diff options
author | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 15:20:36 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@ppc970.osdl.org> | 2005-04-16 15:20:36 -0700 |
commit | 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch) | |
tree | 0bba044c4ce775e45a88a51686b5d9f90697ea9d /drivers/net/tokenring/ibmtr.c | |
download | blackbird-op-linux-1da177e4c3f41524e886b7f1b8a0c1fc7321cac2.tar.gz blackbird-op-linux-1da177e4c3f41524e886b7f1b8a0c1fc7321cac2.zip |
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Diffstat (limited to 'drivers/net/tokenring/ibmtr.c')
-rw-r--r-- | drivers/net/tokenring/ibmtr.c | 1987 |
1 files changed, 1987 insertions, 0 deletions
diff --git a/drivers/net/tokenring/ibmtr.c b/drivers/net/tokenring/ibmtr.c new file mode 100644 index 000000000000..c098863bdd9d --- /dev/null +++ b/drivers/net/tokenring/ibmtr.c @@ -0,0 +1,1987 @@ +/* ibmtr.c: A shared-memory IBM Token Ring 16/4 driver for linux + * + * Written 1993 by Mark Swanson and Peter De Schrijver. + * This software may be used and distributed according to the terms + * of the GNU General Public License, incorporated herein by reference. + * + * This device driver should work with Any IBM Token Ring Card that does + * not use DMA. + * + * I used Donald Becker's (becker@scyld.com) device driver work + * as a base for most of my initial work. + * + * Changes by Peter De Schrijver + * (Peter.Deschrijver@linux.cc.kuleuven.ac.be) : + * + * + changed name to ibmtr.c in anticipation of other tr boards. + * + changed reset code and adapter open code. + * + added SAP open code. + * + a first attempt to write interrupt, transmit and receive routines. + * + * Changes by David W. Morris (dwm@shell.portal.com) : + * 941003 dwm: - Restructure tok_probe for multiple adapters, devices. + * + Add comments, misc reorg for clarity. + * + Flatten interrupt handler levels. + * + * Changes by Farzad Farid (farzy@zen.via.ecp.fr) + * and Pascal Andre (andre@chimay.via.ecp.fr) (March 9 1995) : + * + multi ring support clean up. + * + RFC1042 compliance enhanced. + * + * Changes by Pascal Andre (andre@chimay.via.ecp.fr) (September 7 1995) : + * + bug correction in tr_tx + * + removed redundant information display + * + some code reworking + * + * Changes by Michel Lespinasse (walken@via.ecp.fr), + * Yann Doussot (doussot@via.ecp.fr) and Pascal Andre (andre@via.ecp.fr) + * (February 18, 1996) : + * + modified shared memory and mmio access port the driver to + * alpha platform (structure access -> readb/writeb) + * + * Changes by Steve Kipisz (bungy@ibm.net or kipisz@vnet.ibm.com) + * (January 18 1996): + * + swapped WWOR and WWCR in ibmtr.h + * + moved some init code from tok_probe into trdev_init. The + * PCMCIA code can call trdev_init to complete initializing + * the driver. + * + added -DPCMCIA to support PCMCIA + * + detecting PCMCIA Card Removal in interrupt handler. If + * ISRP is FF, then a PCMCIA card has been removed + * 10/2000 Burt needed a new method to avoid crashing the OS + * + * Changes by Paul Norton (pnorton@cts.com) : + * + restructured the READ.LOG logic to prevent the transmit SRB + * from being rudely overwritten before the transmit cycle is + * complete. (August 15 1996) + * + completed multiple adapter support. (November 20 1996) + * + implemented csum_partial_copy in tr_rx and increased receive + * buffer size and count. Minor fixes. (March 15, 1997) + * + * Changes by Christopher Turcksin <wabbit@rtfc.demon.co.uk> + * + Now compiles ok as a module again. + * + * Changes by Paul Norton (pnorton@ieee.org) : + * + moved the header manipulation code in tr_tx and tr_rx to + * net/802/tr.c. (July 12 1997) + * + add retry and timeout on open if cable disconnected. (May 5 1998) + * + lifted 2000 byte mtu limit. now depends on shared-RAM size. + * May 25 1998) + * + can't allocate 2k recv buff at 8k shared-RAM. (20 October 1998) + * + * Changes by Joel Sloan (jjs@c-me.com) : + * + disable verbose debug messages by default - to enable verbose + * debugging, edit the IBMTR_DEBUG_MESSAGES define below + * + * Changes by Mike Phillips <phillim@amtrak.com> : + * + Added extra #ifdef's to work with new PCMCIA Token Ring Code. + * The PCMCIA code now just sets up the card so it can be recognized + * by ibmtr_probe. Also checks allocated memory vs. on-board memory + * for correct figure to use. + * + * Changes by Tim Hockin (thockin@isunix.it.ilstu.edu) : + * + added spinlocks for SMP sanity (10 March 1999) + * + * Changes by Jochen Friedrich to enable RFC1469 Option 2 multicasting + * i.e. using functional address C0 00 00 04 00 00 to transmit and + * receive multicast packets. + * + * Changes by Mike Sullivan (based on original sram patch by Dave Grothe + * to support windowing into on adapter shared ram. + * i.e. Use LANAID to setup a PnP configuration with 16K RAM. Paging + * will shift this 16K window over the entire available shared RAM. + * + * Changes by Peter De Schrijver (p2@mind.be) : + * + fixed a problem with PCMCIA card removal + * + * Change by Mike Sullivan et al.: + * + added turbo card support. No need to use lanaid to configure + * the adapter into isa compatiblity mode. + * + * Changes by Burt Silverman to allow the computer to behave nicely when + * a cable is pulled or not in place, or a PCMCIA card is removed hot. + */ + +/* change the define of IBMTR_DEBUG_MESSAGES to a nonzero value +in the event that chatty debug messages are desired - jjs 12/30/98 */ + +#define IBMTR_DEBUG_MESSAGES 0 + +#include <linux/module.h> + +#ifdef PCMCIA /* required for ibmtr_cs.c to build */ +#undef MODULE /* yes, really */ +#undef ENABLE_PAGING +#else +#define ENABLE_PAGING 1 +#endif + +#define FALSE 0 +#define TRUE (!FALSE) + +/* changes the output format of driver initialization */ +#define TR_VERBOSE 0 + +/* some 95 OS send many non UI frame; this allow removing the warning */ +#define TR_FILTERNONUI 1 + +#include <linux/ioport.h> +#include <linux/netdevice.h> +#include <linux/ip.h> +#include <linux/trdevice.h> +#include <linux/ibmtr.h> + +#include <net/checksum.h> + +#include <asm/io.h> + +#define DPRINTK(format, args...) printk("%s: " format, dev->name , ## args) +#define DPRINTD(format, args...) DummyCall("%s: " format, dev->name , ## args) + +/* version and credits */ +#ifndef PCMCIA +static char version[] __initdata = + "\nibmtr.c: v1.3.57 8/ 7/94 Peter De Schrijver and Mark Swanson\n" + " v2.1.125 10/20/98 Paul Norton <pnorton@ieee.org>\n" + " v2.2.0 12/30/98 Joel Sloan <jjs@c-me.com>\n" + " v2.2.1 02/08/00 Mike Sullivan <sullivam@us.ibm.com>\n" + " v2.2.2 07/27/00 Burt Silverman <burts@us.ibm.com>\n" + " v2.4.0 03/01/01 Mike Sullivan <sullivan@us.ibm.com>\n"; +#endif + +/* this allows displaying full adapter information */ + +char *channel_def[] __devinitdata = { "ISA", "MCA", "ISA P&P" }; + +static char pcchannelid[] __devinitdata = { + 0x05, 0x00, 0x04, 0x09, + 0x04, 0x03, 0x04, 0x0f, + 0x03, 0x06, 0x03, 0x01, + 0x03, 0x01, 0x03, 0x00, + 0x03, 0x09, 0x03, 0x09, + 0x03, 0x00, 0x02, 0x00 +}; + +static char mcchannelid[] __devinitdata = { + 0x04, 0x0d, 0x04, 0x01, + 0x05, 0x02, 0x05, 0x03, + 0x03, 0x06, 0x03, 0x03, + 0x05, 0x08, 0x03, 0x04, + 0x03, 0x05, 0x03, 0x01, + 0x03, 0x08, 0x02, 0x00 +}; + +char __devinit *adapter_def(char type) +{ + switch (type) { + case 0xF: return "PC Adapter | PC Adapter II | Adapter/A"; + case 0xE: return "16/4 Adapter | 16/4 Adapter/A (long)"; + case 0xD: return "16/4 Adapter/A (short) | 16/4 ISA-16 Adapter"; + case 0xC: return "Auto 16/4 Adapter"; + default: return "adapter (unknown type)"; + }; +}; + +#define TRC_INIT 0x01 /* Trace initialization & PROBEs */ +#define TRC_INITV 0x02 /* verbose init trace points */ +unsigned char ibmtr_debug_trace = 0; + +static int ibmtr_probe(struct net_device *dev); +static int ibmtr_probe1(struct net_device *dev, int ioaddr); +static unsigned char get_sram_size(struct tok_info *adapt_info); +static int trdev_init(struct net_device *dev); +static int tok_open(struct net_device *dev); +static int tok_init_card(struct net_device *dev); +void tok_open_adapter(unsigned long dev_addr); +static void open_sap(unsigned char type, struct net_device *dev); +static void tok_set_multicast_list(struct net_device *dev); +static int tok_send_packet(struct sk_buff *skb, struct net_device *dev); +static int tok_close(struct net_device *dev); +irqreturn_t tok_interrupt(int irq, void *dev_id, struct pt_regs *regs); +static void initial_tok_int(struct net_device *dev); +static void tr_tx(struct net_device *dev); +static void tr_rx(struct net_device *dev); +void ibmtr_reset_timer(struct timer_list*tmr,struct net_device *dev); +static void tok_rerun(unsigned long dev_addr); +void ibmtr_readlog(struct net_device *dev); +static struct net_device_stats *tok_get_stats(struct net_device *dev); +int ibmtr_change_mtu(struct net_device *dev, int mtu); +static void find_turbo_adapters(int *iolist); + +static int ibmtr_portlist[IBMTR_MAX_ADAPTERS+1] __devinitdata = { + 0xa20, 0xa24, 0, 0, 0 +}; +static int __devinitdata turbo_io[IBMTR_MAX_ADAPTERS] = {0}; +static int __devinitdata turbo_irq[IBMTR_MAX_ADAPTERS] = {0}; +static int __devinitdata turbo_searched = 0; + +#ifndef PCMCIA +static __u32 ibmtr_mem_base __initdata = 0xd0000; +#endif + +static void __devinit PrtChanID(char *pcid, short stride) +{ + short i, j; + for (i = 0, j = 0; i < 24; i++, j += stride) + printk("%1x", ((int) pcid[j]) & 0x0f); + printk("\n"); +} + +static void __devinit HWPrtChanID(void __iomem *pcid, short stride) +{ + short i, j; + for (i = 0, j = 0; i < 24; i++, j += stride) + printk("%1x", ((int) readb(pcid + j)) & 0x0f); + printk("\n"); +} + +/* We have to ioremap every checked address, because isa_readb is + * going away. + */ + +static void __devinit find_turbo_adapters(int *iolist) +{ + int ram_addr; + int index=0; + void __iomem *chanid; + int found_turbo=0; + unsigned char *tchanid, ctemp; + int i, j; + unsigned long jif; + void __iomem *ram_mapped ; + + if (turbo_searched == 1) return; + turbo_searched=1; + for (ram_addr=0xC0000; ram_addr < 0xE0000; ram_addr+=0x2000) { + + __u32 intf_tbl=0; + + found_turbo=1; + ram_mapped = ioremap((u32)ram_addr,0x1fff) ; + if (ram_mapped==NULL) + continue ; + chanid=(CHANNEL_ID + ram_mapped); + tchanid=pcchannelid; + ctemp=readb(chanid) & 0x0f; + if (ctemp != *tchanid) continue; + for (i=2,j=1; i<=46; i=i+2,j++) { + if ((readb(chanid+i) & 0x0f) != tchanid[j]){ + found_turbo=0; + break; + } + } + if (!found_turbo) continue; + + writeb(0x90, ram_mapped+0x1E01); + for(i=2; i<0x0f; i++) { + writeb(0x00, ram_mapped+0x1E01+i); + } + writeb(0x00, ram_mapped+0x1E01); + for(jif=jiffies+TR_BUSY_INTERVAL; time_before_eq(jiffies,jif);); + intf_tbl=ntohs(readw(ram_mapped+ACA_OFFSET+ACA_RW+WRBR_EVEN)); + if (intf_tbl) { +#if IBMTR_DEBUG_MESSAGES + printk("ibmtr::find_turbo_adapters, Turbo found at " + "ram_addr %x\n",ram_addr); + printk("ibmtr::find_turbo_adapters, interface_table "); + for(i=0; i<6; i++) { + printk("%x:",readb(ram_addr+intf_tbl+i)); + } + printk("\n"); +#endif + turbo_io[index]=ntohs(readw(ram_mapped+intf_tbl+4)); + turbo_irq[index]=readb(ram_mapped+intf_tbl+3); + outb(0, turbo_io[index] + ADAPTRESET); + for(jif=jiffies+TR_RST_TIME;time_before_eq(jiffies,jif);); + outb(0, turbo_io[index] + ADAPTRESETREL); + index++; + continue; + } +#if IBMTR_DEBUG_MESSAGES + printk("ibmtr::find_turbo_adapters, ibmtr card found at" + " %x but not a Turbo model\n",ram_addr); +#endif + iounmap(ram_mapped) ; + } /* for */ + for(i=0; i<IBMTR_MAX_ADAPTERS; i++) { + if(!turbo_io[i]) break; + for (j=0; j<IBMTR_MAX_ADAPTERS; j++) { + if ( iolist[j] && iolist[j] != turbo_io[i]) continue; + iolist[j]=turbo_io[i]; + break; + } + } +} + +static void ibmtr_cleanup_card(struct net_device *dev) +{ + if (dev->base_addr) { + outb(0,dev->base_addr+ADAPTRESET); + + schedule_timeout(TR_RST_TIME); /* wait 50ms */ + + outb(0,dev->base_addr+ADAPTRESETREL); + } + +#ifndef PCMCIA + free_irq(dev->irq, dev); + release_region(dev->base_addr, IBMTR_IO_EXTENT); + + { + struct tok_info *ti = (struct tok_info *) dev->priv; + iounmap(ti->mmio); + iounmap(ti->sram_virt); + } +#endif +} + +int ibmtr_probe_card(struct net_device *dev) +{ + int err = ibmtr_probe(dev); + if (!err) { + err = register_netdev(dev); + if (err) + ibmtr_cleanup_card(dev); + } + return err; +} + +/**************************************************************************** + * ibmtr_probe(): Routine specified in the network device structure + * to probe for an IBM Token Ring Adapter. Routine outline: + * I. Interrogate hardware to determine if an adapter exists + * and what the speeds and feeds are + * II. Setup data structures to control execution based upon + * adapter characteristics. + * + * We expect ibmtr_probe to be called once for each device entry + * which references it. + ****************************************************************************/ + +static int ibmtr_probe(struct net_device *dev) +{ + int i; + int base_addr = dev->base_addr; + + if (base_addr && base_addr <= 0x1ff) /* Don't probe at all. */ + return -ENXIO; + if (base_addr > 0x1ff) { /* Check a single specified location. */ + if (!ibmtr_probe1(dev, base_addr)) return 0; + return -ENODEV; + } + find_turbo_adapters(ibmtr_portlist); + for (i = 0; ibmtr_portlist[i]; i++) { + int ioaddr = ibmtr_portlist[i]; + + if (!ibmtr_probe1(dev, ioaddr)) return 0; + } + return -ENODEV; +} + +/*****************************************************************************/ + +static int __devinit ibmtr_probe1(struct net_device *dev, int PIOaddr) +{ + + unsigned char segment, intr=0, irq=0, i, j, cardpresent=NOTOK, temp=0; + void __iomem * t_mmio = NULL; + struct tok_info *ti = dev->priv; + void __iomem *cd_chanid; + unsigned char *tchanid, ctemp; +#ifndef PCMCIA + unsigned char t_irq=0; + unsigned long timeout; + static int version_printed; +#endif + + /* Query the adapter PIO base port which will return + * indication of where MMIO was placed. We also have a + * coded interrupt number. + */ + segment = inb(PIOaddr); + if (segment < 0x40 || segment > 0xe0) { + /* Out of range values so we'll assume non-existent IO device + * but this is not necessarily a problem, esp if a turbo + * adapter is being used. */ +#if IBMTR_DEBUG_MESSAGES + DPRINTK("ibmtr_probe1(): unhappy that inb(0x%X) == 0x%X, " + "Hardware Problem?\n",PIOaddr,segment); +#endif + return -ENODEV; + } + /* + * Compute the linear base address of the MMIO area + * as LINUX doesn't care about segments + */ + t_mmio = ioremap(((__u32) (segment & 0xfc) << 11) + 0x80000,2048); + if (!t_mmio) { + DPRINTK("Cannot remap mmiobase memory area") ; + return -ENODEV ; + } + intr = segment & 0x03; /* low bits is coded interrupt # */ + if (ibmtr_debug_trace & TRC_INIT) + DPRINTK("PIOaddr: %4hx seg/intr: %2x mmio base: %p intr: %d\n" + , PIOaddr, (int) segment, t_mmio, (int) intr); + + /* + * Now we will compare expected 'channelid' strings with + * what we is there to learn of ISA/MCA or not TR card + */ +#ifdef PCMCIA + iounmap(t_mmio); + t_mmio = ti->mmio; /*BMS to get virtual address */ + irq = ti->irq; /*BMS to display the irq! */ +#endif + cd_chanid = (CHANNEL_ID + t_mmio); /* for efficiency */ + tchanid = pcchannelid; + cardpresent = TR_ISA; /* try ISA */ + + /* Suboptimize knowing first byte different */ + ctemp = readb(cd_chanid) & 0x0f; + if (ctemp != *tchanid) { /* NOT ISA card, try MCA */ + tchanid = mcchannelid; + cardpresent = TR_MCA; + if (ctemp != *tchanid) /* Neither ISA nor MCA */ + cardpresent = NOTOK; + } + if (cardpresent != NOTOK) { + /* Know presumed type, try rest of ID */ + for (i = 2, j = 1; i <= 46; i = i + 2, j++) { + if( (readb(cd_chanid+i)&0x0f) == tchanid[j]) continue; + /* match failed, not TR card */ + cardpresent = NOTOK; + break; + } + } + /* + * If we have an ISA board check for the ISA P&P version, + * as it has different IRQ settings + */ + if (cardpresent == TR_ISA && (readb(AIPFID + t_mmio) == 0x0e)) + cardpresent = TR_ISAPNP; + if (cardpresent == NOTOK) { /* "channel_id" did not match, report */ + if (!(ibmtr_debug_trace & TRC_INIT)) { +#ifndef PCMCIA + iounmap(t_mmio); +#endif + return -ENODEV; + } + DPRINTK( "Channel ID string not found for PIOaddr: %4hx\n", + PIOaddr); + DPRINTK("Expected for ISA: "); + PrtChanID(pcchannelid, 1); + DPRINTK(" found: "); +/* BMS Note that this can be misleading, when hardware is flaky, because you + are reading it a second time here. So with my flaky hardware, I'll see my- + self in this block, with the HW ID matching the ISA ID exactly! */ + HWPrtChanID(cd_chanid, 2); + DPRINTK("Expected for MCA: "); + PrtChanID(mcchannelid, 1); + } + /* Now, setup some of the pl0 buffers for this driver.. */ + /* If called from PCMCIA, it is already set up, so no need to + waste the memory, just use the existing structure */ +#ifndef PCMCIA + ti->mmio = t_mmio; + for (i = 0; i < IBMTR_MAX_ADAPTERS; i++) { + if (turbo_io[i] != PIOaddr) + continue; +#if IBMTR_DEBUG_MESSAGES + printk("ibmtr::tr_probe1, setting PIOaddr %x to Turbo\n", + PIOaddr); +#endif + ti->turbo = 1; + t_irq = turbo_irq[i]; + } +#endif /* !PCMCIA */ + ti->readlog_pending = 0; + init_waitqueue_head(&ti->wait_for_reset); + + /* if PCMCIA, the card can be recognized as either TR_ISA or TR_ISAPNP + * depending which card is inserted. */ + +#ifndef PCMCIA + switch (cardpresent) { + case TR_ISA: + if (intr == 0) irq = 9; /* irq2 really is irq9 */ + if (intr == 1) irq = 3; + if (intr == 2) irq = 6; + if (intr == 3) irq = 7; + ti->adapter_int_enable = PIOaddr + ADAPTINTREL; + break; + case TR_MCA: + if (intr == 0) irq = 9; + if (intr == 1) irq = 3; + if (intr == 2) irq = 10; + if (intr == 3) irq = 11; + ti->global_int_enable = 0; + ti->adapter_int_enable = 0; + ti->sram_phys=(__u32)(inb(PIOaddr+ADAPTRESETREL) & 0xfe) << 12; + break; + case TR_ISAPNP: + if (!t_irq) { + if (intr == 0) irq = 9; + if (intr == 1) irq = 3; + if (intr == 2) irq = 10; + if (intr == 3) irq = 11; + } else + irq=t_irq; + timeout = jiffies + TR_SPIN_INTERVAL; + while (!readb(ti->mmio + ACA_OFFSET + ACA_RW + RRR_EVEN)){ + if (!time_after(jiffies, timeout)) continue; + DPRINTK( "Hardware timeout during initialization.\n"); + iounmap(t_mmio); + kfree(ti); + return -ENODEV; + } + ti->sram_phys = + ((__u32)readb(ti->mmio+ACA_OFFSET+ACA_RW+RRR_EVEN)<<12); + ti->adapter_int_enable = PIOaddr + ADAPTINTREL; + break; + } /*end switch (cardpresent) */ +#endif /*not PCMCIA */ + + if (ibmtr_debug_trace & TRC_INIT) { /* just report int */ + DPRINTK("irq=%d", irq); + printk(", sram_phys=0x%x", ti->sram_phys); + if(ibmtr_debug_trace&TRC_INITV){ /* full chat in verbose only */ + DPRINTK(", ti->mmio=%p", ti->mmio); + printk(", segment=%02X", segment); + } + printk(".\n"); + } + + /* Get hw address of token ring card */ + j = 0; + for (i = 0; i < 0x18; i = i + 2) { + /* technical reference states to do this */ + temp = readb(ti->mmio + AIP + i) & 0x0f; + ti->hw_address[j] = temp; + if (j & 1) + dev->dev_addr[(j / 2)] = + ti->hw_address[j]+ (ti->hw_address[j - 1] << 4); + ++j; + } + /* get Adapter type: 'F' = Adapter/A, 'E' = 16/4 Adapter II,... */ + ti->adapter_type = readb(ti->mmio + AIPADAPTYPE); + + /* get Data Rate: F=4Mb, E=16Mb, D=4Mb & 16Mb ?? */ + ti->data_rate = readb(ti->mmio + AIPDATARATE); + + /* Get Early Token Release support?: F=no, E=4Mb, D=16Mb, C=4&16Mb */ + ti->token_release = readb(ti->mmio + AIPEARLYTOKEN); + + /* How much shared RAM is on adapter ? */ + if (ti->turbo) { + ti->avail_shared_ram=127; + } else { + ti->avail_shared_ram = get_sram_size(ti);/*in 512 byte units */ + } + /* We need to set or do a bunch of work here based on previous results*/ + /* Support paging? What sizes?: F=no, E=16k, D=32k, C=16 & 32k */ + ti->shared_ram_paging = readb(ti->mmio + AIPSHRAMPAGE); + + /* Available DHB 4Mb size: F=2048, E=4096, D=4464 */ + switch (readb(ti->mmio + AIP4MBDHB)) { + case 0xe: ti->dhb_size4mb = 4096; break; + case 0xd: ti->dhb_size4mb = 4464; break; + default: ti->dhb_size4mb = 2048; break; + } + + /* Available DHB 16Mb size: F=2048, E=4096, D=8192, C=16384, B=17960 */ + switch (readb(ti->mmio + AIP16MBDHB)) { + case 0xe: ti->dhb_size16mb = 4096; break; + case 0xd: ti->dhb_size16mb = 8192; break; + case 0xc: ti->dhb_size16mb = 16384; break; + case 0xb: ti->dhb_size16mb = 17960; break; + default: ti->dhb_size16mb = 2048; break; + } + + /* We must figure out how much shared memory space this adapter + * will occupy so that if there are two adapters we can fit both + * in. Given a choice, we will limit this adapter to 32K. The + * maximum space will will use for two adapters is 64K so if the + * adapter we are working on demands 64K (it also doesn't support + * paging), then only one adapter can be supported. + */ + + /* + * determine how much of total RAM is mapped into PC space + */ + ti->mapped_ram_size= /*sixteen to onehundredtwentyeight 512byte blocks*/ + 1<< ((readb(ti->mmio+ACA_OFFSET+ACA_RW+RRR_ODD) >> 2 & 0x03) + 4); + ti->page_mask = 0; + if (ti->turbo) ti->page_mask=0xf0; + else if (ti->shared_ram_paging == 0xf); /* No paging in adapter */ + else { +#ifdef ENABLE_PAGING + unsigned char pg_size = 0; + /* BMS: page size: PCMCIA, use configuration register; + ISAPNP, use LANAIDC config tool from www.ibm.com */ + switch (ti->shared_ram_paging) { + case 0xf: + break; + case 0xe: + ti->page_mask = (ti->mapped_ram_size == 32) ? 0xc0 : 0; + pg_size = 32; /* 16KB page size */ + break; + case 0xd: + ti->page_mask = (ti->mapped_ram_size == 64) ? 0x80 : 0; + pg_size = 64; /* 32KB page size */ + break; + case 0xc: + switch (ti->mapped_ram_size) { + case 32: + ti->page_mask = 0xc0; + pg_size = 32; + break; + case 64: + ti->page_mask = 0x80; + pg_size = 64; + break; + } + break; + default: + DPRINTK("Unknown shared ram paging info %01X\n", + ti->shared_ram_paging); + iounmap(t_mmio); + kfree(ti); + return -ENODEV; + break; + } /*end switch shared_ram_paging */ + + if (ibmtr_debug_trace & TRC_INIT) + DPRINTK("Shared RAM paging code: %02X, " + "mapped RAM size: %dK, shared RAM size: %dK, " + "page mask: %02X\n:", + ti->shared_ram_paging, ti->mapped_ram_size / 2, + ti->avail_shared_ram / 2, ti->page_mask); +#endif /*ENABLE_PAGING */ + } + +#ifndef PCMCIA + /* finish figuring the shared RAM address */ + if (cardpresent == TR_ISA) { + static __u32 ram_bndry_mask[] = + { 0xffffe000, 0xffffc000, 0xffff8000, 0xffff0000 }; + __u32 new_base, rrr_32, chk_base, rbm; + + rrr_32=readb(ti->mmio+ACA_OFFSET+ACA_RW+RRR_ODD) >> 2 & 0x03; + rbm = ram_bndry_mask[rrr_32]; + new_base = (ibmtr_mem_base + (~rbm)) & rbm;/* up to boundary */ + chk_base = new_base + (ti->mapped_ram_size << 9); + if (chk_base > (ibmtr_mem_base + IBMTR_SHARED_RAM_SIZE)) { + DPRINTK("Shared RAM for this adapter (%05x) exceeds " + "driver limit (%05x), adapter not started.\n", + chk_base, ibmtr_mem_base + IBMTR_SHARED_RAM_SIZE); + iounmap(t_mmio); + kfree(ti); + return -ENODEV; + } else { /* seems cool, record what we have figured out */ + ti->sram_base = new_base >> 12; + ibmtr_mem_base = chk_base; + } + } + else ti->sram_base = ti->sram_phys >> 12; + + /* The PCMCIA has already got the interrupt line and the io port, + so no chance of anybody else getting it - MLP */ + if (request_irq(dev->irq = irq, &tok_interrupt, 0, "ibmtr", dev) != 0) { + DPRINTK("Could not grab irq %d. Halting Token Ring driver.\n", + irq); + iounmap(t_mmio); + kfree(ti); + return -ENODEV; + } + /*?? Now, allocate some of the PIO PORTs for this driver.. */ + /* record PIOaddr range as busy */ + if (!request_region(PIOaddr, IBMTR_IO_EXTENT, "ibmtr")) { + DPRINTK("Could not grab PIO range. Halting driver.\n"); + free_irq(dev->irq, dev); + iounmap(t_mmio); + kfree(ti); + return -EBUSY; + } + + if (!version_printed++) { + printk(version); + } +#endif /* !PCMCIA */ + DPRINTK("%s %s found\n", + channel_def[cardpresent - 1], adapter_def(ti->adapter_type)); + DPRINTK("using irq %d, PIOaddr %hx, %dK shared RAM.\n", + irq, PIOaddr, ti->mapped_ram_size / 2); + DPRINTK("Hardware address : %02X:%02X:%02X:%02X:%02X:%02X\n", + dev->dev_addr[0], dev->dev_addr[1], dev->dev_addr[2], + dev->dev_addr[3], dev->dev_addr[4], dev->dev_addr[5]); + if (ti->page_mask) + DPRINTK("Shared RAM paging enabled. " + "Page size: %uK Shared Ram size %dK\n", + ((ti->page_mask^0xff)+1) >>2, ti->avail_shared_ram / 2); + else + DPRINTK("Shared RAM paging disabled. ti->page_mask %x\n", + ti->page_mask); + + /* Calculate the maximum DHB we can use */ + /* two cases where avail_shared_ram doesn't equal mapped_ram_size: + 1. avail_shared_ram is 127 but mapped_ram_size is 128 (typical) + 2. user has configured adapter for less than avail_shared_ram + but is not using paging (she should use paging, I believe) + */ + if (!ti->page_mask) { + ti->avail_shared_ram= + min(ti->mapped_ram_size,ti->avail_shared_ram); + } + + switch (ti->avail_shared_ram) { + case 16: /* 8KB shared RAM */ + ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)2048); + ti->rbuf_len4 = 1032; + ti->rbuf_cnt4=2; + ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)2048); + ti->rbuf_len16 = 1032; + ti->rbuf_cnt16=2; + break; + case 32: /* 16KB shared RAM */ + ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464); + ti->rbuf_len4 = 1032; + ti->rbuf_cnt4=4; + ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)4096); + ti->rbuf_len16 = 1032; /*1024 usable */ + ti->rbuf_cnt16=4; + break; + case 64: /* 32KB shared RAM */ + ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464); + ti->rbuf_len4 = 1032; + ti->rbuf_cnt4=6; + ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)10240); + ti->rbuf_len16 = 1032; + ti->rbuf_cnt16=6; + break; + case 127: /* 63.5KB shared RAM */ + ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464); + ti->rbuf_len4 = 1032; + ti->rbuf_cnt4=6; + ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)16384); + ti->rbuf_len16 = 1032; + ti->rbuf_cnt16=16; + break; + case 128: /* 64KB shared RAM */ + ti->dhb_size4mb = min(ti->dhb_size4mb, (unsigned short)4464); + ti->rbuf_len4 = 1032; + ti->rbuf_cnt4=6; + ti->dhb_size16mb = min(ti->dhb_size16mb, (unsigned short)17960); + ti->rbuf_len16 = 1032; + ti->rbuf_cnt16=16; + break; + default: + ti->dhb_size4mb = 2048; + ti->rbuf_len4 = 1032; + ti->rbuf_cnt4=2; + ti->dhb_size16mb = 2048; + ti->rbuf_len16 = 1032; + ti->rbuf_cnt16=2; + break; + } + /* this formula is not smart enough for the paging case + ti->rbuf_cnt<x> = (ti->avail_shared_ram * BLOCKSZ - ADAPT_PRIVATE - + ARBLENGTH - SSBLENGTH - DLC_MAX_SAP * SAPLENGTH - + DLC_MAX_STA * STALENGTH - ti->dhb_size<x>mb * NUM_DHB - + SRBLENGTH - ASBLENGTH) / ti->rbuf_len<x>; + */ + ti->maxmtu16 = (ti->rbuf_len16 - 8) * ti->rbuf_cnt16 - TR_HLEN; + ti->maxmtu4 = (ti->rbuf_len4 - 8) * ti->rbuf_cnt4 - TR_HLEN; + /*BMS assuming 18 bytes of Routing Information (usually works) */ + DPRINTK("Maximum Receive Internet Protocol MTU 16Mbps: %d, 4Mbps: %d\n", + ti->maxmtu16, ti->maxmtu4); + + dev->base_addr = PIOaddr; /* set the value for device */ + dev->mem_start = ti->sram_base << 12; + dev->mem_end = dev->mem_start + (ti->mapped_ram_size << 9) - 1; + trdev_init(dev); + return 0; /* Return 0 to indicate we have found a Token Ring card. */ +} /*ibmtr_probe1() */ + +/*****************************************************************************/ + +/* query the adapter for the size of shared RAM */ +/* the function returns the RAM size in units of 512 bytes */ + +static unsigned char __devinit get_sram_size(struct tok_info *adapt_info) +{ + unsigned char avail_sram_code; + static unsigned char size_code[] = { 0, 16, 32, 64, 127, 128 }; + /* Adapter gives + 'F' -- use RRR bits 3,2 + 'E' -- 8kb 'D' -- 16kb + 'C' -- 32kb 'A' -- 64KB + 'B' - 64KB less 512 bytes at top + (WARNING ... must zero top bytes in INIT */ + + avail_sram_code = 0xf - readb(adapt_info->mmio + AIPAVAILSHRAM); + if (avail_sram_code) return size_code[avail_sram_code]; + else /* for code 'F', must compute size from RRR(3,2) bits */ + return 1 << + ((readb(adapt_info->mmio+ACA_OFFSET+ACA_RW+RRR_ODD)>>2&3)+4); +} + +/*****************************************************************************/ + +static int __devinit trdev_init(struct net_device *dev) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + + SET_PAGE(ti->srb_page); + ti->open_failure = NO ; + dev->open = tok_open; + dev->stop = tok_close; + dev->hard_start_xmit = tok_send_packet; + dev->get_stats = tok_get_stats; + dev->set_multicast_list = tok_set_multicast_list; + dev->change_mtu = ibmtr_change_mtu; + + return 0; +} + +/*****************************************************************************/ + +static int tok_init_card(struct net_device *dev) +{ + struct tok_info *ti; + short PIOaddr; + unsigned long i; + + PIOaddr = dev->base_addr; + ti = (struct tok_info *) dev->priv; + /* Special processing for first interrupt after reset */ + ti->do_tok_int = FIRST_INT; + /* Reset adapter */ + writeb(~INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_EVEN); + outb(0, PIOaddr + ADAPTRESET); + + current->state=TASK_UNINTERRUPTIBLE; + schedule_timeout(TR_RST_TIME); /* wait 50ms */ + + outb(0, PIOaddr + ADAPTRESETREL); +#ifdef ENABLE_PAGING + if (ti->page_mask) + writeb(SRPR_ENABLE_PAGING,ti->mmio+ACA_OFFSET+ACA_RW+SRPR_EVEN); +#endif + writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN); + i = sleep_on_timeout(&ti->wait_for_reset, 4 * HZ); + return i? 0 : -EAGAIN; +} + +/*****************************************************************************/ +static int tok_open(struct net_device *dev) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + int i; + + /*the case we were left in a failure state during a previous open */ + if (ti->open_failure == YES) { + DPRINTK("Last time you were disconnected, how about now?\n"); + printk("You can't insert with an ICS connector half-cocked.\n"); + } + + ti->open_status = CLOSED; /* CLOSED or OPEN */ + ti->sap_status = CLOSED; /* CLOSED or OPEN */ + ti->open_failure = NO; /* NO or YES */ + ti->open_mode = MANUAL; /* MANUAL or AUTOMATIC */ + /* 12/2000 not typical Linux, but we can use RUNNING to let us know when + the network has crapped out or cables are disconnected. Useful because + the IFF_UP flag stays up the whole time, until ifconfig tr0 down. + */ + dev->flags &= ~IFF_RUNNING; + + ti->sram_phys &= ~1; /* to reverse what we do in tok_close */ + /* init the spinlock */ + spin_lock_init(&ti->lock); + init_timer(&ti->tr_timer); + + i = tok_init_card(dev); + if (i) return i; + + while (1){ + tok_open_adapter((unsigned long) dev); + i= interruptible_sleep_on_timeout(&ti->wait_for_reset, 25 * HZ); + /* sig catch: estimate opening adapter takes more than .5 sec*/ + if (i>(245*HZ)/10) break; /* fancier than if (i==25*HZ) */ + if (i==0) break; + if (ti->open_status == OPEN && ti->sap_status==OPEN) { + netif_start_queue(dev); + DPRINTK("Adapter is up and running\n"); + return 0; + } + current->state=TASK_INTERRUPTIBLE; + i=schedule_timeout(TR_RETRY_INTERVAL); /* wait 30 seconds */ + if(i!=0) break; /*prob. a signal, like the i>24*HZ case above */ + } + outb(0, dev->base_addr + ADAPTRESET);/* kill pending interrupts*/ + DPRINTK("TERMINATED via signal\n"); /*BMS useful */ + return -EAGAIN; +} + +/*****************************************************************************/ + +#define COMMAND_OFST 0 +#define OPEN_OPTIONS_OFST 8 +#define NUM_RCV_BUF_OFST 24 +#define RCV_BUF_LEN_OFST 26 +#define DHB_LENGTH_OFST 28 +#define NUM_DHB_OFST 30 +#define DLC_MAX_SAP_OFST 32 +#define DLC_MAX_STA_OFST 33 + +void tok_open_adapter(unsigned long dev_addr) +{ + struct net_device *dev = (struct net_device *) dev_addr; + struct tok_info *ti; + int i; + + ti = (struct tok_info *) dev->priv; + SET_PAGE(ti->init_srb_page); + writeb(~SRB_RESP_INT, ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_ODD); + for (i = 0; i < sizeof(struct dir_open_adapter); i++) + writeb(0, ti->init_srb + i); + writeb(DIR_OPEN_ADAPTER, ti->init_srb + COMMAND_OFST); + writew(htons(OPEN_PASS_BCON_MAC), ti->init_srb + OPEN_OPTIONS_OFST); + if (ti->ring_speed == 16) { + writew(htons(ti->dhb_size16mb), ti->init_srb + DHB_LENGTH_OFST); + writew(htons(ti->rbuf_cnt16), ti->init_srb + NUM_RCV_BUF_OFST); + writew(htons(ti->rbuf_len16), ti->init_srb + RCV_BUF_LEN_OFST); + } else { + writew(htons(ti->dhb_size4mb), ti->init_srb + DHB_LENGTH_OFST); + writew(htons(ti->rbuf_cnt4), ti->init_srb + NUM_RCV_BUF_OFST); + writew(htons(ti->rbuf_len4), ti->init_srb + RCV_BUF_LEN_OFST); + } + writeb(NUM_DHB, /* always 2 */ ti->init_srb + NUM_DHB_OFST); + writeb(DLC_MAX_SAP, ti->init_srb + DLC_MAX_SAP_OFST); + writeb(DLC_MAX_STA, ti->init_srb + DLC_MAX_STA_OFST); + ti->srb = ti->init_srb; /* We use this one in the interrupt handler */ + ti->srb_page = ti->init_srb_page; + DPRINTK("Opening adapter: Xmit bfrs: %d X %d, Rcv bfrs: %d X %d\n", + readb(ti->init_srb + NUM_DHB_OFST), + ntohs(readw(ti->init_srb + DHB_LENGTH_OFST)), + ntohs(readw(ti->init_srb + NUM_RCV_BUF_OFST)), + ntohs(readw(ti->init_srb + RCV_BUF_LEN_OFST))); + writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN); + writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); +} + +/*****************************************************************************/ + +static void open_sap(unsigned char type, struct net_device *dev) +{ + int i; + struct tok_info *ti = (struct tok_info *) dev->priv; + + SET_PAGE(ti->srb_page); + for (i = 0; i < sizeof(struct dlc_open_sap); i++) + writeb(0, ti->srb + i); + +#define MAX_I_FIELD_OFST 14 +#define SAP_VALUE_OFST 16 +#define SAP_OPTIONS_OFST 17 +#define STATION_COUNT_OFST 18 + + writeb(DLC_OPEN_SAP, ti->srb + COMMAND_OFST); + writew(htons(MAX_I_FIELD), ti->srb + MAX_I_FIELD_OFST); + writeb(SAP_OPEN_IND_SAP | SAP_OPEN_PRIORITY, ti->srb+ SAP_OPTIONS_OFST); + writeb(SAP_OPEN_STATION_CNT, ti->srb + STATION_COUNT_OFST); + writeb(type, ti->srb + SAP_VALUE_OFST); + writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); +} + + +/*****************************************************************************/ + +static void tok_set_multicast_list(struct net_device *dev) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + struct dev_mc_list *mclist; + unsigned char address[4]; + + int i; + + /*BMS the next line is CRUCIAL or you may be sad when you */ + /*BMS ifconfig tr down or hot unplug a PCMCIA card ??hownowbrowncow*/ + if (/*BMSHELPdev->start == 0 ||*/ ti->open_status != OPEN) return; + address[0] = address[1] = address[2] = address[3] = 0; + mclist = dev->mc_list; + for (i = 0; i < dev->mc_count; i++) { + address[0] |= mclist->dmi_addr[2]; + address[1] |= mclist->dmi_addr[3]; + address[2] |= mclist->dmi_addr[4]; + address[3] |= mclist->dmi_addr[5]; + mclist = mclist->next; + } + SET_PAGE(ti->srb_page); + for (i = 0; i < sizeof(struct srb_set_funct_addr); i++) + writeb(0, ti->srb + i); + +#define FUNCT_ADDRESS_OFST 6 + + writeb(DIR_SET_FUNC_ADDR, ti->srb + COMMAND_OFST); + for (i = 0; i < 4; i++) + writeb(address[i], ti->srb + FUNCT_ADDRESS_OFST + i); + writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); +#if TR_VERBOSE + DPRINTK("Setting functional address: "); + for (i=0;i<4;i++) printk("%02X ", address[i]); + printk("\n"); +#endif +} + +/*****************************************************************************/ + +#define STATION_ID_OFST 4 + +static int tok_send_packet(struct sk_buff *skb, struct net_device *dev) +{ + struct tok_info *ti; + unsigned long flags; + ti = (struct tok_info *) dev->priv; + + netif_stop_queue(dev); + + /* lock against other CPUs */ + spin_lock_irqsave(&(ti->lock), flags); + + /* Save skb; we'll need it when the adapter asks for the data */ + ti->current_skb = skb; + SET_PAGE(ti->srb_page); + writeb(XMIT_UI_FRAME, ti->srb + COMMAND_OFST); + writew(ti->exsap_station_id, ti->srb + STATION_ID_OFST); + writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + spin_unlock_irqrestore(&(ti->lock), flags); + dev->trans_start = jiffies; + return 0; +} + +/*****************************************************************************/ + +static int tok_close(struct net_device *dev) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + + /* Important for PCMCIA hot unplug, otherwise, we'll pull the card, */ + /* unloading the module from memory, and then if a timer pops, ouch */ + del_timer_sync(&ti->tr_timer); + outb(0, dev->base_addr + ADAPTRESET); + ti->sram_phys |= 1; + ti->open_status = CLOSED; + + netif_stop_queue(dev); + DPRINTK("Adapter is closed.\n"); + return 0; +} + +/*****************************************************************************/ + +#define RETCODE_OFST 2 +#define OPEN_ERROR_CODE_OFST 6 +#define ASB_ADDRESS_OFST 8 +#define SRB_ADDRESS_OFST 10 +#define ARB_ADDRESS_OFST 12 +#define SSB_ADDRESS_OFST 14 + +static char *printphase[]= {"Lobe media test","Physical insertion", + "Address verification","Roll call poll","Request Parameters"}; +static char *printerror[]={"Function failure","Signal loss","Reserved", + "Frequency error","Timeout","Ring failure","Ring beaconing", + "Duplicate node address", + "Parameter request-retry count exceeded","Remove received", + "IMPL force received","Duplicate modifier", + "No monitor detected","Monitor contention failed for RPL"}; + +static void __iomem *map_address(struct tok_info *ti, unsigned index, __u8 *page) +{ + if (ti->page_mask) { + *page = (index >> 8) & ti->page_mask; + index &= ~(ti->page_mask << 8); + } + return ti->sram_virt + index; +} + +void dir_open_adapter (struct net_device *dev) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + unsigned char ret_code; + __u16 err; + + ti->srb = map_address(ti, + ntohs(readw(ti->init_srb + SRB_ADDRESS_OFST)), + &ti->srb_page); + ti->ssb = map_address(ti, + ntohs(readw(ti->init_srb + SSB_ADDRESS_OFST)), + &ti->ssb_page); + ti->arb = map_address(ti, + ntohs(readw(ti->init_srb + ARB_ADDRESS_OFST)), + &ti->arb_page); + ti->asb = map_address(ti, + ntohs(readw(ti->init_srb + ASB_ADDRESS_OFST)), + &ti->asb_page); + ti->current_skb = NULL; + ret_code = readb(ti->init_srb + RETCODE_OFST); + err = ntohs(readw(ti->init_srb + OPEN_ERROR_CODE_OFST)); + if (!ret_code) { + ti->open_status = OPEN; /* TR adapter is now available */ + if (ti->open_mode == AUTOMATIC) { + DPRINTK("Adapter reopened.\n"); + } + writeb(~SRB_RESP_INT, ti->mmio+ACA_OFFSET+ACA_RESET+ISRP_ODD); + open_sap(EXTENDED_SAP, dev); + return; + } + ti->open_failure = YES; + if (ret_code == 7){ + if (err == 0x24) { + if (!ti->auto_speedsave) { + DPRINTK("Open failed: Adapter speed must match " + "ring speed if Automatic Ring Speed Save is " + "disabled.\n"); + ti->open_action = FAIL; + }else + DPRINTK("Retrying open to adjust to " + "ring speed, "); + } else if (err == 0x2d) { + DPRINTK("Physical Insertion: No Monitor Detected, "); + printk("retrying after %ds delay...\n", + TR_RETRY_INTERVAL/HZ); + } else if (err == 0x11) { + DPRINTK("Lobe Media Function Failure (0x11), "); + printk(" retrying after %ds delay...\n", + TR_RETRY_INTERVAL/HZ); + } else { + char **prphase = printphase; + char **prerror = printerror; + DPRINTK("TR Adapter misc open failure, error code = "); + printk("0x%x, Phase: %s, Error: %s\n", + err, prphase[err/16 -1], prerror[err%16 -1]); + printk(" retrying after %ds delay...\n", + TR_RETRY_INTERVAL/HZ); + } + } else DPRINTK("open failed: ret_code = %02X..., ", ret_code); + if (ti->open_action != FAIL) { + if (ti->open_mode==AUTOMATIC){ + ti->open_action = REOPEN; + ibmtr_reset_timer(&(ti->tr_timer), dev); + return; + } + wake_up(&ti->wait_for_reset); + return; + } + DPRINTK("FAILURE, CAPUT\n"); +} + +/******************************************************************************/ + +irqreturn_t tok_interrupt(int irq, void *dev_id, struct pt_regs *regs) +{ + unsigned char status; + /* unsigned char status_even ; */ + struct tok_info *ti; + struct net_device *dev; +#ifdef ENABLE_PAGING + unsigned char save_srpr; +#endif + + dev = dev_id; +#if TR_VERBOSE + DPRINTK("Int from tok_driver, dev : %p irq%d regs=%p\n", dev,irq,regs); +#endif + ti = (struct tok_info *) dev->priv; + if (ti->sram_phys & 1) + return IRQ_NONE; /* PCMCIA card extraction flag */ + spin_lock(&(ti->lock)); +#ifdef ENABLE_PAGING + save_srpr = readb(ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN); +#endif + + /* Disable interrupts till processing is finished */ + writeb((~INT_ENABLE), ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_EVEN); + + /* Reset interrupt for ISA boards */ + if (ti->adapter_int_enable) + outb(0, ti->adapter_int_enable); + else /* used for PCMCIA cards */ + outb(0, ti->global_int_enable); + if (ti->do_tok_int == FIRST_INT){ + initial_tok_int(dev); +#ifdef ENABLE_PAGING + writeb(save_srpr, ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN); +#endif + spin_unlock(&(ti->lock)); + return IRQ_HANDLED; + } + /* Begin interrupt handler HERE inline to avoid the extra + levels of logic and call depth for the original solution. */ + status = readb(ti->mmio + ACA_OFFSET + ACA_RW + ISRP_ODD); + /*BMSstatus_even = readb (ti->mmio + ACA_OFFSET + ACA_RW + ISRP_EVEN) */ + /*BMSdebugprintk("tok_interrupt: ISRP_ODD = 0x%x ISRP_EVEN = 0x%x\n", */ + /*BMS status,status_even); */ + + if (status & ADAP_CHK_INT) { + int i; + void __iomem *check_reason; + __u8 check_reason_page = 0; + check_reason = map_address(ti, + ntohs(readw(ti->mmio+ ACA_OFFSET+ACA_RW + WWCR_EVEN)), + &check_reason_page); + SET_PAGE(check_reason_page); + + DPRINTK("Adapter check interrupt\n"); + DPRINTK("8 reason bytes follow: "); + for (i = 0; i < 8; i++, check_reason++) + printk("%02X ", (int) readb(check_reason)); + printk("\n"); + writeb(~ADAP_CHK_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD); + status = readb(ti->mmio + ACA_OFFSET + ACA_RW + ISRA_EVEN); + DPRINTK("ISRA_EVEN == 0x02%x\n",status); + ti->open_status = CLOSED; + ti->sap_status = CLOSED; + ti->open_mode = AUTOMATIC; + dev->flags &= ~IFF_RUNNING; + netif_stop_queue(dev); + ti->open_action = RESTART; + outb(0, dev->base_addr + ADAPTRESET); + ibmtr_reset_timer(&(ti->tr_timer), dev);/*BMS try to reopen*/ + spin_unlock(&(ti->lock)); + return IRQ_HANDLED; + } + if (readb(ti->mmio + ACA_OFFSET + ACA_RW + ISRP_EVEN) + & (TCR_INT | ERR_INT | ACCESS_INT)) { + DPRINTK("adapter error: ISRP_EVEN : %02x\n", + (int)readb(ti->mmio+ ACA_OFFSET + ACA_RW + ISRP_EVEN)); + writeb(~(TCR_INT | ERR_INT | ACCESS_INT), + ti->mmio + ACA_OFFSET + ACA_RESET + ISRP_EVEN); + status= readb(ti->mmio+ ACA_OFFSET + ACA_RW + ISRA_EVEN);/*BMS*/ + DPRINTK("ISRA_EVEN == 0x02%x\n",status);/*BMS*/ + writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN); +#ifdef ENABLE_PAGING + writeb(save_srpr, ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN); +#endif + spin_unlock(&(ti->lock)); + return IRQ_HANDLED; + } + if (status & SRB_RESP_INT) { /* SRB response */ + SET_PAGE(ti->srb_page); +#if TR_VERBOSE + DPRINTK("SRB resp: cmd=%02X rsp=%02X\n", + readb(ti->srb), readb(ti->srb + RETCODE_OFST)); +#endif + switch (readb(ti->srb)) { /* SRB command check */ + case XMIT_DIR_FRAME:{ + unsigned char xmit_ret_code; + xmit_ret_code = readb(ti->srb + RETCODE_OFST); + if (xmit_ret_code == 0xff) break; + DPRINTK("error on xmit_dir_frame request: %02X\n", + xmit_ret_code); + if (ti->current_skb) { + dev_kfree_skb_irq(ti->current_skb); + ti->current_skb = NULL; + } + /*dev->tbusy = 0;*/ + netif_wake_queue(dev); + if (ti->readlog_pending) + ibmtr_readlog(dev); + break; + } + case XMIT_UI_FRAME:{ + unsigned char xmit_ret_code; + + xmit_ret_code = readb(ti->srb + RETCODE_OFST); + if (xmit_ret_code == 0xff) break; + DPRINTK("error on xmit_ui_frame request: %02X\n", + xmit_ret_code); + if (ti->current_skb) { + dev_kfree_skb_irq(ti->current_skb); + ti->current_skb = NULL; + } + netif_wake_queue(dev); + if (ti->readlog_pending) + ibmtr_readlog(dev); + break; + } + case DIR_OPEN_ADAPTER: + dir_open_adapter(dev); + break; + case DLC_OPEN_SAP: + if (readb(ti->srb + RETCODE_OFST)) { + DPRINTK("open_sap failed: ret_code = %02X, " + "retrying\n", + (int) readb(ti->srb + RETCODE_OFST)); + ti->open_action = REOPEN; + ibmtr_reset_timer(&(ti->tr_timer), dev); + break; + } + ti->exsap_station_id = readw(ti->srb + STATION_ID_OFST); + ti->sap_status = OPEN;/* TR adapter is now available */ + if (ti->open_mode==MANUAL){ + wake_up(&ti->wait_for_reset); + break; + } + netif_wake_queue(dev); + dev->flags |= IFF_RUNNING;/*BMS 12/2000*/ + break; + case DIR_INTERRUPT: + case DIR_MOD_OPEN_PARAMS: + case DIR_SET_GRP_ADDR: + case DIR_SET_FUNC_ADDR: + case DLC_CLOSE_SAP: + if (readb(ti->srb + RETCODE_OFST)) + DPRINTK("error on %02X: %02X\n", + (int) readb(ti->srb + COMMAND_OFST), + (int) readb(ti->srb + RETCODE_OFST)); + break; + case DIR_READ_LOG: + if (readb(ti->srb + RETCODE_OFST)){ + DPRINTK("error on dir_read_log: %02X\n", + (int) readb(ti->srb + RETCODE_OFST)); + netif_wake_queue(dev); + break; + } +#if IBMTR_DEBUG_MESSAGES + +#define LINE_ERRORS_OFST 0 +#define INTERNAL_ERRORS_OFST 1 +#define BURST_ERRORS_OFST 2 +#define AC_ERRORS_OFST 3 +#define ABORT_DELIMITERS_OFST 4 +#define LOST_FRAMES_OFST 6 +#define RECV_CONGEST_COUNT_OFST 7 +#define FRAME_COPIED_ERRORS_OFST 8 +#define FREQUENCY_ERRORS_OFST 9 +#define TOKEN_ERRORS_OFST 10 + + DPRINTK("Line errors %02X, Internal errors %02X, " + "Burst errors %02X\n" "A/C errors %02X, " + "Abort delimiters %02X, Lost frames %02X\n" + "Receive congestion count %02X, " + "Frame copied errors %02X\nFrequency errors %02X, " + "Token errors %02X\n", + (int) readb(ti->srb + LINE_ERRORS_OFST), + (int) readb(ti->srb + INTERNAL_ERRORS_OFST), + (int) readb(ti->srb + BURST_ERRORS_OFST), + (int) readb(ti->srb + AC_ERRORS_OFST), + (int) readb(ti->srb + ABORT_DELIMITERS_OFST), + (int) readb(ti->srb + LOST_FRAMES_OFST), + (int) readb(ti->srb + RECV_CONGEST_COUNT_OFST), + (int) readb(ti->srb + FRAME_COPIED_ERRORS_OFST), + (int) readb(ti->srb + FREQUENCY_ERRORS_OFST), + (int) readb(ti->srb + TOKEN_ERRORS_OFST)); +#endif + netif_wake_queue(dev); + break; + default: + DPRINTK("Unknown command %02X encountered\n", + (int) readb(ti->srb)); + } /* end switch SRB command check */ + writeb(~SRB_RESP_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD); + } /* if SRB response */ + if (status & ASB_FREE_INT) { /* ASB response */ + SET_PAGE(ti->asb_page); +#if TR_VERBOSE + DPRINTK("ASB resp: cmd=%02X\n", readb(ti->asb)); +#endif + + switch (readb(ti->asb)) { /* ASB command check */ + case REC_DATA: + case XMIT_UI_FRAME: + case XMIT_DIR_FRAME: + break; + default: + DPRINTK("unknown command in asb %02X\n", + (int) readb(ti->asb)); + } /* switch ASB command check */ + if (readb(ti->asb + 2) != 0xff) /* checks ret_code */ + DPRINTK("ASB error %02X in cmd %02X\n", + (int) readb(ti->asb + 2), (int) readb(ti->asb)); + writeb(~ASB_FREE_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD); + } /* if ASB response */ + +#define STATUS_OFST 6 +#define NETW_STATUS_OFST 6 + + if (status & ARB_CMD_INT) { /* ARB response */ + SET_PAGE(ti->arb_page); +#if TR_VERBOSE + DPRINTK("ARB resp: cmd=%02X\n", readb(ti->arb)); +#endif + + switch (readb(ti->arb)) { /* ARB command check */ + case DLC_STATUS: + DPRINTK("DLC_STATUS new status: %02X on station %02X\n", + ntohs(readw(ti->arb + STATUS_OFST)), + ntohs(readw(ti->arb+ STATION_ID_OFST))); + break; + case REC_DATA: + tr_rx(dev); + break; + case RING_STAT_CHANGE:{ + unsigned short ring_status; + ring_status= ntohs(readw(ti->arb + NETW_STATUS_OFST)); + if (ibmtr_debug_trace & TRC_INIT) + DPRINTK("Ring Status Change...(0x%x)\n", + ring_status); + if(ring_status& (REMOVE_RECV|AUTO_REMOVAL|LOBE_FAULT)){ + netif_stop_queue(dev); + dev->flags &= ~IFF_RUNNING;/*not typical Linux*/ + DPRINTK("Remove received, or Auto-removal error" + ", or Lobe fault\n"); + DPRINTK("We'll try to reopen the closed adapter" + " after a %d second delay.\n", + TR_RETRY_INTERVAL/HZ); + /*I was confused: I saw the TR reopening but */ + /*forgot:with an RJ45 in an RJ45/ICS adapter */ + /*but adapter not in the ring, the TR will */ + /* open, and then soon close and come here. */ + ti->open_mode = AUTOMATIC; + ti->open_status = CLOSED; /*12/2000 BMS*/ + ti->open_action = REOPEN; + ibmtr_reset_timer(&(ti->tr_timer), dev); + } else if (ring_status & LOG_OVERFLOW) { + if(netif_queue_stopped(dev)) + ti->readlog_pending = 1; + else + ibmtr_readlog(dev); + } + break; + } + case XMIT_DATA_REQ: + tr_tx(dev); + break; + default: + DPRINTK("Unknown command %02X in arb\n", + (int) readb(ti->arb)); + break; + } /* switch ARB command check */ + writeb(~ARB_CMD_INT, ti->mmio+ ACA_OFFSET+ACA_RESET + ISRP_ODD); + writeb(ARB_FREE, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + } /* if ARB response */ + if (status & SSB_RESP_INT) { /* SSB response */ + unsigned char retcode; + SET_PAGE(ti->ssb_page); +#if TR_VERBOSE + DPRINTK("SSB resp: cmd=%02X rsp=%02X\n", + readb(ti->ssb), readb(ti->ssb + 2)); +#endif + + switch (readb(ti->ssb)) { /* SSB command check */ + case XMIT_DIR_FRAME: + case XMIT_UI_FRAME: + retcode = readb(ti->ssb + 2); + if (retcode && (retcode != 0x22))/* checks ret_code */ + DPRINTK("xmit ret_code: %02X xmit error code: " + "%02X\n", + (int)retcode, (int)readb(ti->ssb + 6)); + else + ti->tr_stats.tx_packets++; + break; + case XMIT_XID_CMD: + DPRINTK("xmit xid ret_code: %02X\n", + (int) readb(ti->ssb + 2)); + default: + DPRINTK("Unknown command %02X in ssb\n", + (int) readb(ti->ssb)); + } /* SSB command check */ + writeb(~SSB_RESP_INT, ti->mmio+ ACA_OFFSET+ACA_RESET+ ISRP_ODD); + writeb(SSB_FREE, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + } /* if SSB response */ +#ifdef ENABLE_PAGING + writeb(save_srpr, ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN); +#endif + writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN); + spin_unlock(&(ti->lock)); + return IRQ_HANDLED; +} /*tok_interrupt */ + +/*****************************************************************************/ + +#define INIT_STATUS_OFST 1 +#define INIT_STATUS_2_OFST 2 +#define ENCODED_ADDRESS_OFST 8 + +static void initial_tok_int(struct net_device *dev) +{ + + __u32 encoded_addr, hw_encoded_addr; + struct tok_info *ti; + unsigned char init_status; /*BMS 12/2000*/ + + ti = (struct tok_info *) dev->priv; + + ti->do_tok_int = NOT_FIRST; + + /* we assign the shared-ram address for ISA devices */ + writeb(ti->sram_base, ti->mmio + ACA_OFFSET + ACA_RW + RRR_EVEN); +#ifndef PCMCIA + ti->sram_virt = ioremap(((__u32)ti->sram_base << 12), ti->avail_shared_ram); +#endif + ti->init_srb = map_address(ti, + ntohs(readw(ti->mmio + ACA_OFFSET + WRBR_EVEN)), + &ti->init_srb_page); + if (ti->page_mask && ti->avail_shared_ram == 127) { + void __iomem *last_512; + __u8 last_512_page=0; + int i; + last_512 = map_address(ti, 0xfe00, &last_512_page); + /* initialize high section of ram (if necessary) */ + SET_PAGE(last_512_page); + for (i = 0; i < 512; i++) + writeb(0, last_512 + i); + } + SET_PAGE(ti->init_srb_page); + +#if TR_VERBOSE + { + int i; + + DPRINTK("ti->init_srb_page=0x%x\n", ti->init_srb_page); + DPRINTK("init_srb(%p):", ti->init_srb ); + for (i = 0; i < 20; i++) + printk("%02X ", (int) readb(ti->init_srb + i)); + printk("\n"); + } +#endif + + hw_encoded_addr = readw(ti->init_srb + ENCODED_ADDRESS_OFST); + encoded_addr = ntohs(hw_encoded_addr); + init_status= /*BMS 12/2000 check for shallow mode possibility (Turbo)*/ + readb(ti->init_srb+offsetof(struct srb_init_response,init_status)); + /*printk("Initial interrupt: init_status= 0x%02x\n",init_status);*/ + ti->ring_speed = init_status & 0x01 ? 16 : 4; + DPRINTK("Initial interrupt : %d Mbps, shared RAM base %08x.\n", + ti->ring_speed, (unsigned int)dev->mem_start); + ti->auto_speedsave=readb(ti->init_srb+INIT_STATUS_2_OFST)&4?TRUE:FALSE; + + if (ti->open_mode == MANUAL) wake_up(&ti->wait_for_reset); + else tok_open_adapter((unsigned long)dev); + +} /*initial_tok_int() */ + +/*****************************************************************************/ + +#define CMD_CORRELATE_OFST 1 +#define DHB_ADDRESS_OFST 6 + +#define FRAME_LENGTH_OFST 6 +#define HEADER_LENGTH_OFST 8 +#define RSAP_VALUE_OFST 9 + +static void tr_tx(struct net_device *dev) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + struct trh_hdr *trhdr = (struct trh_hdr *) ti->current_skb->data; + unsigned int hdr_len; + __u32 dhb=0,dhb_base; + void __iomem *dhbuf = NULL; + unsigned char xmit_command; + int i,dhb_len=0x4000,src_len,src_offset; + struct trllc *llc; + struct srb_xmit xsrb; + __u8 dhb_page = 0; + __u8 llc_ssap; + + SET_PAGE(ti->asb_page); + + if (readb(ti->asb+RETCODE_OFST) != 0xFF) DPRINTK("ASB not free !!!\n"); + + /* in providing the transmit interrupts, is telling us it is ready for + data and providing a shared memory address for us to stuff with data. + Here we compute the effective address where we will place data. + */ + SET_PAGE(ti->arb_page); + dhb=dhb_base=ntohs(readw(ti->arb + DHB_ADDRESS_OFST)); + if (ti->page_mask) { + dhb_page = (dhb_base >> 8) & ti->page_mask; + dhb=dhb_base & ~(ti->page_mask << 8); + } + dhbuf = ti->sram_virt + dhb; + + /* Figure out the size of the 802.5 header */ + if (!(trhdr->saddr[0] & 0x80)) /* RIF present? */ + hdr_len = sizeof(struct trh_hdr) - TR_MAXRIFLEN; + else + hdr_len = ((ntohs(trhdr->rcf) & TR_RCF_LEN_MASK) >> 8) + + sizeof(struct trh_hdr) - TR_MAXRIFLEN; + + llc = (struct trllc *) (ti->current_skb->data + hdr_len); + + llc_ssap = llc->ssap; + SET_PAGE(ti->srb_page); + memcpy_fromio(&xsrb, ti->srb, sizeof(xsrb)); + SET_PAGE(ti->asb_page); + xmit_command = xsrb.command; + + writeb(xmit_command, ti->asb + COMMAND_OFST); + writew(xsrb.station_id, ti->asb + STATION_ID_OFST); + writeb(llc_ssap, ti->asb + RSAP_VALUE_OFST); + writeb(xsrb.cmd_corr, ti->asb + CMD_CORRELATE_OFST); + writeb(0, ti->asb + RETCODE_OFST); + if ((xmit_command == XMIT_XID_CMD) || (xmit_command == XMIT_TEST_CMD)) { + writew(htons(0x11), ti->asb + FRAME_LENGTH_OFST); + writeb(0x0e, ti->asb + HEADER_LENGTH_OFST); + SET_PAGE(dhb_page); + writeb(AC, dhbuf); + writeb(LLC_FRAME, dhbuf + 1); + for (i = 0; i < TR_ALEN; i++) + writeb((int) 0x0FF, dhbuf + i + 2); + for (i = 0; i < TR_ALEN; i++) + writeb(0, dhbuf + i + TR_ALEN + 2); + writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + return; + } + /* + * the token ring packet is copied from sk_buff to the adapter + * buffer identified in the command data received with the interrupt. + */ + writeb(hdr_len, ti->asb + HEADER_LENGTH_OFST); + writew(htons(ti->current_skb->len), ti->asb + FRAME_LENGTH_OFST); + src_len=ti->current_skb->len; + src_offset=0; + dhb=dhb_base; + while(1) { + if (ti->page_mask) { + dhb_page=(dhb >> 8) & ti->page_mask; + dhb=dhb & ~(ti->page_mask << 8); + dhb_len=0x4000-dhb; /* remaining size of this page */ + } + dhbuf = ti->sram_virt + dhb; + SET_PAGE(dhb_page); + if (src_len > dhb_len) { + memcpy_toio(dhbuf,&ti->current_skb->data[src_offset], + dhb_len); + src_len -= dhb_len; + src_offset += dhb_len; + dhb_base+=dhb_len; + dhb=dhb_base; + continue; + } + memcpy_toio(dhbuf, &ti->current_skb->data[src_offset], src_len); + break; + } + writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + ti->tr_stats.tx_bytes += ti->current_skb->len; + dev_kfree_skb_irq(ti->current_skb); + ti->current_skb = NULL; + netif_wake_queue(dev); + if (ti->readlog_pending) + ibmtr_readlog(dev); +} /*tr_tx */ + +/*****************************************************************************/ + + +#define RECEIVE_BUFFER_OFST 6 +#define LAN_HDR_LENGTH_OFST 8 +#define DLC_HDR_LENGTH_OFST 9 + +#define DSAP_OFST 0 +#define SSAP_OFST 1 +#define LLC_OFST 2 +#define PROTID_OFST 3 +#define ETHERTYPE_OFST 6 + +static void tr_rx(struct net_device *dev) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + __u32 rbuffer; + void __iomem *rbuf, *rbufdata, *llc; + __u8 rbuffer_page = 0; + unsigned char *data; + unsigned int rbuffer_len, lan_hdr_len, hdr_len, ip_len, length; + unsigned char dlc_hdr_len; + struct sk_buff *skb; + unsigned int skb_size = 0; + int IPv4_p = 0; + unsigned int chksum = 0; + struct iphdr *iph; + struct arb_rec_req rarb; + + SET_PAGE(ti->arb_page); + memcpy_fromio(&rarb, ti->arb, sizeof(rarb)); + rbuffer = ntohs(rarb.rec_buf_addr) ; + rbuf = map_address(ti, rbuffer, &rbuffer_page); + + SET_PAGE(ti->asb_page); + + if (readb(ti->asb + RETCODE_OFST) !=0xFF) DPRINTK("ASB not free !!!\n"); + + writeb(REC_DATA, ti->asb + COMMAND_OFST); + writew(rarb.station_id, ti->asb + STATION_ID_OFST); + writew(rarb.rec_buf_addr, ti->asb + RECEIVE_BUFFER_OFST); + + lan_hdr_len = rarb.lan_hdr_len; + if (lan_hdr_len > sizeof(struct trh_hdr)) { + DPRINTK("Linux cannot handle greater than 18 bytes RIF\n"); + return; + } /*BMS I added this above just to be very safe */ + dlc_hdr_len = readb(ti->arb + DLC_HDR_LENGTH_OFST); + hdr_len = lan_hdr_len + sizeof(struct trllc) + sizeof(struct iphdr); + + SET_PAGE(rbuffer_page); + llc = rbuf + offsetof(struct rec_buf, data) + lan_hdr_len; + +#if TR_VERBOSE + DPRINTK("offsetof data: %02X lan_hdr_len: %02X\n", + (__u32) offsetof(struct rec_buf, data), (unsigned int) lan_hdr_len); + DPRINTK("llc: %08X rec_buf_addr: %04X dev->mem_start: %lX\n", + llc, ntohs(rarb.rec_buf_addr), dev->mem_start); + DPRINTK("dsap: %02X, ssap: %02X, llc: %02X, protid: %02X%02X%02X, " + "ethertype: %04X\n", + (int) readb(llc + DSAP_OFST), (int) readb(llc + SSAP_OFST), + (int) readb(llc + LLC_OFST), (int) readb(llc + PROTID_OFST), + (int) readb(llc+PROTID_OFST+1),(int)readb(llc+PROTID_OFST + 2), + (int) ntohs(readw(llc + ETHERTYPE_OFST))); +#endif + if (readb(llc + offsetof(struct trllc, llc)) != UI_CMD) { + SET_PAGE(ti->asb_page); + writeb(DATA_LOST, ti->asb + RETCODE_OFST); + ti->tr_stats.rx_dropped++; + writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + return; + } + length = ntohs(rarb.frame_len); + if (readb(llc + DSAP_OFST) == EXTENDED_SAP && + readb(llc + SSAP_OFST) == EXTENDED_SAP && + length >= hdr_len) IPv4_p = 1; +#if TR_VERBOSE +#define SADDR_OFST 8 +#define DADDR_OFST 2 + + if (!IPv4_p) { + + void __iomem *trhhdr = rbuf + offsetof(struct rec_buf, data); + + DPRINTK("Probably non-IP frame received.\n"); + DPRINTK("ssap: %02X dsap: %02X " + "saddr: %02X:%02X:%02X:%02X:%02X:%02X " + "daddr: %02X:%02X:%02X:%02X:%02X:%02X\n", + readb(llc + SSAP_OFST), readb(llc + DSAP_OFST), + readb(trhhdr+SADDR_OFST), readb(trhhdr+ SADDR_OFST+1), + readb(trhhdr+SADDR_OFST+2), readb(trhhdr+SADDR_OFST+3), + readb(trhhdr+SADDR_OFST+4), readb(trhhdr+SADDR_OFST+5), + readb(trhhdr+DADDR_OFST), readb(trhhdr+DADDR_OFST + 1), + readb(trhhdr+DADDR_OFST+2), readb(trhhdr+DADDR_OFST+3), + readb(trhhdr+DADDR_OFST+4), readb(trhhdr+DADDR_OFST+5)); + } +#endif + + /*BMS handle the case she comes in with few hops but leaves with many */ + skb_size=length-lan_hdr_len+sizeof(struct trh_hdr)+sizeof(struct trllc); + + if (!(skb = dev_alloc_skb(skb_size))) { + DPRINTK("out of memory. frame dropped.\n"); + ti->tr_stats.rx_dropped++; + SET_PAGE(ti->asb_page); + writeb(DATA_LOST, ti->asb + offsetof(struct asb_rec, ret_code)); + writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + return; + } + /*BMS again, if she comes in with few but leaves with many */ + skb_reserve(skb, sizeof(struct trh_hdr) - lan_hdr_len); + skb_put(skb, length); + skb->dev = dev; + data = skb->data; + rbuffer_len = ntohs(readw(rbuf + offsetof(struct rec_buf, buf_len))); + rbufdata = rbuf + offsetof(struct rec_buf, data); + + if (IPv4_p) { + /* Copy the headers without checksumming */ + memcpy_fromio(data, rbufdata, hdr_len); + + /* Watch for padded packets and bogons */ + iph= (struct iphdr *)(data+ lan_hdr_len + sizeof(struct trllc)); + ip_len = ntohs(iph->tot_len) - sizeof(struct iphdr); + length -= hdr_len; + if ((ip_len <= length) && (ip_len > 7)) + length = ip_len; + data += hdr_len; + rbuffer_len -= hdr_len; + rbufdata += hdr_len; + } + /* Copy the payload... */ +#define BUFFER_POINTER_OFST 2 +#define BUFFER_LENGTH_OFST 6 + for (;;) { + if (ibmtr_debug_trace&TRC_INITV && length < rbuffer_len) + DPRINTK("CURIOUS, length=%d < rbuffer_len=%d\n", + length,rbuffer_len); + if (IPv4_p) + chksum=csum_partial_copy_nocheck((void*)rbufdata, + data,length<rbuffer_len?length:rbuffer_len,chksum); + else + memcpy_fromio(data, rbufdata, rbuffer_len); + rbuffer = ntohs(readw(rbuf+BUFFER_POINTER_OFST)) ; + if (!rbuffer) + break; + rbuffer -= 2; + length -= rbuffer_len; + data += rbuffer_len; + rbuf = map_address(ti, rbuffer, &rbuffer_page); + SET_PAGE(rbuffer_page); + rbuffer_len = ntohs(readw(rbuf + BUFFER_LENGTH_OFST)); + rbufdata = rbuf + offsetof(struct rec_buf, data); + } + + SET_PAGE(ti->asb_page); + writeb(0, ti->asb + offsetof(struct asb_rec, ret_code)); + + writeb(RESP_IN_ASB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + + ti->tr_stats.rx_bytes += skb->len; + ti->tr_stats.rx_packets++; + + skb->protocol = tr_type_trans(skb, dev); + if (IPv4_p) { + skb->csum = chksum; + skb->ip_summed = 1; + } + netif_rx(skb); + dev->last_rx = jiffies; +} /*tr_rx */ + +/*****************************************************************************/ + +void ibmtr_reset_timer(struct timer_list *tmr, struct net_device *dev) +{ + tmr->expires = jiffies + TR_RETRY_INTERVAL; + tmr->data = (unsigned long) dev; + tmr->function = tok_rerun; + init_timer(tmr); + add_timer(tmr); +} + +/*****************************************************************************/ + +void tok_rerun(unsigned long dev_addr){ + + struct net_device *dev = (struct net_device *)dev_addr; + struct tok_info *ti = (struct tok_info *) dev->priv; + + if ( ti->open_action == RESTART){ + ti->do_tok_int = FIRST_INT; + outb(0, dev->base_addr + ADAPTRESETREL); +#ifdef ENABLE_PAGING + if (ti->page_mask) + writeb(SRPR_ENABLE_PAGING, + ti->mmio + ACA_OFFSET + ACA_RW + SRPR_EVEN); +#endif + + writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN); + } else + tok_open_adapter(dev_addr); +} + +/*****************************************************************************/ + +void ibmtr_readlog(struct net_device *dev) +{ + struct tok_info *ti; + + ti = (struct tok_info *) dev->priv; + + ti->readlog_pending = 0; + SET_PAGE(ti->srb_page); + writeb(DIR_READ_LOG, ti->srb); + writeb(INT_ENABLE, ti->mmio + ACA_OFFSET + ACA_SET + ISRP_EVEN); + writeb(CMD_IN_SRB, ti->mmio + ACA_OFFSET + ACA_SET + ISRA_ODD); + + netif_stop_queue(dev); + +} + +/*****************************************************************************/ + +/* tok_get_stats(): Basically a scaffold routine which will return + the address of the tr_statistics structure associated with + this device -- the tr.... structure is an ethnet look-alike + so at least for this iteration may suffice. */ + +static struct net_device_stats *tok_get_stats(struct net_device *dev) +{ + + struct tok_info *toki; + toki = (struct tok_info *) dev->priv; + return (struct net_device_stats *) &toki->tr_stats; +} + +/*****************************************************************************/ + +int ibmtr_change_mtu(struct net_device *dev, int mtu) +{ + struct tok_info *ti = (struct tok_info *) dev->priv; + + if (ti->ring_speed == 16 && mtu > ti->maxmtu16) + return -EINVAL; + if (ti->ring_speed == 4 && mtu > ti->maxmtu4) + return -EINVAL; + dev->mtu = mtu; + return 0; +} + +/*****************************************************************************/ +#ifdef MODULE + +/* 3COM 3C619C supports 8 interrupts, 32 I/O ports */ +static struct net_device *dev_ibmtr[IBMTR_MAX_ADAPTERS]; +static int io[IBMTR_MAX_ADAPTERS] = { 0xa20, 0xa24 }; +static int irq[IBMTR_MAX_ADAPTERS]; +static int mem[IBMTR_MAX_ADAPTERS]; + +MODULE_LICENSE("GPL"); + +module_param_array(io, int, NULL, 0); +module_param_array(irq, int, NULL, 0); +module_param_array(mem, int, NULL, 0); + +static int __init ibmtr_init(void) +{ + int i; + int count=0; + + find_turbo_adapters(io); + + for (i = 0; io[i] && (i < IBMTR_MAX_ADAPTERS); i++) { + struct net_device *dev; + irq[i] = 0; + mem[i] = 0; + dev = alloc_trdev(sizeof(struct tok_info)); + if (dev == NULL) { + if (i == 0) + return -ENOMEM; + break; + } + dev->base_addr = io[i]; + dev->irq = irq[i]; + dev->mem_start = mem[i]; + + if (ibmtr_probe_card(dev)) { + free_netdev(dev); + continue; + } + dev_ibmtr[i] = dev; + count++; + } + if (count) return 0; + printk("ibmtr: register_netdev() returned non-zero.\n"); + return -EIO; +} +module_init(ibmtr_init); + +static void __exit ibmtr_cleanup(void) +{ + int i; + + for (i = 0; i < IBMTR_MAX_ADAPTERS; i++){ + if (!dev_ibmtr[i]) + continue; + unregister_netdev(dev_ibmtr[i]); + ibmtr_cleanup_card(dev_ibmtr[i]); + free_netdev(dev_ibmtr[i]); + } +} +module_exit(ibmtr_cleanup); +#endif |