/* * ti_pcilynx.c - Texas Instruments PCILynx driver * Copyright (C) 1999,2000 Andreas Bombe , * Stephan Linz * * 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. */ /* * Lynx DMA usage: * * 0 is used for Lynx local bus transfers * 1 is async/selfid receive * 2 is iso receive * 3 is async transmit */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "ieee1394.h" #include "ieee1394_types.h" #include "hosts.h" #include "ieee1394_core.h" #include "pcilynx.h" #if MAX_PCILYNX_CARDS > PCILYNX_MINOR_ROM_START #error Max number of cards is bigger than PCILYNX_MINOR_ROM_START - this does not work. #endif /* print general (card independent) information */ #define PRINT_G(level, fmt, args...) printk(level "pcilynx: " fmt "\n" , ## args) /* print card specific information */ #define PRINT(level, card, fmt, args...) printk(level "pcilynx%d: " fmt "\n" , card , ## args) #ifdef CONFIG_IEEE1394_VERBOSEDEBUG #define PRINT_GD(level, fmt, args...) printk(level "pcilynx: " fmt "\n" , ## args) #define PRINTD(level, card, fmt, args...) printk(level "pcilynx%d: " fmt "\n" , card , ## args) #else #define PRINT_GD(level, fmt, args...) #define PRINTD(level, card, fmt, args...) #endif static struct ti_lynx cards[MAX_PCILYNX_CARDS]; static int num_of_cards = 0; /* * PCL handling functions. */ static pcl_t alloc_pcl(struct ti_lynx *lynx) { u8 m; int i, j; spin_lock(&lynx->lock); /* FIXME - use ffz() to make this readable */ for (i = 0; i < LOCALRAM_SIZE; i++) { m = lynx->pcl_bmap[i]; for (j = 0; j < 8; j++) { if (m & 1<pcl_bmap[i] = m; spin_unlock(&lynx->lock); return 8 * i + j; } } spin_unlock(&lynx->lock); return -1; } static void free_pcl(struct ti_lynx *lynx, pcl_t pclid) { int off, bit; off = pclid / 8; bit = pclid % 8; if (pclid < 0) { return; } spin_lock(&lynx->lock); if (lynx->pcl_bmap[off] & 1<pcl_bmap[off] &= ~(1<id, "attempted to free unallocated PCL %d", pclid); } spin_unlock(&lynx->lock); } /* functions useful for debugging */ static void pretty_print_pcl(const struct ti_pcl *pcl) { int i; printk("PCL next %08x, userdata %08x, status %08x, remtrans %08x, nextbuf %08x\n", pcl->next, pcl->user_data, pcl->pcl_status, pcl->remaining_transfer_count, pcl->next_data_buffer); printk("PCL"); for (i=0; i<13; i++) { printk(" c%x:%08x d%x:%08x", i, pcl->buffer[i].control, i, pcl->buffer[i].pointer); if (!(i & 0x3) && (i != 12)) printk("\nPCL"); } printk("\n"); } static void print_pcl(const struct ti_lynx *lynx, pcl_t pclid) { struct ti_pcl pcl; get_pcl(lynx, pclid, &pcl); pretty_print_pcl(&pcl); } static int add_card(struct pci_dev *dev); static void remove_card(struct ti_lynx *lynx); static int init_driver(void); /*********************************** * IEEE-1394 functionality section * ***********************************/ static int get_phy_reg(struct ti_lynx *lynx, int addr) { int retval; int i = 0; unsigned long flags; if (addr > 15) { PRINT(KERN_ERR, lynx->id, __FUNCTION__ ": PHY register address %d out of range", addr); return -1; } spin_lock_irqsave(&lynx->phy_reg_lock, flags); do { reg_write(lynx, LINK_PHY, LINK_PHY_READ | LINK_PHY_ADDR(addr)); retval = reg_read(lynx, LINK_PHY); if (i > 10000) { PRINT(KERN_ERR, lynx->id, __FUNCTION__ ": runaway loop, aborting"); retval = -1; break; } i++; } while ((retval & 0xf00) != LINK_PHY_RADDR(addr)); reg_write(lynx, LINK_INT_STATUS, LINK_INT_PHY_REG_RCVD); spin_unlock_irqrestore(&lynx->phy_reg_lock, flags); if (retval != -1) { return retval & 0xff; } else { return -1; } } static int set_phy_reg(struct ti_lynx *lynx, int addr, int val) { unsigned long flags; if (addr > 15) { PRINT(KERN_ERR, lynx->id, __FUNCTION__ ": PHY register address %d out of range", addr); return -1; } if (val > 0xff) { PRINT(KERN_ERR, lynx->id, __FUNCTION__ ": PHY register value %d out of range", val); return -1; } spin_lock_irqsave(&lynx->phy_reg_lock, flags); reg_write(lynx, LINK_PHY, LINK_PHY_WRITE | LINK_PHY_ADDR(addr) | LINK_PHY_WDATA(val)); spin_unlock_irqrestore(&lynx->phy_reg_lock, flags); return 0; } static int sel_phy_reg_page(struct ti_lynx *lynx, int page) { int reg; if (page > 7) { PRINT(KERN_ERR, lynx->id, __FUNCTION__ ": PHY page %d out of range", page); return -1; } reg = get_phy_reg(lynx, 7); if (reg != -1) { reg &= 0x1f; reg |= (page << 5); set_phy_reg(lynx, 7, reg); return 0; } else { return -1; } } #if 0 /* not needed at this time */ static int sel_phy_reg_port(struct ti_lynx *lynx, int port) { int reg; if (port > 15) { PRINT(KERN_ERR, lynx->id, __FUNCTION__ ": PHY port %d out of range", port); return -1; } reg = get_phy_reg(lynx, 7); if (reg != -1) { reg &= 0xf0; reg |= port; set_phy_reg(lynx, 7, reg); return 0; } else { return -1; } } #endif static u32 get_phy_vendorid(struct ti_lynx *lynx) { u32 pvid = 0; sel_phy_reg_page(lynx, 1); pvid |= (get_phy_reg(lynx, 10) << 16); pvid |= (get_phy_reg(lynx, 11) << 8); pvid |= get_phy_reg(lynx, 12); PRINT(KERN_INFO, lynx->id, "PHY vendor id 0x%06x", pvid); return pvid; } static u32 get_phy_productid(struct ti_lynx *lynx) { u32 id = 0; sel_phy_reg_page(lynx, 1); id |= (get_phy_reg(lynx, 13) << 16); id |= (get_phy_reg(lynx, 14) << 8); id |= get_phy_reg(lynx, 15); PRINT(KERN_INFO, lynx->id, "PHY product id 0x%06x", id); return id; } static quadlet_t generate_own_selfid(struct ti_lynx *lynx, struct hpsb_host *host) { quadlet_t lsid; char phyreg[7]; int i; for (i = 0; i < 7; i++) { phyreg[i] = get_phy_reg(lynx, i); } /* FIXME? We assume a TSB21LV03A phy here. This code doesn't support more than 3 ports on the PHY anyway. */ lsid = 0x80400000 | ((phyreg[0] & 0xfc) << 22); lsid |= (phyreg[1] & 0x3f) << 16; /* gap count */ lsid |= (phyreg[2] & 0xc0) << 8; /* max speed */ /* lsid |= (phyreg[6] & 0x01) << 11; *//* contender (phy dependent) */ lsid |= 1 << 11; /* set contender (hack) */ lsid |= (phyreg[6] & 0x10) >> 3; /* initiated reset */ //for (i = 0; i < (phyreg[2] & 0xf); i++) { /* ports */ for (i = 0; i < (phyreg[2] & 0x1f); i++) { /* ports */ if (phyreg[3 + i] & 0x4) { lsid |= (((phyreg[3 + i] & 0x8) | 0x10) >> 3) << (6 - i*2); } else { lsid |= 1 << (6 - i*2); } } cpu_to_be32s(&lsid); PRINT(KERN_DEBUG, lynx->id, "generated own selfid 0x%x", lsid); return lsid; } static void handle_selfid(struct ti_lynx *lynx, struct hpsb_host *host, size_t size) { quadlet_t *q = lynx->rcv_page; int phyid, isroot; quadlet_t lsid = 0; int i; i = (size > 16 ? 16 : size) / 4 - 1; while (i >= 0) { cpu_to_be32s(&q[i]); i--; } if (!lynx->phyic.reg_1394a) { lsid = generate_own_selfid(lynx, host); } phyid = get_phy_reg(lynx, 0); isroot = (phyid & 2) != 0; phyid >>= 2; PRINT(KERN_INFO, lynx->id, "SelfID process finished (phyid %d, %s)", phyid, (isroot ? "root" : "not root")); reg_write(lynx, LINK_ID, (0xffc0 | phyid) << 16); if (!lynx->phyic.reg_1394a && !size) { hpsb_selfid_received(host, lsid); } while (size > 0) { struct selfid *sid = (struct selfid *)q; if (!lynx->phyic.reg_1394a && !sid->extended && (sid->phy_id == (phyid + 1))) { hpsb_selfid_received(host, lsid); } if (q[0] == ~q[1]) { PRINT(KERN_DEBUG, lynx->id, "selfid packet 0x%x rcvd", q[0]); hpsb_selfid_received(host, q[0]); } else { PRINT(KERN_INFO, lynx->id, "inconsistent selfid 0x%x/0x%x", q[0], q[1]); } q += 2; size -= 8; } if (!lynx->phyic.reg_1394a && isroot && phyid != 0) { hpsb_selfid_received(host, lsid); } hpsb_selfid_complete(host, phyid, isroot); } /* This must be called with the async_queue_lock held. */ static void send_next_async(struct ti_lynx *lynx) { struct ti_pcl pcl; struct hpsb_packet *packet = lynx->async_queue; pcl.next = PCL_NEXT_INVALID; pcl.async_error_next = PCL_NEXT_INVALID; #ifdef __BIG_ENDIAN pcl.buffer[0].control = PCL_CMD_XMT | packet->speed_code << 14 | packet->header_size; #else pcl.buffer[0].control = PCL_CMD_XMT | packet->speed_code << 14 | packet->header_size | PCL_BIGENDIAN; #endif pcl.buffer[0].pointer = virt_to_bus(packet->header); pcl.buffer[1].control = PCL_LAST_BUFF | packet->data_size; pcl.buffer[1].pointer = virt_to_bus(packet->data); if (!packet->data_be) { pcl.buffer[1].control |= PCL_BIGENDIAN; } put_pcl(lynx, lynx->async_pcl, &pcl); run_pcl(lynx, lynx->async_pcl_start, 3); } static int lynx_detect(struct hpsb_host_template *tmpl) { struct hpsb_host *host; int i; init_driver(); for (i = 0; i < num_of_cards; i++) { host = hpsb_get_host(tmpl, 0); if (host == NULL) { /* simply don't init more after out of mem */ return i; } host->hostdata = &cards[i]; cards[i].host = host; } return num_of_cards; } static int lynx_initialize(struct hpsb_host *host) { struct ti_lynx *lynx = host->hostdata; struct ti_pcl pcl; int i; u32 *pcli; lynx->async_queue = NULL; spin_lock_init(&lynx->async_queue_lock); spin_lock_init(&lynx->phy_reg_lock); pcl.next = pcl_bus(lynx, lynx->rcv_pcl); put_pcl(lynx, lynx->rcv_pcl_start, &pcl); pcl.next = PCL_NEXT_INVALID; pcl.async_error_next = PCL_NEXT_INVALID; #ifdef __BIG_ENDIAN pcl.buffer[0].control = PCL_CMD_RCV | 16; pcl.buffer[1].control = PCL_LAST_BUFF | 4080; #else pcl.buffer[0].control = PCL_CMD_RCV | PCL_BIGENDIAN | 16; pcl.buffer[1].control = PCL_LAST_BUFF | 4080; #endif pcl.buffer[0].pointer = virt_to_bus(lynx->rcv_page); pcl.buffer[1].pointer = virt_to_bus(lynx->rcv_page) + 16; put_pcl(lynx, lynx->rcv_pcl, &pcl); pcl.next = pcl_bus(lynx, lynx->async_pcl); pcl.async_error_next = pcl_bus(lynx, lynx->async_pcl); put_pcl(lynx, lynx->async_pcl_start, &pcl); pcl.next = PCL_NEXT_INVALID; pcl.async_error_next = PCL_NEXT_INVALID; pcl.buffer[0].control = PCL_CMD_RCV | 4; #ifndef __BIG_ENDIAN pcl.buffer[0].control |= PCL_BIGENDIAN; #endif pcl.buffer[1].control = PCL_LAST_BUFF | 2044; for (i = 0; i < NUM_ISORCV_PCL; i++) { int page = i / ISORCV_PER_PAGE; int sec = i % ISORCV_PER_PAGE; pcl.buffer[0].pointer = virt_to_bus(lynx->iso_rcv.page[page]) + sec * MAX_ISORCV_SIZE; pcl.buffer[1].pointer = pcl.buffer[0].pointer + 4; put_pcl(lynx, lynx->iso_rcv.pcl[i], &pcl); } pcli = (u32 *)&pcl; for (i = 0; i < NUM_ISORCV_PCL; i++) { pcli[i] = pcl_bus(lynx, lynx->iso_rcv.pcl[i]); } put_pcl(lynx, lynx->iso_rcv.pcl_start, &pcl); /* 85 bytes for each FIFO - FIXME - optimize or make configurable */ reg_write(lynx, FIFO_SIZES, 0x00555555); /* 20 byte threshold before triggering PCI transfer */ reg_write(lynx, DMA_GLOBAL_REGISTER, 0x2<<24); /* 69 byte threshold on both send FIFOs before transmitting */ reg_write(lynx, FIFO_XMIT_THRESHOLD, 0x4545); reg_set_bits(lynx, PCI_INT_ENABLE, PCI_INT_1394); reg_write(lynx, LINK_INT_ENABLE, LINK_INT_PHY_TIMEOUT | LINK_INT_PHY_REG_RCVD | LINK_INT_PHY_BUSRESET | LINK_INT_ISO_STUCK | LINK_INT_ASYNC_STUCK | LINK_INT_SENT_REJECT | LINK_INT_TX_INVALID_TC | LINK_INT_GRF_OVERFLOW | LINK_INT_ITF_UNDERFLOW | LINK_INT_ATF_UNDERFLOW); reg_write(lynx, DMA1_WORD0_CMP_VALUE, 0); reg_write(lynx, DMA1_WORD0_CMP_ENABLE, 0xa<<4); reg_write(lynx, DMA1_WORD1_CMP_VALUE, 0); reg_write(lynx, DMA1_WORD1_CMP_ENABLE, DMA_WORD1_CMP_MATCH_NODE_BCAST | DMA_WORD1_CMP_MATCH_BROADCAST | DMA_WORD1_CMP_MATCH_LOCAL | DMA_WORD1_CMP_MATCH_BUS_BCAST | DMA_WORD1_CMP_ENABLE_SELF_ID | DMA_WORD1_CMP_ENABLE_MASTER); run_pcl(lynx, lynx->rcv_pcl_start, 1); reg_write(lynx, DMA_WORD0_CMP_VALUE(CHANNEL_ISO_RCV), 0); reg_write(lynx, DMA_WORD0_CMP_ENABLE(CHANNEL_ISO_RCV), 0x9<<4); reg_write(lynx, DMA_WORD1_CMP_VALUE(CHANNEL_ISO_RCV), 0); reg_write(lynx, DMA_WORD1_CMP_ENABLE(CHANNEL_ISO_RCV), 0); run_sub_pcl(lynx, lynx->iso_rcv.pcl_start, 0, CHANNEL_ISO_RCV); reg_write(lynx, LINK_CONTROL, LINK_CONTROL_RCV_CMP_VALID | LINK_CONTROL_TX_ISO_EN | LINK_CONTROL_RX_ISO_EN | LINK_CONTROL_TX_ASYNC_EN | LINK_CONTROL_RX_ASYNC_EN | LINK_CONTROL_RESET_TX | LINK_CONTROL_RESET_RX | LINK_CONTROL_CYCSOURCE | LINK_CONTROL_CYCTIMEREN); /* attempt to enable contender bit -FIXME- would this work elsewhere? */ reg_set_bits(lynx, GPIO_CTRL_A, 0x1); reg_write(lynx, GPIO_DATA_BASE + 0x3c, 0x1); return 1; } static void lynx_release(struct hpsb_host *host) { struct ti_lynx *lynx; if (host != NULL) { lynx = host->hostdata; remove_card(lynx); } else { unregister_chrdev(PCILYNX_MAJOR, PCILYNX_DRIVER_NAME); } } /* * FIXME - does not support iso/raw transmits yet and will choke on them. */ static int lynx_transmit(struct hpsb_host *host, struct hpsb_packet *packet) { struct ti_lynx *lynx = host->hostdata; struct hpsb_packet *p; unsigned long flags; if (packet->data_size >= 4096) { PRINT(KERN_ERR, lynx->id, "transmit packet data too big (%d)", packet->data_size); return 0; } packet->xnext = NULL; if (packet->tcode == TCODE_WRITEQ || packet->tcode == TCODE_READQ_RESPONSE) { cpu_to_be32s(&packet->header[3]); } spin_lock_irqsave(&lynx->async_queue_lock, flags); if (lynx->async_queue == NULL) { lynx->async_queue = packet; send_next_async(lynx); } else { p = lynx->async_queue; while (p->xnext != NULL) { p = p->xnext; } p->xnext = packet; } spin_unlock_irqrestore(&lynx->async_queue_lock, flags); return 1; } static int lynx_devctl(struct hpsb_host *host, enum devctl_cmd cmd, int arg) { struct ti_lynx *lynx = host->hostdata; int retval = 0; struct hpsb_packet *packet, *lastpacket; unsigned long flags; switch (cmd) { case RESET_BUS: if (arg) { arg = 3 << 6; } else { arg = 1 << 6; } PRINT(KERN_INFO, lynx->id, "resetting bus on request%s", (host->attempt_root ? " and attempting to become root" : "")); spin_lock_irqsave(&lynx->phy_reg_lock, flags); reg_write(lynx, LINK_PHY, LINK_PHY_WRITE | LINK_PHY_ADDR(1) | LINK_PHY_WDATA(arg)); spin_unlock_irqrestore(&lynx->phy_reg_lock, flags); break; case GET_CYCLE_COUNTER: retval = reg_read(lynx, CYCLE_TIMER); break; case SET_CYCLE_COUNTER: reg_write(lynx, CYCLE_TIMER, arg); break; case SET_BUS_ID: reg_write(lynx, LINK_ID, (arg << 22) | (reg_read(lynx, LINK_ID) & 0x003f0000)); break; case ACT_CYCLE_MASTER: if (arg) { reg_set_bits(lynx, LINK_CONTROL, LINK_CONTROL_CYCMASTER); } else { reg_clear_bits(lynx, LINK_CONTROL, LINK_CONTROL_CYCMASTER); } break; case CANCEL_REQUESTS: spin_lock_irqsave(&lynx->async_queue_lock, flags); reg_write(lynx, DMA3_CHAN_CTRL, 0); packet = lynx->async_queue; lynx->async_queue = NULL; spin_unlock_irqrestore(&lynx->async_queue_lock, flags); while (packet != NULL) { lastpacket = packet; packet = packet->xnext; hpsb_packet_sent(host, lastpacket, ACKX_ABORTED); } break; case MODIFY_USAGE: if (arg) { MOD_INC_USE_COUNT; } else { MOD_DEC_USE_COUNT; } break; case ISO_LISTEN_CHANNEL: spin_lock_irqsave(&lynx->iso_rcv.lock, flags); if (lynx->iso_rcv.chan_count++ == 0) { reg_write(lynx, DMA_WORD1_CMP_ENABLE(CHANNEL_ISO_RCV), DMA_WORD1_CMP_ENABLE_MASTER); } spin_unlock_irqrestore(&lynx->iso_rcv.lock, flags); break; case ISO_UNLISTEN_CHANNEL: spin_lock_irqsave(&lynx->iso_rcv.lock, flags); if (--lynx->iso_rcv.chan_count == 0) { reg_write(lynx, DMA_WORD1_CMP_ENABLE(CHANNEL_ISO_RCV), 0); } spin_unlock_irqrestore(&lynx->iso_rcv.lock, flags); break; default: PRINT(KERN_ERR, lynx->id, "unknown devctl command %d", cmd); retval = -1; } return retval; } /*************************************** * IEEE-1394 functionality section END * ***************************************/ /* VFS functions for local bus / aux device access. Access to those * is implemented as a character device instead of block devices * because buffers are not wanted for this. Therefore llseek (from * VFS) can be used for these char devices with obvious effects. */ static int mem_open(struct inode*, struct file*); static int mem_release(struct inode*, struct file*); static unsigned int aux_poll(struct file*, struct poll_table_struct*); static ssize_t mem_read (struct file*, char*, size_t, loff_t*); static ssize_t mem_write(struct file*, const char*, size_t, loff_t*); static struct file_operations aux_ops = { /* FIXME: should have custom llseek with bounds checking */ read: mem_read, write: mem_write, poll: aux_poll, open: mem_open, release: mem_release, }; static void aux_setup_pcls(struct ti_lynx *lynx) { struct ti_pcl pcl; unsigned long membufbus = virt_to_bus(lynx->mem_dma_buffer); int i; /* This pcl is used to start any aux transfers, the pointer to next points to itself to avoid a dummy pcl (the PCL engine only executes the next pcl on startup. The real chain is done by branch */ pcl.next = pcl_bus(lynx, lynx->mem_pcl.start); pcl.buffer[0].control = PCL_CMD_BRANCH | PCL_COND_DMARDY_SET; pcl.buffer[0].pointer = pcl_bus(lynx, lynx->mem_pcl.max); pcl.buffer[1].control = PCL_CMD_BRANCH | PCL_COND_DMARDY_CLEAR; pcl.buffer[1].pointer = pcl_bus(lynx, lynx->mem_pcl.cmd); put_pcl(lynx, lynx->mem_pcl.start, &pcl); /* let maxpcl transfer exactly 32kB */ pcl.next = PCL_NEXT_INVALID; for (i=0; i<8; i++) { pcl.buffer[i].control = 4000; pcl.buffer[i].pointer = membufbus + i * 4000; } pcl.buffer[0].control |= PCL_CMD_LBUS_TO_PCI /*| PCL_GEN_INTR*/; pcl.buffer[8].control = 768 | PCL_LAST_BUFF; pcl.buffer[8].pointer = membufbus + 8 * 4000; put_pcl(lynx, lynx->mem_pcl.max, &pcl); /* magic stuff - self and modpcl modifying pcl */ pcl.next = pcl_bus(lynx, lynx->mem_pcl.mod); pcl.user_data = 4000; pcl.buffer[0].control = PCL_CMD_LOAD; pcl.buffer[0].pointer = pcl_bus(lynx, lynx->mem_pcl.cmd) + pcloffs(user_data); pcl.buffer[1].control = PCL_CMD_STOREQ; pcl.buffer[1].pointer = pcl_bus(lynx, lynx->mem_pcl.mod) + pcloffs(buffer[1].control); pcl.buffer[2].control = PCL_CMD_LOAD; pcl.buffer[2].pointer = membufbus; pcl.buffer[3].control = PCL_CMD_STOREQ; pcl.buffer[3].pointer = pcl_bus(lynx, lynx->mem_pcl.cmd) + pcloffs(buffer[1].pointer); pcl.buffer[4].control = PCL_CMD_STOREQ; pcl.buffer[4].pointer = pcl_bus(lynx, lynx->mem_pcl.cmd) + pcloffs(buffer[6].pointer); pcl.buffer[5].control = PCL_CMD_LOAD; pcl.buffer[5].pointer = membufbus + 4; pcl.buffer[6].control = PCL_CMD_STOREQ | PCL_LAST_CMD; put_pcl(lynx, lynx->mem_pcl.cmd, &pcl); /* modified by cmdpcl when actual transfer occurs */ pcl.next = PCL_NEXT_INVALID; pcl.buffer[0].control = PCL_CMD_LBUS_TO_PCI; /* null transfer */ for (i=1; i<13; i++) { pcl.buffer[i].control = 4000; pcl.buffer[i].pointer = membufbus + (i-1) * 4000; } put_pcl(lynx, lynx->mem_pcl.mod, &pcl); } static int mem_open(struct inode *inode, struct file *file) { int cid = MINOR(inode->i_rdev); enum { rom, aux, ram } type; struct memdata *md; MOD_INC_USE_COUNT; if (cid < PCILYNX_MINOR_AUX_START) { /* just for completeness */ MOD_DEC_USE_COUNT; return -ENXIO; } else if (cid < PCILYNX_MINOR_ROM_START) { cid -= PCILYNX_MINOR_AUX_START; if (cid >= num_of_cards || !cards[cid].aux_port) { MOD_DEC_USE_COUNT; return -ENXIO; } type = aux; } else if (cid < PCILYNX_MINOR_RAM_START) { cid -= PCILYNX_MINOR_ROM_START; if (cid >= num_of_cards || !cards[cid].local_rom) { MOD_DEC_USE_COUNT; return -ENXIO; } type = rom; } else { /* WARNING: Know what you are doing when opening RAM. * It is currently used inside the driver! */ cid -= PCILYNX_MINOR_RAM_START; if (cid >= num_of_cards || !cards[cid].local_ram) { MOD_DEC_USE_COUNT; return -ENXIO; } type = ram; } md = (struct memdata *)vmalloc(sizeof(struct memdata)); if (md == NULL) { MOD_DEC_USE_COUNT; return -ENOMEM; } md->lynx = &cards[cid]; md->cid = cid; switch (type) { case rom: md->type = rom; break; case ram: md->type = ram; break; case aux: md->aux_intr_last_seen = atomic_read(&cards[cid].aux_intr_seen); md->type = aux; break; } file->private_data = md; return 0; } static int mem_release(struct inode *inode, struct file *file) { struct memdata *md = (struct memdata *)file->private_data; vfree(md); MOD_DEC_USE_COUNT; return 0; } static unsigned int aux_poll(struct file *file, poll_table *pt) { struct memdata *md = (struct memdata *)file->private_data; int cid = md->cid; unsigned int mask; int intr_seen; /* reading and writing is always allowed */ mask = POLLIN | POLLRDNORM | POLLOUT | POLLWRNORM; if (md->type == aux) { poll_wait(file, &cards[cid].aux_intr_wait, pt); intr_seen = atomic_read(&cards[cid].aux_intr_seen); if (md->aux_intr_last_seen != intr_seen) { mask |= POLLPRI; /* md->aux_intr_last_seen = intr_seen; */ md->aux_intr_last_seen++; /* don't miss interrupts */ /* FIXME - make ioctl for configuring this */ } } return mask; } /* * do not DMA if count is too small because this will have a serious impact * on performance - the value 2400 was found by experiment and may not work * everywhere as good as here - use mem_mindma option for modules to change */ short mem_mindma = 2400; MODULE_PARM(mem_mindma, "h"); static ssize_t mem_read(struct file *file, char *buffer, size_t count, loff_t *offset) { struct memdata *md = (struct memdata *)file->private_data; size_t bcount; size_t alignfix; int off = (int)*offset; /* avoid useless 64bit-arithmetic */ void *membase; DECLARE_WAITQUEUE(wait, current); if ((off + count) > PCILYNX_MAX_MEMORY+1) { count = PCILYNX_MAX_MEMORY+1 - off; } if (count <= 0) { return 0; } down(&md->lynx->mem_dma_mutex); switch (md->type) { case rom: reg_write(md->lynx, LBUS_ADDR, LBUS_ADDR_SEL_ROM | off); membase = md->lynx->local_rom; break; case ram: reg_write(md->lynx, LBUS_ADDR, LBUS_ADDR_SEL_RAM | off); membase = md->lynx->local_ram; break; case aux: reg_write(md->lynx, LBUS_ADDR, LBUS_ADDR_SEL_AUX | off); membase = md->lynx->aux_port; break; default: panic("pcilynx%d: unsupported md->type %d in " __FUNCTION__, md->lynx->id, md->type); } if (count < mem_mindma) { memcpy_fromio(md->lynx->mem_dma_buffer, membase+off, count); copy_to_user(buffer, md->lynx->mem_dma_buffer, count); bcount = 0; goto done; } bcount = count; alignfix = 4 - (off % 4); if (alignfix != 4) { if (bcount < alignfix) { alignfix = bcount; } memcpy_fromio(md->lynx->mem_dma_buffer, membase+off, alignfix); copy_to_user(buffer, md->lynx->mem_dma_buffer, alignfix); if (bcount == alignfix) { goto done; } bcount -= alignfix; buffer += alignfix; off += alignfix; } if (reg_read(md->lynx, DMA0_CHAN_CTRL) & DMA_CHAN_CTRL_BUSY) { PRINT(KERN_WARNING, md->lynx->id, "DMA ALREADY ACTIVE!"); } add_wait_queue(&md->lynx->mem_dma_intr_wait, &wait); if (bcount > 32768) { current->state = TASK_INTERRUPTIBLE; reg_write(md->lynx, DMA0_READY, 1); /* select maxpcl */ run_pcl(md->lynx, md->lynx->mem_pcl.start, 0); while (reg_read(md->lynx, DMA0_CHAN_CTRL) & DMA_CHAN_CTRL_BUSY) { if (signal_pending(current)) { reg_write(md->lynx, DMA0_CHAN_CTRL, 0); goto rmwait_done; } schedule(); } copy_to_user(buffer, md->lynx->mem_dma_buffer, 32768); buffer += 32768; bcount -= 32768; } *(u32 *)(md->lynx->mem_dma_buffer) = pcl_bus(md->lynx, md->lynx->mem_pcl.mod) + pcloffs(buffer[bcount/4000+1].control); *(u32 *)(md->lynx->mem_dma_buffer+4) = PCL_LAST_BUFF | (bcount % 4000); current->state = TASK_INTERRUPTIBLE; reg_write(md->lynx, DMA0_READY, 0); run_pcl(md->lynx, md->lynx->mem_pcl.start, 0); while (reg_read(md->lynx, DMA0_CHAN_CTRL) & DMA_CHAN_CTRL_BUSY) { if (signal_pending(current)) { reg_write(md->lynx, DMA0_CHAN_CTRL, 0); goto rmwait_done; } schedule(); } copy_to_user(buffer, md->lynx->mem_dma_buffer, bcount); bcount = 0; if (reg_read(md->lynx, DMA0_CHAN_CTRL) & DMA_CHAN_CTRL_BUSY) { PRINT(KERN_ERR, md->lynx->id, "DMA STILL ACTIVE!"); } rmwait_done: reg_write(md->lynx, DMA0_CHAN_CTRL, 0); remove_wait_queue(&md->lynx->mem_dma_intr_wait, &wait); done: up(&md->lynx->mem_dma_mutex); count -= bcount; *offset += count; return (count ? count : -EINTR); } static ssize_t mem_write(struct file *file, const char *buffer, size_t count, loff_t *offset) { struct memdata *md = (struct memdata *)file->private_data; if (((*offset) + count) > PCILYNX_MAX_MEMORY+1) { count = PCILYNX_MAX_MEMORY+1 - *offset; } if (count == 0 || *offset > PCILYNX_MAX_MEMORY) { return -ENOSPC; } /* FIXME: dereferencing pointers to PCI mem doesn't work everywhere */ switch (md->type) { case aux: copy_from_user(md->lynx->aux_port+(*offset), buffer, count); break; case ram: copy_from_user(md->lynx->local_ram+(*offset), buffer, count); break; case rom: /* the ROM may be writeable */ copy_from_user(md->lynx->local_rom+(*offset), buffer, count); break; } file->f_pos += count; return count; } /******************************************************** * Global stuff (interrupt handler, init/shutdown code) * ********************************************************/ static void lynx_irq_handler(int irq, void *dev_id, struct pt_regs *regs_are_unused) { struct ti_lynx *lynx = (struct ti_lynx *)dev_id; struct hpsb_host *host = lynx->host; u32 intmask = reg_read(lynx, PCI_INT_STATUS); u32 linkint = reg_read(lynx, LINK_INT_STATUS); reg_write(lynx, PCI_INT_STATUS, intmask); reg_write(lynx, LINK_INT_STATUS, linkint); //printk("-%d- one interrupt: 0x%08x / 0x%08x\n", lynx->id, intmask, linkint); if (intmask & PCI_INT_AUX_INT) { atomic_inc(&lynx->aux_intr_seen); wake_up_interruptible(&lynx->aux_intr_wait); } if (intmask & PCI_INT_DMA0_HLT) { wake_up_interruptible(&lynx->mem_dma_intr_wait); } if (intmask & PCI_INT_1394) { if (linkint & LINK_INT_PHY_TIMEOUT) { PRINT(KERN_INFO, lynx->id, "PHY timeout occured"); } if (linkint & LINK_INT_PHY_BUSRESET) { PRINT(KERN_INFO, lynx->id, "bus reset interrupt"); if (!host->in_bus_reset) { hpsb_bus_reset(host); } } if (linkint & LINK_INT_PHY_REG_RCVD) { if (!host->in_bus_reset) { PRINT(KERN_INFO, lynx->id, "phy reg received without reset"); } } if (linkint & LINK_INT_ISO_STUCK) { PRINT(KERN_INFO, lynx->id, "isochronous transmitter stuck"); } if (linkint & LINK_INT_ASYNC_STUCK) { PRINT(KERN_INFO, lynx->id, "asynchronous transmitter stuck"); } if (linkint & LINK_INT_SENT_REJECT) { PRINT(KERN_INFO, lynx->id, "sent reject"); } if (linkint & LINK_INT_TX_INVALID_TC) { PRINT(KERN_INFO, lynx->id, "invalid transaction code"); } if (linkint & LINK_INT_GRF_OVERFLOW) { PRINT(KERN_INFO, lynx->id, "GRF overflow"); } if (linkint & LINK_INT_ITF_UNDERFLOW) { PRINT(KERN_INFO, lynx->id, "ITF underflow"); } if (linkint & LINK_INT_ATF_UNDERFLOW) { PRINT(KERN_INFO, lynx->id, "ATF underflow"); } } if (intmask & PCI_INT_DMA_HLT(CHANNEL_ISO_RCV)) { PRINTD(KERN_DEBUG, lynx->id, "iso receive"); spin_lock(&lynx->iso_rcv.lock); lynx->iso_rcv.stat[lynx->iso_rcv.next] = reg_read(lynx, DMA_CHAN_STAT(CHANNEL_ISO_RCV)); lynx->iso_rcv.used++; lynx->iso_rcv.next = (lynx->iso_rcv.next + 1) % NUM_ISORCV_PCL; if ((lynx->iso_rcv.next == lynx->iso_rcv.last) || !lynx->iso_rcv.chan_count) { PRINTD(KERN_DEBUG, lynx->id, "stopped"); reg_write(lynx, DMA_WORD1_CMP_ENABLE(CHANNEL_ISO_RCV), 0); } run_sub_pcl(lynx, lynx->iso_rcv.pcl_start, lynx->iso_rcv.next, CHANNEL_ISO_RCV); spin_unlock(&lynx->iso_rcv.lock); queue_task(&lynx->iso_rcv.tq, &tq_immediate); mark_bh(IMMEDIATE_BH); } if (intmask & PCI_INT_DMA3_HLT) { /* async send DMA completed */ u32 ack; struct hpsb_packet *packet; spin_lock(&lynx->async_queue_lock); ack = reg_read(lynx, DMA3_CHAN_STAT); packet = lynx->async_queue; lynx->async_queue = packet->xnext; if (lynx->async_queue != NULL) { send_next_async(lynx); } spin_unlock(&lynx->async_queue_lock); if (ack & DMA_CHAN_STAT_SPECIALACK) { ack = (ack >> 15) & 0xf; PRINTD(KERN_INFO, lynx->id, "special ack %d", ack); ack = (ack == 1 ? ACKX_TIMEOUT : ACKX_SEND_ERROR); } else { ack = (ack >> 15) & 0xf; } hpsb_packet_sent(host, packet, ack); } if (intmask & (PCI_INT_DMA1_HLT | PCI_INT_DMA1_PCL)) { /* general receive DMA completed */ int stat = reg_read(lynx, DMA1_CHAN_STAT); PRINTD(KERN_DEBUG, lynx->id, "received packet size %d", stat & 0x1fff); if (stat & DMA_CHAN_STAT_SELFID) { handle_selfid(lynx, host, stat & 0x1fff); reg_set_bits(lynx, LINK_CONTROL, LINK_CONTROL_RCV_CMP_VALID | LINK_CONTROL_TX_ASYNC_EN | LINK_CONTROL_RX_ASYNC_EN); } else { quadlet_t *q_data = lynx->rcv_page; if ((*q_data >> 4 & 0xf) == TCODE_READQ_RESPONSE || (*q_data >> 4 & 0xf) == TCODE_WRITEQ) { cpu_to_be32s(q_data + 3); } hpsb_packet_received(host, q_data, stat & 0x1fff); } run_pcl(lynx, lynx->rcv_pcl_start, 1); } } static void iso_rcv_bh(struct ti_lynx *lynx) { unsigned int idx; quadlet_t *data; unsigned long flags; spin_lock_irqsave(&lynx->iso_rcv.lock, flags); while (lynx->iso_rcv.used) { idx = lynx->iso_rcv.last; spin_unlock_irqrestore(&lynx->iso_rcv.lock, flags); data = lynx->iso_rcv.page[idx / ISORCV_PER_PAGE] + (idx % ISORCV_PER_PAGE) * MAX_ISORCV_SIZE; if (lynx->iso_rcv.stat[idx] & (DMA_CHAN_STAT_PCIERR | DMA_CHAN_STAT_PKTERR)) { PRINT(KERN_INFO, lynx->id, "iso receive error on %d to 0x%p", idx, data); } else { hpsb_packet_received(lynx->host, data, lynx->iso_rcv.stat[idx] & 0x1fff); } spin_lock_irqsave(&lynx->iso_rcv.lock, flags); lynx->iso_rcv.last = (idx + 1) % NUM_ISORCV_PCL; lynx->iso_rcv.used--; } if (lynx->iso_rcv.chan_count) { reg_write(lynx, DMA_WORD1_CMP_ENABLE(CHANNEL_ISO_RCV), DMA_WORD1_CMP_ENABLE_MASTER); } spin_unlock_irqrestore(&lynx->iso_rcv.lock, flags); } static int add_card(struct pci_dev *dev) { #define FAIL(fmt, args...) \ PRINT_G(KERN_ERR, fmt , ## args); \ num_of_cards--; \ remove_card(lynx); \ return 1 struct ti_lynx *lynx; /* shortcut to currently handled device */ unsigned long page; unsigned int i; if (num_of_cards == MAX_PCILYNX_CARDS) { PRINT_G(KERN_WARNING, "cannot handle more than %d cards. " "Adjust MAX_PCILYNX_CARDS in ti_pcilynx.h.", MAX_PCILYNX_CARDS); return 1; } lynx = &cards[num_of_cards++]; lynx->id = num_of_cards-1; lynx->dev = dev; pci_set_master(dev); if (!request_irq(dev->irq, lynx_irq_handler, SA_SHIRQ, PCILYNX_DRIVER_NAME, lynx)) { PRINT(KERN_INFO, lynx->id, "allocated interrupt %d", dev->irq); lynx->state = have_intr; } else { FAIL("failed to allocate shared interrupt %d", dev->irq); } #ifndef CONFIG_IEEE1394_PCILYNX_LOCALRAM lynx->pcl_mem = kmalloc(8 * sizeof(lynx->pcl_bmap) * sizeof(struct ti_pcl), GFP_KERNEL); if (lynx->pcl_mem != NULL) { lynx->state = have_pcl_mem; PRINT(KERN_INFO, lynx->id, "allocated PCL memory %d Bytes @ 0x%p", 8 * sizeof(lynx->pcl_bmap) * sizeof(struct ti_pcl), lynx->pcl_mem); } else { FAIL("failed to allocate PCL memory area"); } #endif lynx->mem_dma_buffer = kmalloc(32768, GFP_KERNEL); if (lynx->mem_dma_buffer != NULL) { lynx->state = have_aux_buf; } else { FAIL("failed to allocate DMA buffer for aux"); } page = get_free_page(GFP_KERNEL); if (page != 0) { lynx->rcv_page = (void *)page; lynx->state = have_1394_buffers; } else { FAIL("failed to allocate receive buffer"); } for (i = 0; i < ISORCV_PAGES; i++) { page = get_free_page(GFP_KERNEL); if (page != 0) { lynx->iso_rcv.page[i] = (void *)page; } else { FAIL("failed to allocate iso receive buffers"); } } #if LINUX_VERSION_CODE < KERNEL_VERSION(2,3,13) lynx->registers = ioremap_nocache(dev->base_address[0], PCILYNX_MAX_REGISTER); lynx->local_ram = ioremap(dev->base_address[1], PCILYNX_MAX_MEMORY); lynx->aux_port = ioremap(dev->base_address[2], PCILYNX_MAX_MEMORY); #else lynx->registers = ioremap_nocache(dev->resource[0].start, PCILYNX_MAX_REGISTER); lynx->local_ram = ioremap(dev->resource[1].start, PCILYNX_MAX_MEMORY); lynx->aux_port = ioremap(dev->resource[2].start, PCILYNX_MAX_MEMORY); #endif #if LINUX_VERSION_CODE < KERNEL_VERSION(2,3,15) lynx->local_rom = ioremap(dev->rom_address, PCILYNX_MAX_MEMORY); #else lynx->local_rom = ioremap(dev->resource[PCI_ROM_RESOURCE].start, PCILYNX_MAX_MEMORY); #endif lynx->state = have_iomappings; if (lynx->registers == NULL) { FAIL("failed to remap registers - card not accessible"); } #ifdef CONFIG_IEEE1394_PCILYNX_LOCALRAM if (lynx->local_ram == NULL) { FAIL("failed to remap local RAM which is required for " "operation"); } #endif /* alloc_pcl return values are not checked, it is expected that the * provided PCL space is sufficient for the initial allocations */ if (lynx->aux_port != NULL) { lynx->mem_pcl.start = alloc_pcl(lynx); lynx->mem_pcl.cmd = alloc_pcl(lynx); lynx->mem_pcl.mod = alloc_pcl(lynx); lynx->mem_pcl.max = alloc_pcl(lynx); aux_setup_pcls(lynx); sema_init(&lynx->mem_dma_mutex, 1); } lynx->rcv_pcl = alloc_pcl(lynx); lynx->rcv_pcl_start = alloc_pcl(lynx); lynx->async_pcl = alloc_pcl(lynx); lynx->async_pcl_start = alloc_pcl(lynx); for (i = 0; i < NUM_ISORCV_PCL; i++) { lynx->iso_rcv.pcl[i] = alloc_pcl(lynx); } lynx->iso_rcv.pcl_start = alloc_pcl(lynx); /* all allocations successful - simple init stuff follows */ lynx->lock = SPIN_LOCK_UNLOCKED; reg_write(lynx, PCI_INT_ENABLE, PCI_INT_AUX_INT | PCI_INT_DMA_ALL); init_waitqueue_head(&lynx->mem_dma_intr_wait); init_waitqueue_head(&lynx->aux_intr_wait); lynx->iso_rcv.tq.routine = (void (*)(void*))iso_rcv_bh; lynx->iso_rcv.tq.data = lynx; lynx->iso_rcv.lock = SPIN_LOCK_UNLOCKED; PRINT(KERN_INFO, lynx->id, "remapped memory spaces reg 0x%p, rom 0x%p, " "ram 0x%p, aux 0x%p", lynx->registers, lynx->local_rom, lynx->local_ram, lynx->aux_port); /* now, looking for PHY register set */ if ((get_phy_reg(lynx, 2) & 0xe0) == 0xe0) { lynx->phyic.reg_1394a = 1; PRINT(KERN_INFO, lynx->id, "found 1394a conform PHY (using extended register set)"); lynx->phyic.vendor = get_phy_vendorid(lynx); lynx->phyic.product = get_phy_productid(lynx); } else { lynx->phyic.reg_1394a = 0; PRINT(KERN_INFO, lynx->id, "found old 1394 PHY"); } return 0; #undef FAIL } static void remove_card(struct ti_lynx *lynx) { int i; switch (lynx->state) { case have_iomappings: reg_write(lynx, PCI_INT_ENABLE, 0); reg_write(lynx, MISC_CONTROL, MISC_CONTROL_SWRESET); iounmap(lynx->registers); iounmap(lynx->local_rom); iounmap(lynx->local_ram); iounmap(lynx->aux_port); case have_1394_buffers: for (i = 0; i < ISORCV_PAGES; i++) { if (lynx->iso_rcv.page[i]) { free_page((unsigned long)lynx->iso_rcv.page[i]); } } free_page((unsigned long)lynx->rcv_page); case have_aux_buf: kfree(lynx->mem_dma_buffer); case have_pcl_mem: #ifndef CONFIG_IEEE1394_PCILYNX_LOCALRAM kfree(lynx->pcl_mem); #endif case have_intr: free_irq(lynx->dev->irq, lynx); case clear: /* do nothing - already freed */ } lynx->state = clear; } static int init_driver() { struct pci_dev *dev = NULL; int success = 0; if (num_of_cards) { PRINT_G(KERN_DEBUG, __PRETTY_FUNCTION__ " called again"); return 0; } PRINT_G(KERN_INFO, "looking for PCILynx cards"); while ((dev = pci_find_device(PCI_VENDOR_ID_TI, PCI_DEVICE_ID_TI_PCILYNX, dev)) != NULL) { if (add_card(dev) == 0) { success = 1; } } if (success == 0) { PRINT_G(KERN_WARNING, "no operable PCILynx cards found"); return -ENXIO; } if (register_chrdev(PCILYNX_MAJOR, PCILYNX_DRIVER_NAME, &aux_ops)) { PRINT_G(KERN_ERR, "allocation of char major number %d failed", PCILYNX_MAJOR); return -EBUSY; } return 0; } static size_t get_lynx_rom(struct hpsb_host *host, const quadlet_t **ptr) { *ptr = lynx_csr_rom; return sizeof(lynx_csr_rom); } struct hpsb_host_template *get_lynx_template(void) { static struct hpsb_host_template tmpl = { name: "pcilynx", detect_hosts: lynx_detect, initialize_host: lynx_initialize, release_host: lynx_release, get_rom: get_lynx_rom, transmit_packet: lynx_transmit, devctl: lynx_devctl }; return &tmpl; } #ifdef MODULE /* EXPORT_NO_SYMBOLS; */ MODULE_AUTHOR("Andreas E. Bombe "); MODULE_DESCRIPTION("driver for Texas Instruments PCI Lynx IEEE-1394 controller"); MODULE_SUPPORTED_DEVICE("pcilynx"); void cleanup_module(void) { hpsb_unregister_lowlevel(get_lynx_template()); PRINT_G(KERN_INFO, "removed " PCILYNX_DRIVER_NAME " module"); } int init_module(void) { if (hpsb_register_lowlevel(get_lynx_template())) { PRINT_G(KERN_ERR, "registering failed"); return -ENXIO; } else { return 0; } } #endif /* MODULE */