEUE(&ifp->if_snd, m); if (es->es_oactive == 0) ecstart(ifp->if_unit); splx(s); return (mcopy ? looutput(&loif, mcopy, dst) : 0); qfull: m0 = m; splx(s); bad: m_freem(m0); if (mcopy) m_freem(mcopy); return (error); } /* * Routine to copy from mbuf chain to transmit * buffer in UNIBUS memory. * If packet size is less than the minimum legal size, * the buffer is expanded. We probably should zero out the extra * bytes for security, but that would slow things down. */ ecput(ecbuf, m) u_char *ecbuf; struct mbuf *m; { register struct mbuf *mp; register int off; u_char *bp; for (off = 2048, mp = m; mp; mp = mp->m_next) off -= mp->m_len; if (2048 - off < ETHERMIN + sizeof (struct ether_header)) off = 2048 - ETHERMIN - sizeof (struct ether_header); *(u_short *)ecbuf = off; bp = (u_char *)(ecbuf + off); for (mp = m; mp; mp = mp->m_next) { register unsigned len = mp->m_len; u_char *mcp; if (len == 0) continue; mcp = mtod(mp, u_char *); if ((unsigned)bp & 01) { *bp++ = *mcp++; len--; } if (off = (len >> 1)) { register u_short *to, *from; to = (u_short *)bp; from = (u_short *)mcp; do *to++ = *from++; while (--off > 0); bp = (u_char *)to, mcp = (u_char *)from; } if (len & 01) *bp++ = *mcp++; } m_freem(m); } /* * Routine to copy from UNIBUS memory into mbufs. * Similar in spirit to if_rubaget. * * Warning: This makes the fairly safe assumption that * mbufs have even lengths. */ struct mbuf * ecget(ecbuf, totlen, off0, ifp) u_char *ecbuf; int totlen, off0; struct ifnet *ifp; { register struct mbuf *m; struct mbuf *top = 0, **mp = ⊤ register int off = off0, len; u_char *cp; cp = ecbuf + ECRDOFF + sizeof (struct ether_header); while (totlen > 0) { register int words; u_char *mcp; MGET(m, M_DONTWAIT, MT_DATA); if (m == 0) goto bad; if (off) { len = totlen - off; cp = ecbuf + ECRDOFF + sizeof (struct ether_header) + off; } else len = totlen; if (ifp) len += sizeof(ifp); if (len >= NBPG) { MCLGET(m); if (m->m_len == CLBYTES) m->m_len = len = MIN(len, CLBYTES); else m->m_len = len = MIN(MLEN, len); } else { m->m_len = len = MIN(MLEN, len); m->m_off = MMINOFF; } mcp = mtod(m, u_char *); if (ifp) { /* * Prepend interface pointer to first mbuf. */ *(mtod(m, struct ifnet **)) = ifp; mcp += sizeof(ifp); len -= sizeof(ifp); ifp = (struct ifnet *)0; } if (words = (len >> 1)) { register u_short *to, *from; to = (u_short *)mcp; from = (u_short *)cp; do *to++ = *from++; while (--words > 0); mcp = (u_char *)to; cp = (u_char *)from; } if (len & 01) *mcp++ = *cp++; *mp = m; mp = &m->m_next; if (off == 0) { totlen -= len; continue; } off += len; if (off == totlen) { cp = ecbuf + ECRDOFF + sizeof (struct ether_header); off = 0; totlen = off0; } } return (top); bad: m_freem(top); return (0); } /* * Process an ioctl request. */ ecioctl(ifp, cmd, data) register struct ifnet *ifp; int cmd; caddr_t data; { register struct ifaddr *ifa = (struct ifaddr *)data; struct ec_softc *es = &ec_softc[ifp->if_unit]; struct ecdevice *addr; int s = splimp(), error = 0; addr = (struct ecdevice *)(ecinfo[ifp->if_unit]->ui_addr); switch (cmd) { case SIOCSIFADDR: ifp->if_flags |= IFF_UP; switch (ifa->ifa_addr.sa_family) { #ifdef INET case AF_INET: ecinit(ifp->if_unit); /* before arpwhohas */ ((struct arpcom *)ifp)->ac_ipaddr = IA_SIN(ifa)->sin_addr; arpwhohas((struct arpcom *)ifp, &IA_SIN(ifa)->sin_addr); break; #endif #ifdef NS case AF_NS: { register struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr); if (ns_nullhost(*ina)) ina->x_host = *(union ns_host *)(es->es_addr); else { /* * The manual says we can't change the address * while the receiver is armed, * so reset everything */ ifp->if_flags &= ~IFF_RUNNING; bcopy((caddr_t)ina->x_host.c_host, (caddr_t)es->es_addr, sizeof(es->es_addr)); } ecinit(ifp->if_unit); /* does ec_setaddr() */ break; } #endif default: ecinit(ifp->if_unit); break; } break; case SIOCSIFFLAGS: if ((ifp->if_flags & IFF_UP) == 0 && ifp->if_flags & IFF_RUNNING) { addr->ec_xcr = EC_UECLR; ifp->if_flags &= ~IFF_RUNNING; } else if (ifp->if_flags & IFF_UP && (ifp->if_flags & IFF_RUNNING) == 0) ecinit(ifp->if_unit); break; default: error = EINVAL; } splx(s); return (error); } ec_setaddr(physaddr,unit) u_char *physaddr; int unit; { struct ec_softc *es = &ec_softc[unit]; struct uba_device *ui = ecinfo[unit]; register struct ecdevice *addr = (struct ecdevice *)ui->ui_addr; register char nibble; register int i, j; /* * Use the ethernet address supplied * Note that we do a UECLR here, so the receive buffers * must be requeued. */ #ifdef DEBUG printf("ec_setaddr: setting address for unit %d = %s", unit, ether_sprintf(physaddr)); #endif addr->ec_xcr = EC_UECLR; addr->ec_rcr = 0; /* load requested address */ for (i = 0; i < 6; i++) { /* 6 bytes of address */ es->es_addr[i] = physaddr[i]; nibble = physaddr[i] & 0xf; /* lower nibble */ addr->ec_rcr = (nibble << 8); addr->ec_rcr = (nibble << 8) + EC_AWCLK; /* latch nibble */ addr->ec_rcr = (nibble << 8); for (j=0; j < 4; j++) { addr->ec_rcr = 0; addr->ec_rcr = EC_ASTEP; /* step counter */ addr->ec_rcr = 0; } nibble = (physaddr[i] >> 4) & 0xf; /* upper nibble */ addr->ec_rcr = (nibble << 8); addr->ec_rcr = (nibble << 8) + EC_AWCLK; /* latch nibble */ addr->ec_rcr = (nibble << 8); for (j=0; j < 4; j++) { addr->ec_rcr = 0; addr->ec_rcr = EC_ASTEP; /* step counter */ addr->ec_rcr = 0; } } #ifdef DEBUG /* * Read the ethernet address off the board, one nibble at a time. */ addr->ec_xcr = EC_UECLR; addr->ec_rcr = 0; /* read RAM */ cp = es->es_addr; #undef NEXTBIT #define NEXTBIT addr->ec_rcr = EC_ASTEP; addr->ec_rcr = 0 for (i=0; i < sizeof (es->es_addr); i++) { *cp = 0; for (j=0; j<=4; j+=4) { *cp |= ((addr->ec_rcr >> 8) & 0xf) << j; NEXTBIT; NEXTBIT; NEXTBIT; NEXTBIT; } cp++; } printf("ec_setaddr: RAM address for unit %d = %s", unit, ether_sprintf(physaddr)); #endif } #endif