drivers/net/s2io.c

   1 /************************************************************************
   2  * s2io.c: A Linux PCI-X Ethernet driver for S2IO 10GbE Server NIC
   3  * Copyright(c) 2002-2005 Neterion Inc.
   4
   5  * This software may be used and distributed according to the terms of
   6  * the GNU General Public License (GPL), incorporated herein by reference.
   7  * Drivers based on or derived from this code fall under the GPL and must
   8  * retain the authorship, copyright and license notice.  This file is not
   9  * a complete program and may only be used when the entire operating
  10  * system is licensed under the GPL.
  11  * See the file COPYING in this distribution for more information.
  12  *
  13  * Credits:
  14  * Jeff Garzik          : For pointing out the improper error condition
  15  *                        check in the s2io_xmit routine and also some
  16  *                        issues in the Tx watch dog function. Also for
  17  *                        patiently answering all those innumerable
  18  *                        questions regaring the 2.6 porting issues.
  19  * Stephen Hemminger    : Providing proper 2.6 porting mechanism for some
  20  *                        macros available only in 2.6 Kernel.
  21  * Francois Romieu      : For pointing out all code part that were
  22  *                        deprecated and also styling related comments.
  23  * Grant Grundler       : For helping me get rid of some Architecture
  24  *                        dependent code.
  25  * Christopher Hellwig  : Some more 2.6 specific issues in the driver.
  26  *
  27  * The module loadable parameters that are supported by the driver and a brief
  28  * explaination of all the variables.
  29  * rx_ring_num : This can be used to program the number of receive rings used
  30  * in the driver.
  31  * rx_ring_len: This defines the number of descriptors each ring can have. This
  32  * is also an array of size 8.
  33  * tx_fifo_num: This defines the number of Tx FIFOs thats used int the driver.
  34  * tx_fifo_len: This too is an array of 8. Each element defines the number of
  35  * Tx descriptors that can be associated with each corresponding FIFO.
  36  * in PCI Configuration space.
  37  ************************************************************************/
  38
  39 #include <linux/config.h>
  40 #include <linux/module.h>
  41 #include <linux/types.h>
  42 #include <linux/errno.h>
  43 #include <linux/ioport.h>
  44 #include <linux/pci.h>
  45 #include <linux/kernel.h>
  46 #include <linux/netdevice.h>
  47 #include <linux/etherdevice.h>
  48 #include <linux/skbuff.h>
  49 #include <linux/init.h>
  50 #include <linux/delay.h>
  51 #include <linux/stddef.h>
  52 #include <linux/ioctl.h>
  53 #include <linux/timex.h>
  54 #include <linux/sched.h>
  55 #include <linux/ethtool.h>
  56 #include <linux/version.h>
  57 #include <linux/workqueue.h>
  58
  59 #include <asm/io.h>
  60 #include <asm/system.h>
  61 #include <asm/uaccess.h>
  62
  63 /* local include */
  64 #include "s2io.h"
  65 #include "s2io-regs.h"
  66
  67 /* S2io Driver name & version. */
  68 static char s2io_driver_name[] = "s2io";
  69 static char s2io_driver_version[] = "Version 1.7.7.1";
  70
  71 /*
  72  * Cards with following subsystem_id have a link state indication
  73  * problem, 600B, 600C, 600D, 640B, 640C and 640D.
  74  * macro below identifies these cards given the subsystem_id.
  75  */
  76 #define CARDS_WITH_FAULTY_LINK_INDICATORS(subid) \
  77                 (((subid >= 0x600B) && (subid <= 0x600D)) || \
  78                  ((subid >= 0x640B) && (subid <= 0x640D))) ? 1 : 0
  79
  80 #define LINK_IS_UP(val64) (!(val64 & (ADAPTER_STATUS_RMAC_REMOTE_FAULT | \
  81                                       ADAPTER_STATUS_RMAC_LOCAL_FAULT)))
  82 #define TASKLET_IN_USE test_and_set_bit(0, (&sp->tasklet_status))
  83 #define PANIC   1
  84 #define LOW     2
  85 static inline int rx_buffer_level(nic_t * sp, int rxb_size, int ring)
  86 {
  87         int level = 0;
  88         if ((sp->pkt_cnt[ring] - rxb_size) > 16) {
  89                 level = LOW;
  90                 if ((sp->pkt_cnt[ring] - rxb_size) < MAX_RXDS_PER_BLOCK) {
  91                         level = PANIC;
  92                 }
  93         }
  94
  95         return level;
  96 }
  97
  98 /* Ethtool related variables and Macros. */
  99 static char s2io_gstrings[][ETH_GSTRING_LEN] = {
 100         "Register test\t(offline)",
 101         "Eeprom test\t(offline)",
 102         "Link test\t(online)",
 103         "RLDRAM test\t(offline)",
 104         "BIST Test\t(offline)"
 105 };
 106
 107 static char ethtool_stats_keys[][ETH_GSTRING_LEN] = {
 108         {"tmac_frms"},
 109         {"tmac_data_octets"},
 110         {"tmac_drop_frms"},
 111         {"tmac_mcst_frms"},
 112         {"tmac_bcst_frms"},
 113         {"tmac_pause_ctrl_frms"},
 114         {"tmac_any_err_frms"},
 115         {"tmac_vld_ip_octets"},
 116         {"tmac_vld_ip"},
 117         {"tmac_drop_ip"},
 118         {"tmac_icmp"},
 119         {"tmac_rst_tcp"},
 120         {"tmac_tcp"},
 121         {"tmac_udp"},
 122         {"rmac_vld_frms"},
 123         {"rmac_data_octets"},
 124         {"rmac_fcs_err_frms"},
 125         {"rmac_drop_frms"},
 126         {"rmac_vld_mcst_frms"},
 127         {"rmac_vld_bcst_frms"},
 128         {"rmac_in_rng_len_err_frms"},
 129         {"rmac_long_frms"},
 130         {"rmac_pause_ctrl_frms"},
 131         {"rmac_discarded_frms"},
 132         {"rmac_usized_frms"},
 133         {"rmac_osized_frms"},
 134         {"rmac_frag_frms"},
 135         {"rmac_jabber_frms"},
 136         {"rmac_ip"},
 137         {"rmac_ip_octets"},
 138         {"rmac_hdr_err_ip"},
 139         {"rmac_drop_ip"},
 140         {"rmac_icmp"},
 141         {"rmac_tcp"},
 142         {"rmac_udp"},
 143         {"rmac_err_drp_udp"},
 144         {"rmac_pause_cnt"},
 145         {"rmac_accepted_ip"},
 146         {"rmac_err_tcp"},
 147 };
 148
 149 #define S2IO_STAT_LEN sizeof(ethtool_stats_keys)/ ETH_GSTRING_LEN
 150 #define S2IO_STAT_STRINGS_LEN S2IO_STAT_LEN * ETH_GSTRING_LEN
 151
 152 #define S2IO_TEST_LEN   sizeof(s2io_gstrings) / ETH_GSTRING_LEN
 153 #define S2IO_STRINGS_LEN        S2IO_TEST_LEN * ETH_GSTRING_LEN
 154
 155
 156 /*
 157  * Constants to be programmed into the Xena's registers, to configure
 158  * the XAUI.
 159  */
 160
 161 #define SWITCH_SIGN     0xA5A5A5A5A5A5A5A5ULL
 162 #define END_SIGN        0x0
 163
 164 static u64 default_mdio_cfg[] = {
 165         /* Reset PMA PLL */
 166         0xC001010000000000ULL, 0xC0010100000000E0ULL,
 167         0xC0010100008000E4ULL,
 168         /* Remove Reset from PMA PLL */
 169         0xC001010000000000ULL, 0xC0010100000000E0ULL,
 170         0xC0010100000000E4ULL,
 171         END_SIGN
 172 };
 173
 174 static u64 default_dtx_cfg[] = {
 175         0x8000051500000000ULL, 0x80000515000000E0ULL,
 176         0x80000515D93500E4ULL, 0x8001051500000000ULL,
 177         0x80010515000000E0ULL, 0x80010515001E00E4ULL,
 178         0x8002051500000000ULL, 0x80020515000000E0ULL,
 179         0x80020515F21000E4ULL,
 180         /* Set PADLOOPBACKN */
 181         0x8002051500000000ULL, 0x80020515000000E0ULL,
 182         0x80020515B20000E4ULL, 0x8003051500000000ULL,
 183         0x80030515000000E0ULL, 0x80030515B20000E4ULL,
 184         0x8004051500000000ULL, 0x80040515000000E0ULL,
 185         0x80040515B20000E4ULL, 0x8005051500000000ULL,
 186         0x80050515000000E0ULL, 0x80050515B20000E4ULL,
 187         SWITCH_SIGN,
 188         /* Remove PADLOOPBACKN */
 189         0x8002051500000000ULL, 0x80020515000000E0ULL,
 190         0x80020515F20000E4ULL, 0x8003051500000000ULL,
 191         0x80030515000000E0ULL, 0x80030515F20000E4ULL,
 192         0x8004051500000000ULL, 0x80040515000000E0ULL,
 193         0x80040515F20000E4ULL, 0x8005051500000000ULL,
 194         0x80050515000000E0ULL, 0x80050515F20000E4ULL,
 195         END_SIGN
 196 };
 197
 198
 199 /*
 200  * Constants for Fixing the MacAddress problem seen mostly on
 201  * Alpha machines.
 202  */
 203 static u64 fix_mac[] = {
 204         0x0060000000000000ULL, 0x0060600000000000ULL,
 205         0x0040600000000000ULL, 0x0000600000000000ULL,
 206         0x0020600000000000ULL, 0x0060600000000000ULL,
 207         0x0020600000000000ULL, 0x0060600000000000ULL,
 208         0x0020600000000000ULL, 0x0060600000000000ULL,
 209         0x0020600000000000ULL, 0x0060600000000000ULL,
 210         0x0020600000000000ULL, 0x0060600000000000ULL,
 211         0x0020600000000000ULL, 0x0060600000000000ULL,
 212         0x0020600000000000ULL, 0x0060600000000000ULL,
 213         0x0020600000000000ULL, 0x0060600000000000ULL,
 214         0x0020600000000000ULL, 0x0060600000000000ULL,
 215         0x0020600000000000ULL, 0x0060600000000000ULL,
 216         0x0020600000000000ULL, 0x0000600000000000ULL,
 217         0x0040600000000000ULL, 0x0060600000000000ULL,
 218         END_SIGN
 219 };
 220
 221 /* Module Loadable parameters. */
 222 static unsigned int tx_fifo_num = 1;
 223 static unsigned int tx_fifo_len[MAX_TX_FIFOS] =
 224     {[0 ...(MAX_TX_FIFOS - 1)] = 0 };
 225 static unsigned int rx_ring_num = 1;
 226 static unsigned int rx_ring_sz[MAX_RX_RINGS] =
 227     {[0 ...(MAX_RX_RINGS - 1)] = 0 };
 228 static unsigned int Stats_refresh_time = 4;
 229 static unsigned int rmac_pause_time = 65535;
 230 static unsigned int mc_pause_threshold_q0q3 = 187;
 231 static unsigned int mc_pause_threshold_q4q7 = 187;
 232 static unsigned int shared_splits;
 233 static unsigned int tmac_util_period = 5;
 234 static unsigned int rmac_util_period = 5;
 235 #ifndef CONFIG_S2IO_NAPI
 236 static unsigned int indicate_max_pkts;
 237 #endif
 238
 239 /*
 240  * S2IO device table.
 241  * This table lists all the devices that this driver supports.
 242  */
 243 static struct pci_device_id s2io_tbl[] __devinitdata = {
 244         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_WIN,
 245          PCI_ANY_ID, PCI_ANY_ID},
 246         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_S2IO_UNI,
 247          PCI_ANY_ID, PCI_ANY_ID},
 248         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_WIN,
 249          PCI_ANY_ID, PCI_ANY_ID},
 250         {PCI_VENDOR_ID_S2IO, PCI_DEVICE_ID_HERC_UNI,
 251          PCI_ANY_ID, PCI_ANY_ID},
 252         {0,}
 253 };
 254
 255 MODULE_DEVICE_TABLE(pci, s2io_tbl);
 256
 257 static struct pci_driver s2io_driver = {
 258       .name = "S2IO",
 259       .id_table = s2io_tbl,
 260       .probe = s2io_init_nic,
 261       .remove = __devexit_p(s2io_rem_nic),
 262 };
 263
 264 /* A simplifier macro used both by init and free shared_mem Fns(). */
 265 #define TXD_MEM_PAGE_CNT(len, per_each) ((len+per_each - 1) / per_each)
 266
 267 /**
 268  * init_shared_mem - Allocation and Initialization of Memory
 269  * @nic: Device private variable.
 270  * Description: The function allocates all the memory areas shared
 271  * between the NIC and the driver. This includes Tx descriptors,
 272  * Rx descriptors and the statistics block.
 273  */
 274
 275 static int init_shared_mem(struct s2io_nic *nic)
 276 {
 277         u32 size;
 278         void *tmp_v_addr, *tmp_v_addr_next;
 279         dma_addr_t tmp_p_addr, tmp_p_addr_next;
 280         RxD_block_t *pre_rxd_blk = NULL;
 281         int i, j, blk_cnt;
 282         int lst_size, lst_per_page;
 283         struct net_device *dev = nic->dev;
 284 #ifdef CONFIG_2BUFF_MODE
 285         unsigned long tmp;
 286         buffAdd_t *ba;
 287 #endif
 288
 289         mac_info_t *mac_control;
 290         struct config_param *config;
 291
 292         mac_control = &nic->mac_control;
 293         config = &nic->config;
 294
 295
 296         /* Allocation and initialization of TXDLs in FIOFs */
 297         size = 0;
 298         for (i = 0; i < config->tx_fifo_num; i++) {
 299                 size += config->tx_cfg[i].fifo_len;
 300         }
 301         if (size > MAX_AVAILABLE_TXDS) {
 302                 DBG_PRINT(ERR_DBG, "%s: Total number of Tx FIFOs ",
 303                           dev->name);
 304                 DBG_PRINT(ERR_DBG, "exceeds the maximum value ");
 305                 DBG_PRINT(ERR_DBG, "that can be used\n");
 306                 return FAILURE;
 307         }
 308
 309         lst_size = (sizeof(TxD_t) * config->max_txds);
 310         lst_per_page = PAGE_SIZE / lst_size;
 311
 312         for (i = 0; i < config->tx_fifo_num; i++) {
 313                 int fifo_len = config->tx_cfg[i].fifo_len;
 314                 int list_holder_size = fifo_len * sizeof(list_info_hold_t);
 315                 nic->list_info[i] = kmalloc(list_holder_size, GFP_KERNEL);
 316                 if (!nic->list_info[i]) {
 317                         DBG_PRINT(ERR_DBG,
 318                                   "Malloc failed for list_info\n");
 319                         return -ENOMEM;
 320                 }
 321                 memset(nic->list_info[i], 0, list_holder_size);
 322         }
 323         for (i = 0; i < config->tx_fifo_num; i++) {
 324                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
 325                                                 lst_per_page);
 326                 mac_control->tx_curr_put_info[i].offset = 0;
 327                 mac_control->tx_curr_put_info[i].fifo_len =
 328                     config->tx_cfg[i].fifo_len - 1;
 329                 mac_control->tx_curr_get_info[i].offset = 0;
 330                 mac_control->tx_curr_get_info[i].fifo_len =
 331                     config->tx_cfg[i].fifo_len - 1;
 332                 for (j = 0; j < page_num; j++) {
 333                         int k = 0;
 334                         dma_addr_t tmp_p;
 335                         void *tmp_v;
 336                         tmp_v = pci_alloc_consistent(nic->pdev,
 337                                                      PAGE_SIZE, &tmp_p);
 338                         if (!tmp_v) {
 339                                 DBG_PRINT(ERR_DBG,
 340                                           "pci_alloc_consistent ");
 341                                 DBG_PRINT(ERR_DBG, "failed for TxDL\n");
 342                                 return -ENOMEM;
 343                         }
 344                         while (k < lst_per_page) {
 345                                 int l = (j * lst_per_page) + k;
 346                                 if (l == config->tx_cfg[i].fifo_len)
 347                                         goto end_txd_alloc;
 348                                 nic->list_info[i][l].list_virt_addr =
 349                                     tmp_v + (k * lst_size);
 350                                 nic->list_info[i][l].list_phy_addr =
 351                                     tmp_p + (k * lst_size);
 352                                 k++;
 353                         }
 354                 }
 355         }
 356       end_txd_alloc:
 357
 358         /* Allocation and initialization of RXDs in Rings */
 359         size = 0;
 360         for (i = 0; i < config->rx_ring_num; i++) {
 361                 if (config->rx_cfg[i].num_rxd % (MAX_RXDS_PER_BLOCK + 1)) {
 362                         DBG_PRINT(ERR_DBG, "%s: RxD count of ", dev->name);
 363                         DBG_PRINT(ERR_DBG, "Ring%d is not a multiple of ",
 364                                   i);
 365                         DBG_PRINT(ERR_DBG, "RxDs per Block");
 366                         return FAILURE;
 367                 }
 368                 size += config->rx_cfg[i].num_rxd;
 369                 nic->block_count[i] =
 370                     config->rx_cfg[i].num_rxd / (MAX_RXDS_PER_BLOCK + 1);
 371                 nic->pkt_cnt[i] =
 372                     config->rx_cfg[i].num_rxd - nic->block_count[i];
 373         }
 374
 375         for (i = 0; i < config->rx_ring_num; i++) {
 376                 mac_control->rx_curr_get_info[i].block_index = 0;
 377                 mac_control->rx_curr_get_info[i].offset = 0;
 378                 mac_control->rx_curr_get_info[i].ring_len =
 379                     config->rx_cfg[i].num_rxd - 1;
 380                 mac_control->rx_curr_put_info[i].block_index = 0;
 381                 mac_control->rx_curr_put_info[i].offset = 0;
 382                 mac_control->rx_curr_put_info[i].ring_len =
 383                     config->rx_cfg[i].num_rxd - 1;
 384                 blk_cnt =
 385                     config->rx_cfg[i].num_rxd / (MAX_RXDS_PER_BLOCK + 1);
 386                 /*  Allocating all the Rx blocks */
 387                 for (j = 0; j < blk_cnt; j++) {
 388 #ifndef CONFIG_2BUFF_MODE
 389                         size = (MAX_RXDS_PER_BLOCK + 1) * (sizeof(RxD_t));
 390 #else
 391                         size = SIZE_OF_BLOCK;
 392 #endif
 393                         tmp_v_addr = pci_alloc_consistent(nic->pdev, size,
 394                                                           &tmp_p_addr);
 395                         if (tmp_v_addr == NULL) {
 396                                 /*
 397                                  * In case of failure, free_shared_mem()
 398                                  * is called, which should free any
 399                                  * memory that was alloced till the
 400                                  * failure happened.
 401                                  */
 402                                 nic->rx_blocks[i][j].block_virt_addr =
 403                                     tmp_v_addr;
 404                                 return -ENOMEM;
 405                         }
 406                         memset(tmp_v_addr, 0, size);
 407                         nic->rx_blocks[i][j].block_virt_addr = tmp_v_addr;
 408                         nic->rx_blocks[i][j].block_dma_addr = tmp_p_addr;
 409                 }
 410                 /* Interlinking all Rx Blocks */
 411                 for (j = 0; j < blk_cnt; j++) {
 412                         tmp_v_addr = nic->rx_blocks[i][j].block_virt_addr;
 413                         tmp_v_addr_next =
 414                             nic->rx_blocks[i][(j + 1) %
 415                                               blk_cnt].block_virt_addr;
 416                         tmp_p_addr = nic->rx_blocks[i][j].block_dma_addr;
 417                         tmp_p_addr_next =
 418                             nic->rx_blocks[i][(j + 1) %
 419                                               blk_cnt].block_dma_addr;
 420
 421                         pre_rxd_blk = (RxD_block_t *) tmp_v_addr;
 422                         pre_rxd_blk->reserved_1 = END_OF_BLOCK; /* last RxD
 423                                                                  * marker.
 424                                                                  */
 425 #ifndef CONFIG_2BUFF_MODE
 426                         pre_rxd_blk->reserved_2_pNext_RxD_block =
 427                             (unsigned long) tmp_v_addr_next;
 428 #endif
 429                         pre_rxd_blk->pNext_RxD_Blk_physical =
 430                             (u64) tmp_p_addr_next;
 431                 }
 432         }
 433
 434 #ifdef CONFIG_2BUFF_MODE
 435         /*
 436          * Allocation of Storages for buffer addresses in 2BUFF mode
 437          * and the buffers as well.
 438          */
 439         for (i = 0; i < config->rx_ring_num; i++) {
 440                 blk_cnt =
 441                     config->rx_cfg[i].num_rxd / (MAX_RXDS_PER_BLOCK + 1);
 442                 nic->ba[i] = kmalloc((sizeof(buffAdd_t *) * blk_cnt),
 443                                      GFP_KERNEL);
 444                 if (!nic->ba[i])
 445                         return -ENOMEM;
 446                 for (j = 0; j < blk_cnt; j++) {
 447                         int k = 0;
 448                         nic->ba[i][j] = kmalloc((sizeof(buffAdd_t) *
 449                                                  (MAX_RXDS_PER_BLOCK + 1)),
 450                                                 GFP_KERNEL);
 451                         if (!nic->ba[i][j])
 452                                 return -ENOMEM;
 453                         while (k != MAX_RXDS_PER_BLOCK) {
 454                                 ba = &nic->ba[i][j][k];
 455
 456                                 ba->ba_0_org = kmalloc
 457                                     (BUF0_LEN + ALIGN_SIZE, GFP_KERNEL);
 458                                 if (!ba->ba_0_org)
 459                                         return -ENOMEM;
 460                                 tmp = (unsigned long) ba->ba_0_org;
 461                                 tmp += ALIGN_SIZE;
 462                                 tmp &= ~((unsigned long) ALIGN_SIZE);
 463                                 ba->ba_0 = (void *) tmp;
 464
 465                                 ba->ba_1_org = kmalloc
 466                                     (BUF1_LEN + ALIGN_SIZE, GFP_KERNEL);
 467                                 if (!ba->ba_1_org)
 468                                         return -ENOMEM;
 469                                 tmp = (unsigned long) ba->ba_1_org;
 470                                 tmp += ALIGN_SIZE;
 471                                 tmp &= ~((unsigned long) ALIGN_SIZE);
 472                                 ba->ba_1 = (void *) tmp;
 473                                 k++;
 474                         }
 475                 }
 476         }
 477 #endif
 478
 479         /* Allocation and initialization of Statistics block */
 480         size = sizeof(StatInfo_t);
 481         mac_control->stats_mem = pci_alloc_consistent
 482             (nic->pdev, size, &mac_control->stats_mem_phy);
 483
 484         if (!mac_control->stats_mem) {
 485                 /*
 486                  * In case of failure, free_shared_mem() is called, which
 487                  * should free any memory that was alloced till the
 488                  * failure happened.
 489                  */
 490                 return -ENOMEM;
 491         }
 492         mac_control->stats_mem_sz = size;
 493
 494         tmp_v_addr = mac_control->stats_mem;
 495         mac_control->stats_info = (StatInfo_t *) tmp_v_addr;
 496         memset(tmp_v_addr, 0, size);
 497
 498         DBG_PRINT(INIT_DBG, "%s:Ring Mem PHY: 0x%llx\n", dev->name,
 499                   (unsigned long long) tmp_p_addr);
 500
 501         return SUCCESS;
 502 }
 503
 504 /**
 505  * free_shared_mem - Free the allocated Memory
 506  * @nic:  Device private variable.
 507  * Description: This function is to free all memory locations allocated by
 508  * the init_shared_mem() function and return it to the kernel.
 509  */
 510
 511 static void free_shared_mem(struct s2io_nic *nic)
 512 {
 513         int i, j, blk_cnt, size;
 514         void *tmp_v_addr;
 515         dma_addr_t tmp_p_addr;
 516         mac_info_t *mac_control;
 517         struct config_param *config;
 518         int lst_size, lst_per_page;
 519
 520
 521         if (!nic)
 522                 return;
 523
 524         mac_control = &nic->mac_control;
 525         config = &nic->config;
 526
 527         lst_size = (sizeof(TxD_t) * config->max_txds);
 528         lst_per_page = PAGE_SIZE / lst_size;
 529
 530         for (i = 0; i < config->tx_fifo_num; i++) {
 531                 int page_num = TXD_MEM_PAGE_CNT(config->tx_cfg[i].fifo_len,
 532                                                 lst_per_page);
 533                 for (j = 0; j < page_num; j++) {
 534                         int mem_blks = (j * lst_per_page);
 535                         if (!nic->list_info[i][mem_blks].list_virt_addr)
 536                                 break;
 537                         pci_free_consistent(nic->pdev, PAGE_SIZE,
 538                                             nic->list_info[i][mem_blks].
 539                                             list_virt_addr,
 540                                             nic->list_info[i][mem_blks].
 541                                             list_phy_addr);
 542                 }
 543                 kfree(nic->list_info[i]);
 544         }
 545
 546 #ifndef CONFIG_2BUFF_MODE
 547         size = (MAX_RXDS_PER_BLOCK + 1) * (sizeof(RxD_t));
 548 #else
 549         size = SIZE_OF_BLOCK;
 550 #endif
 551         for (i = 0; i < config->rx_ring_num; i++) {
 552                 blk_cnt = nic->block_count[i];
 553                 for (j = 0; j < blk_cnt; j++) {
 554                         tmp_v_addr = nic->rx_blocks[i][j].block_virt_addr;
 555                         tmp_p_addr = nic->rx_blocks[i][j].block_dma_addr;
 556                         if (tmp_v_addr == NULL)
 557                                 break;
 558                         pci_free_consistent(nic->pdev, size,
 559                                             tmp_v_addr, tmp_p_addr);
 560                 }
 561         }
 562
 563 #ifdef CONFIG_2BUFF_MODE
 564         /* Freeing buffer storage addresses in 2BUFF mode. */
 565         for (i = 0; i < config->rx_ring_num; i++) {
 566                 blk_cnt =
 567                     config->rx_cfg[i].num_rxd / (MAX_RXDS_PER_BLOCK + 1);
 568                 if (!nic->ba[i])
 569                         goto end_free;
 570                 for (j = 0; j < blk_cnt; j++) {
 571                         int k = 0;
 572                         if (!nic->ba[i][j]) {
 573                                 kfree(nic->ba[i]);
 574                                 goto end_free;
 575                         }
 576                         while (k != MAX_RXDS_PER_BLOCK) {
 577                                 buffAdd_t *ba = &nic->ba[i][j][k];
 578                                 if (!ba || !ba->ba_0_org || !ba->ba_1_org)
 579                                 {
 580                                         kfree(nic->ba[i]);
 581                                         kfree(nic->ba[i][j]);
 582                                         if(ba->ba_0_org)
 583                                                 kfree(ba->ba_0_org);
 584                                         if(ba->ba_1_org)
 585                                                 kfree(ba->ba_1_org);
 586                                         goto end_free;
 587                                 }
 588                                 kfree(ba->ba_0_org);
 589                                 kfree(ba->ba_1_org);
 590                                 k++;
 591                         }
 592                         kfree(nic->ba[i][j]);
 593                 }
 594                 kfree(nic->ba[i]);
 595         }
 596 end_free:
 597 #endif
 598
 599         if (mac_control->stats_mem) {
 600                 pci_free_consistent(nic->pdev,
 601                                     mac_control->stats_mem_sz,
 602                                     mac_control->stats_mem,
 603                                     mac_control->stats_mem_phy);
 604         }
 605 }
 606
 607 /**
 608  *  init_nic - Initialization of hardware
 609  *  @nic: device peivate variable
 610  *  Description: The function sequentially configures every block
 611  *  of the H/W from their reset values.
 612  *  Return Value:  SUCCESS on success and
 613  *  '-1' on failure (endian settings incorrect).
 614  */
 615
 616 static int init_nic(struct s2io_nic *nic)
 617 {
 618         XENA_dev_config_t __iomem *bar0 = nic->bar0;
 619         struct net_device *dev = nic->dev;
 620         register u64 val64 = 0;
 621         void __iomem *add;
 622         u32 time;
 623         int i, j;
 624         mac_info_t *mac_control;
 625         struct config_param *config;
 626         int mdio_cnt = 0, dtx_cnt = 0;
 627         unsigned long long mem_share;
 628
 629         mac_control = &nic->mac_control;
 630         config = &nic->config;
 631
 632         /* Initialize swapper control register */
 633         if (s2io_set_swapper(nic)) {
 634                 DBG_PRINT(ERR_DBG,"ERROR: Setting Swapper failed\n");
 635                 return -1;
 636         }
 637
 638         /* Remove XGXS from reset state */
 639         val64 = 0;
 640         writeq(val64, &bar0->sw_reset);
 641         val64 = readq(&bar0->sw_reset);
 642         msleep(500);
 643
 644         /*  Enable Receiving broadcasts */
 645         add = &bar0->mac_cfg;
 646         val64 = readq(&bar0->mac_cfg);
 647         val64 |= MAC_RMAC_BCAST_ENABLE;
 648         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
 649         writel((u32) val64, add);
 650         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
 651         writel((u32) (val64 >> 32), (add + 4));
 652
 653         /* Read registers in all blocks */
 654         val64 = readq(&bar0->mac_int_mask);
 655         val64 = readq(&bar0->mc_int_mask);
 656         val64 = readq(&bar0->xgxs_int_mask);
 657
 658         /*  Set MTU */
 659         val64 = dev->mtu;
 660         writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
 661
 662         /*
 663          * Configuring the XAUI Interface of Xena.
 664          * ***************************************
 665          * To Configure the Xena's XAUI, one has to write a series
 666          * of 64 bit values into two registers in a particular
 667          * sequence. Hence a macro 'SWITCH_SIGN' has been defined
 668          * which will be defined in the array of configuration values
 669          * (default_dtx_cfg & default_mdio_cfg) at appropriate places
 670          * to switch writing from one regsiter to another. We continue
 671          * writing these values until we encounter the 'END_SIGN' macro.
 672          * For example, After making a series of 21 writes into
 673          * dtx_control register the 'SWITCH_SIGN' appears and hence we
 674          * start writing into mdio_control until we encounter END_SIGN.
 675          */
 676         while (1) {
 677               dtx_cfg:
 678                 while (default_dtx_cfg[dtx_cnt] != END_SIGN) {
 679                         if (default_dtx_cfg[dtx_cnt] == SWITCH_SIGN) {
 680                                 dtx_cnt++;
 681                                 goto mdio_cfg;
 682                         }
 683                         SPECIAL_REG_WRITE(default_dtx_cfg[dtx_cnt],
 684                                           &bar0->dtx_control, UF);
 685                         val64 = readq(&bar0->dtx_control);
 686                         dtx_cnt++;
 687                 }
 688               mdio_cfg:
 689                 while (default_mdio_cfg[mdio_cnt] != END_SIGN) {
 690                         if (default_mdio_cfg[mdio_cnt] == SWITCH_SIGN) {
 691                                 mdio_cnt++;
 692                                 goto dtx_cfg;
 693                         }
 694                         SPECIAL_REG_WRITE(default_mdio_cfg[mdio_cnt],
 695                                           &bar0->mdio_control, UF);
 696                         val64 = readq(&bar0->mdio_control);
 697                         mdio_cnt++;
 698                 }
 699                 if ((default_dtx_cfg[dtx_cnt] == END_SIGN) &&
 700                     (default_mdio_cfg[mdio_cnt] == END_SIGN)) {
 701                         break;
 702                 } else {
 703                         goto dtx_cfg;
 704                 }
 705         }
 706
 707         /*  Tx DMA Initialization */
 708         val64 = 0;
 709         writeq(val64, &bar0->tx_fifo_partition_0);
 710         writeq(val64, &bar0->tx_fifo_partition_1);
 711         writeq(val64, &bar0->tx_fifo_partition_2);
 712         writeq(val64, &bar0->tx_fifo_partition_3);
 713
 714
 715         for (i = 0, j = 0; i < config->tx_fifo_num; i++) {
 716                 val64 |=
 717                     vBIT(config->tx_cfg[i].fifo_len - 1, ((i * 32) + 19),
 718                          13) | vBIT(config->tx_cfg[i].fifo_priority,
 719                                     ((i * 32) + 5), 3);
 720
 721                 if (i == (config->tx_fifo_num - 1)) {
 722                         if (i % 2 == 0)
 723                                 i++;
 724                 }
 725
 726                 switch (i) {
 727                 case 1:
 728                         writeq(val64, &bar0->tx_fifo_partition_0);
 729                         val64 = 0;
 730                         break;
 731                 case 3:
 732                         writeq(val64, &bar0->tx_fifo_partition_1);
 733                         val64 = 0;
 734                         break;
 735                 case 5:
 736                         writeq(val64, &bar0->tx_fifo_partition_2);
 737                         val64 = 0;
 738                         break;
 739                 case 7:
 740                         writeq(val64, &bar0->tx_fifo_partition_3);
 741                         break;
 742                 }
 743         }
 744
 745         /* Enable Tx FIFO partition 0. */
 746         val64 = readq(&bar0->tx_fifo_partition_0);
 747         val64 |= BIT(0);        /* To enable the FIFO partition. */
 748         writeq(val64, &bar0->tx_fifo_partition_0);
 749
 750         val64 = readq(&bar0->tx_fifo_partition_0);
 751         DBG_PRINT(INIT_DBG, "Fifo partition at: 0x%p is: 0x%llx\n",
 752                   &bar0->tx_fifo_partition_0, (unsigned long long) val64);
 753
 754         /*
 755          * Initialization of Tx_PA_CONFIG register to ignore packet
 756          * integrity checking.
 757          */
 758         val64 = readq(&bar0->tx_pa_cfg);
 759         val64 |= TX_PA_CFG_IGNORE_FRM_ERR | TX_PA_CFG_IGNORE_SNAP_OUI |
 760             TX_PA_CFG_IGNORE_LLC_CTRL | TX_PA_CFG_IGNORE_L2_ERR;
 761         writeq(val64, &bar0->tx_pa_cfg);
 762
 763         /* Rx DMA intialization. */
 764         val64 = 0;
 765         for (i = 0; i < config->rx_ring_num; i++) {
 766                 val64 |=
 767                     vBIT(config->rx_cfg[i].ring_priority, (5 + (i * 8)),
 768                          3);
 769         }
 770         writeq(val64, &bar0->rx_queue_priority);
 771
 772         /*
 773          * Allocating equal share of memory to all the
 774          * configured Rings.
 775          */
 776         val64 = 0;
 777         for (i = 0; i < config->rx_ring_num; i++) {
 778                 switch (i) {
 779                 case 0:
 780                         mem_share = (64 / config->rx_ring_num +
 781                                      64 % config->rx_ring_num);
 782                         val64 |= RX_QUEUE_CFG_Q0_SZ(mem_share);
 783                         continue;
 784                 case 1:
 785                         mem_share = (64 / config->rx_ring_num);
 786                         val64 |= RX_QUEUE_CFG_Q1_SZ(mem_share);
 787                         continue;
 788                 case 2:
 789                         mem_share = (64 / config->rx_ring_num);
 790                         val64 |= RX_QUEUE_CFG_Q2_SZ(mem_share);
 791                         continue;
 792                 case 3:
 793                         mem_share = (64 / config->rx_ring_num);
 794                         val64 |= RX_QUEUE_CFG_Q3_SZ(mem_share);
 795                         continue;
 796                 case 4:
 797                         mem_share = (64 / config->rx_ring_num);
 798                         val64 |= RX_QUEUE_CFG_Q4_SZ(mem_share);
 799                         continue;
 800                 case 5:
 801                         mem_share = (64 / config->rx_ring_num);
 802                         val64 |= RX_QUEUE_CFG_Q5_SZ(mem_share);
 803                         continue;
 804                 case 6:
 805                         mem_share = (64 / config->rx_ring_num);
 806                         val64 |= RX_QUEUE_CFG_Q6_SZ(mem_share);
 807                         continue;
 808                 case 7:
 809                         mem_share = (64 / config->rx_ring_num);
 810                         val64 |= RX_QUEUE_CFG_Q7_SZ(mem_share);
 811                         continue;
 812                 }
 813         }
 814         writeq(val64, &bar0->rx_queue_cfg);
 815
 816         /*
 817          * Initializing the Tx round robin registers to 0.
 818          * Filling Tx and Rx round robin registers as per the
 819          * number of FIFOs and Rings is still TODO.
 820          */
 821         writeq(0, &bar0->tx_w_round_robin_0);
 822         writeq(0, &bar0->tx_w_round_robin_1);
 823         writeq(0, &bar0->tx_w_round_robin_2);
 824         writeq(0, &bar0->tx_w_round_robin_3);
 825         writeq(0, &bar0->tx_w_round_robin_4);
 826
 827         /*
 828          * TODO
 829          * Disable Rx steering. Hard coding all packets be steered to
 830          * Queue 0 for now.
 831          */
 832         val64 = 0x8080808080808080ULL;
 833         writeq(val64, &bar0->rts_qos_steering);
 834
 835         /* UDP Fix */
 836         val64 = 0;
 837         for (i = 1; i < 8; i++)
 838                 writeq(val64, &bar0->rts_frm_len_n[i]);
 839
 840         /* Set rts_frm_len register for fifo 0 */
 841         writeq(MAC_RTS_FRM_LEN_SET(dev->mtu + 22),
 842                &bar0->rts_frm_len_n[0]);
 843
 844         /* Enable statistics */
 845         writeq(mac_control->stats_mem_phy, &bar0->stat_addr);
 846         val64 = SET_UPDT_PERIOD(Stats_refresh_time) |
 847             STAT_CFG_STAT_RO | STAT_CFG_STAT_EN;
 848         writeq(val64, &bar0->stat_cfg);
 849
 850         /*
 851          * Initializing the sampling rate for the device to calculate the
 852          * bandwidth utilization.
 853          */
 854         val64 = MAC_TX_LINK_UTIL_VAL(tmac_util_period) |
 855             MAC_RX_LINK_UTIL_VAL(rmac_util_period);
 856         writeq(val64, &bar0->mac_link_util);
 857
 858
 859         /*
 860          * Initializing the Transmit and Receive Traffic Interrupt
 861          * Scheme.
 862          */
 863         /* TTI Initialization. Default Tx timer gets us about
 864          * 250 interrupts per sec. Continuous interrupts are enabled
 865          * by default.
 866          */
 867         val64 = TTI_DATA1_MEM_TX_TIMER_VAL(0x2078) |
 868             TTI_DATA1_MEM_TX_URNG_A(0xA) |
 869             TTI_DATA1_MEM_TX_URNG_B(0x10) |
 870             TTI_DATA1_MEM_TX_URNG_C(0x30) | TTI_DATA1_MEM_TX_TIMER_AC_EN |
 871                 TTI_DATA1_MEM_TX_TIMER_CI_EN;
 872         writeq(val64, &bar0->tti_data1_mem);
 873
 874         val64 = TTI_DATA2_MEM_TX_UFC_A(0x10) |
 875             TTI_DATA2_MEM_TX_UFC_B(0x20) |
 876             TTI_DATA2_MEM_TX_UFC_C(0x40) | TTI_DATA2_MEM_TX_UFC_D(0x80);
 877         writeq(val64, &bar0->tti_data2_mem);
 878
 879         val64 = TTI_CMD_MEM_WE | TTI_CMD_MEM_STROBE_NEW_CMD;
 880         writeq(val64, &bar0->tti_command_mem);
 881
 882         /*
 883          * Once the operation completes, the Strobe bit of the command
 884          * register will be reset. We poll for this particular condition
 885          * We wait for a maximum of 500ms for the operation to complete,
 886          * if it's not complete by then we return error.
 887          */
 888         time = 0;
 889         while (TRUE) {
 890                 val64 = readq(&bar0->tti_command_mem);
 891                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
 892                         break;
 893                 }
 894                 if (time > 10) {
 895                         DBG_PRINT(ERR_DBG, "%s: TTI init Failed\n",
 896                                   dev->name);
 897                         return -1;
 898                 }
 899                 msleep(50);
 900                 time++;
 901         }
 902
 903         /* RTI Initialization */
 904         val64 = RTI_DATA1_MEM_RX_TIMER_VAL(0xFFF) |
 905             RTI_DATA1_MEM_RX_URNG_A(0xA) |
 906             RTI_DATA1_MEM_RX_URNG_B(0x10) |
 907             RTI_DATA1_MEM_RX_URNG_C(0x30) | RTI_DATA1_MEM_RX_TIMER_AC_EN;
 908
 909         writeq(val64, &bar0->rti_data1_mem);
 910
 911         val64 = RTI_DATA2_MEM_RX_UFC_A(0x1) |
 912             RTI_DATA2_MEM_RX_UFC_B(0x2) |
 913             RTI_DATA2_MEM_RX_UFC_C(0x40) | RTI_DATA2_MEM_RX_UFC_D(0x80);
 914         writeq(val64, &bar0->rti_data2_mem);
 915
 916         val64 = RTI_CMD_MEM_WE | RTI_CMD_MEM_STROBE_NEW_CMD;
 917         writeq(val64, &bar0->rti_command_mem);
 918
 919         /*
 920          * Once the operation completes, the Strobe bit of the command
 921          * register will be reset. We poll for this particular condition
 922          * We wait for a maximum of 500ms for the operation to complete,
 923          * if it's not complete by then we return error.
 924          */
 925         time = 0;
 926         while (TRUE) {
 927                 val64 = readq(&bar0->rti_command_mem);
 928                 if (!(val64 & TTI_CMD_MEM_STROBE_NEW_CMD)) {
 929                         break;
 930                 }
 931                 if (time > 10) {
 932                         DBG_PRINT(ERR_DBG, "%s: RTI init Failed\n",
 933                                   dev->name);
 934                         return -1;
 935                 }
 936                 time++;
 937                 msleep(50);
 938         }
 939
 940         /*
 941          * Initializing proper values as Pause threshold into all
 942          * the 8 Queues on Rx side.
 943          */
 944         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q0q3);
 945         writeq(0xffbbffbbffbbffbbULL, &bar0->mc_pause_thresh_q4q7);
 946
 947         /* Disable RMAC PAD STRIPPING */
 948         add = &bar0->mac_cfg;
 949         val64 = readq(&bar0->mac_cfg);
 950         val64 &= ~(MAC_CFG_RMAC_STRIP_PAD);
 951         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
 952         writel((u32) (val64), add);
 953         writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
 954         writel((u32) (val64 >> 32), (add + 4));
 955         val64 = readq(&bar0->mac_cfg);
 956
 957         /*
 958          * Set the time value to be inserted in the pause frame
 959          * generated by xena.
 960          */
 961         val64 = readq(&bar0->rmac_pause_cfg);
 962         val64 &= ~(RMAC_PAUSE_HG_PTIME(0xffff));
 963         val64 |= RMAC_PAUSE_HG_PTIME(nic->mac_control.rmac_pause_time);
 964         writeq(val64, &bar0->rmac_pause_cfg);
 965
 966         /*
 967          * Set the Threshold Limit for Generating the pause frame
 968          * If the amount of data in any Queue exceeds ratio of
 969          * (mac_control.mc_pause_threshold_q0q3 or q4q7)/256
 970          * pause frame is generated
 971          */
 972         val64 = 0;
 973         for (i = 0; i < 4; i++) {
 974                 val64 |=
 975                     (((u64) 0xFF00 | nic->mac_control.
 976                       mc_pause_threshold_q0q3)
 977                      << (i * 2 * 8));
 978         }
 979         writeq(val64, &bar0->mc_pause_thresh_q0q3);
 980
 981         val64 = 0;
 982         for (i = 0; i < 4; i++) {
 983                 val64 |=
 984                     (((u64) 0xFF00 | nic->mac_control.
 985                       mc_pause_threshold_q4q7)
 986                      << (i * 2 * 8));
 987         }
 988         writeq(val64, &bar0->mc_pause_thresh_q4q7);
 989
 990         /*
 991          * TxDMA will stop Read request if the number of read split has
 992          * exceeded the limit pointed by shared_splits
 993          */
 994         val64 = readq(&bar0->pic_control);
 995         val64 |= PIC_CNTL_SHARED_SPLITS(shared_splits);
 996         writeq(val64, &bar0->pic_control);
 997
 998         return SUCCESS;
 999 }
1000
1001 /**
1002  *  en_dis_able_nic_intrs - Enable or Disable the interrupts
1003  *  @nic: device private variable,
1004  *  @mask: A mask indicating which Intr block must be modified and,
1005  *  @flag: A flag indicating whether to enable or disable the Intrs.
1006  *  Description: This function will either disable or enable the interrupts
1007  *  depending on the flag argument. The mask argument can be used to
1008  *  enable/disable any Intr block.
1009  *  Return Value: NONE.
1010  */
1011
1012 static void en_dis_able_nic_intrs(struct s2io_nic *nic, u16 mask, int flag)
1013 {
1014         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1015         register u64 val64 = 0, temp64 = 0;
1016
1017         /*  Top level interrupt classification */
1018         /*  PIC Interrupts */
1019         if ((mask & (TX_PIC_INTR | RX_PIC_INTR))) {
1020                 /*  Enable PIC Intrs in the general intr mask register */
1021                 val64 = TXPIC_INT_M | PIC_RX_INT_M;
1022                 if (flag == ENABLE_INTRS) {
1023                         temp64 = readq(&bar0->general_int_mask);
1024                         temp64 &= ~((u64) val64);
1025                         writeq(temp64, &bar0->general_int_mask);
1026                         /*
1027                          * Disabled all PCIX, Flash, MDIO, IIC and GPIO
1028                          * interrupts for now.
1029                          * TODO
1030                          */
1031                         writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1032                         /*
1033                          * No MSI Support is available presently, so TTI and
1034                          * RTI interrupts are also disabled.
1035                          */
1036                 } else if (flag == DISABLE_INTRS) {
1037                         /*
1038                          * Disable PIC Intrs in the general
1039                          * intr mask register
1040                          */
1041                         writeq(DISABLE_ALL_INTRS, &bar0->pic_int_mask);
1042                         temp64 = readq(&bar0->general_int_mask);
1043                         val64 |= temp64;
1044                         writeq(val64, &bar0->general_int_mask);
1045                 }
1046         }
1047
1048         /*  DMA Interrupts */
1049         /*  Enabling/Disabling Tx DMA interrupts */
1050         if (mask & TX_DMA_INTR) {
1051                 /* Enable TxDMA Intrs in the general intr mask register */
1052                 val64 = TXDMA_INT_M;
1053                 if (flag == ENABLE_INTRS) {
1054                         temp64 = readq(&bar0->general_int_mask);
1055                         temp64 &= ~((u64) val64);
1056                         writeq(temp64, &bar0->general_int_mask);
1057                         /*
1058                          * Keep all interrupts other than PFC interrupt
1059                          * and PCC interrupt disabled in DMA level.
1060                          */
1061                         val64 = DISABLE_ALL_INTRS & ~(TXDMA_PFC_INT_M |
1062                                                       TXDMA_PCC_INT_M);
1063                         writeq(val64, &bar0->txdma_int_mask);
1064                         /*
1065                          * Enable only the MISC error 1 interrupt in PFC block
1066                          */
1067                         val64 = DISABLE_ALL_INTRS & (~PFC_MISC_ERR_1);
1068                         writeq(val64, &bar0->pfc_err_mask);
1069                         /*
1070                          * Enable only the FB_ECC error interrupt in PCC block
1071                          */
1072                         val64 = DISABLE_ALL_INTRS & (~PCC_FB_ECC_ERR);
1073                         writeq(val64, &bar0->pcc_err_mask);
1074                 } else if (flag == DISABLE_INTRS) {
1075                         /*
1076                          * Disable TxDMA Intrs in the general intr mask
1077                          * register
1078                          */
1079                         writeq(DISABLE_ALL_INTRS, &bar0->txdma_int_mask);
1080                         writeq(DISABLE_ALL_INTRS, &bar0->pfc_err_mask);
1081                         temp64 = readq(&bar0->general_int_mask);
1082                         val64 |= temp64;
1083                         writeq(val64, &bar0->general_int_mask);
1084                 }
1085         }
1086
1087         /*  Enabling/Disabling Rx DMA interrupts */
1088         if (mask & RX_DMA_INTR) {
1089                 /*  Enable RxDMA Intrs in the general intr mask register */
1090                 val64 = RXDMA_INT_M;
1091                 if (flag == ENABLE_INTRS) {
1092                         temp64 = readq(&bar0->general_int_mask);
1093                         temp64 &= ~((u64) val64);
1094                         writeq(temp64, &bar0->general_int_mask);
1095                         /*
1096                          * All RxDMA block interrupts are disabled for now
1097                          * TODO
1098                          */
1099                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1100                 } else if (flag == DISABLE_INTRS) {
1101                         /*
1102                          * Disable RxDMA Intrs in the general intr mask
1103                          * register
1104                          */
1105                         writeq(DISABLE_ALL_INTRS, &bar0->rxdma_int_mask);
1106                         temp64 = readq(&bar0->general_int_mask);
1107                         val64 |= temp64;
1108                         writeq(val64, &bar0->general_int_mask);
1109                 }
1110         }
1111
1112         /*  MAC Interrupts */
1113         /*  Enabling/Disabling MAC interrupts */
1114         if (mask & (TX_MAC_INTR | RX_MAC_INTR)) {
1115                 val64 = TXMAC_INT_M | RXMAC_INT_M;
1116                 if (flag == ENABLE_INTRS) {
1117                         temp64 = readq(&bar0->general_int_mask);
1118                         temp64 &= ~((u64) val64);
1119                         writeq(temp64, &bar0->general_int_mask);
1120                         /*
1121                          * All MAC block error interrupts are disabled for now
1122                          * except the link status change interrupt.
1123                          * TODO
1124                          */
1125                         val64 = MAC_INT_STATUS_RMAC_INT;
1126                         temp64 = readq(&bar0->mac_int_mask);
1127                         temp64 &= ~((u64) val64);
1128                         writeq(temp64, &bar0->mac_int_mask);
1129
1130                         val64 = readq(&bar0->mac_rmac_err_mask);
1131                         val64 &= ~((u64) RMAC_LINK_STATE_CHANGE_INT);
1132                         writeq(val64, &bar0->mac_rmac_err_mask);
1133                 } else if (flag == DISABLE_INTRS) {
1134                         /*
1135                          * Disable MAC Intrs in the general intr mask register
1136                          */
1137                         writeq(DISABLE_ALL_INTRS, &bar0->mac_int_mask);
1138                         writeq(DISABLE_ALL_INTRS,
1139                                &bar0->mac_rmac_err_mask);
1140
1141                         temp64 = readq(&bar0->general_int_mask);
1142                         val64 |= temp64;
1143                         writeq(val64, &bar0->general_int_mask);
1144                 }
1145         }
1146
1147         /*  XGXS Interrupts */
1148         if (mask & (TX_XGXS_INTR | RX_XGXS_INTR)) {
1149                 val64 = TXXGXS_INT_M | RXXGXS_INT_M;
1150                 if (flag == ENABLE_INTRS) {
1151                         temp64 = readq(&bar0->general_int_mask);
1152                         temp64 &= ~((u64) val64);
1153                         writeq(temp64, &bar0->general_int_mask);
1154                         /*
1155                          * All XGXS block error interrupts are disabled for now
1156                          * TODO
1157                          */
1158                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1159                 } else if (flag == DISABLE_INTRS) {
1160                         /*
1161                          * Disable MC Intrs in the general intr mask register
1162                          */
1163                         writeq(DISABLE_ALL_INTRS, &bar0->xgxs_int_mask);
1164                         temp64 = readq(&bar0->general_int_mask);
1165                         val64 |= temp64;
1166                         writeq(val64, &bar0->general_int_mask);
1167                 }
1168         }
1169
1170         /*  Memory Controller(MC) interrupts */
1171         if (mask & MC_INTR) {
1172                 val64 = MC_INT_M;
1173                 if (flag == ENABLE_INTRS) {
1174                         temp64 = readq(&bar0->general_int_mask);
1175                         temp64 &= ~((u64) val64);
1176                         writeq(temp64, &bar0->general_int_mask);
1177                         /*
1178                          * All MC block error interrupts are disabled for now
1179                          * TODO
1180                          */
1181                         writeq(DISABLE_ALL_INTRS, &bar0->mc_int_mask);
1182                 } else if (flag == DISABLE_INTRS) {
1183                         /*
1184                          * Disable MC Intrs in the general intr mask register
1185                          */
1186                         writeq(DISABLE_ALL_INTRS, &bar0->mc_int_mask);
1187                         temp64 = readq(&bar0->general_int_mask);
1188                         val64 |= temp64;
1189                         writeq(val64, &bar0->general_int_mask);
1190                 }
1191         }
1192
1193
1194         /*  Tx traffic interrupts */
1195         if (mask & TX_TRAFFIC_INTR) {
1196                 val64 = TXTRAFFIC_INT_M;
1197                 if (flag == ENABLE_INTRS) {
1198                         temp64 = readq(&bar0->general_int_mask);
1199                         temp64 &= ~((u64) val64);
1200                         writeq(temp64, &bar0->general_int_mask);
1201                         /*
1202                          * Enable all the Tx side interrupts
1203                          * writing 0 Enables all 64 TX interrupt levels
1204                          */
1205                         writeq(0x0, &bar0->tx_traffic_mask);
1206                 } else if (flag == DISABLE_INTRS) {
1207                         /*
1208                          * Disable Tx Traffic Intrs in the general intr mask
1209                          * register.
1210                          */
1211                         writeq(DISABLE_ALL_INTRS, &bar0->tx_traffic_mask);
1212                         temp64 = readq(&bar0->general_int_mask);
1213                         val64 |= temp64;
1214                         writeq(val64, &bar0->general_int_mask);
1215                 }
1216         }
1217
1218         /*  Rx traffic interrupts */
1219         if (mask & RX_TRAFFIC_INTR) {
1220                 val64 = RXTRAFFIC_INT_M;
1221                 if (flag == ENABLE_INTRS) {
1222                         temp64 = readq(&bar0->general_int_mask);
1223                         temp64 &= ~((u64) val64);
1224                         writeq(temp64, &bar0->general_int_mask);
1225                         /* writing 0 Enables all 8 RX interrupt levels */
1226                         writeq(0x0, &bar0->rx_traffic_mask);
1227                 } else if (flag == DISABLE_INTRS) {
1228                         /*
1229                          * Disable Rx Traffic Intrs in the general intr mask
1230                          * register.
1231                          */
1232                         writeq(DISABLE_ALL_INTRS, &bar0->rx_traffic_mask);
1233                         temp64 = readq(&bar0->general_int_mask);
1234                         val64 |= temp64;
1235                         writeq(val64, &bar0->general_int_mask);
1236                 }
1237         }
1238 }
1239
1240 /**
1241  *  verify_xena_quiescence - Checks whether the H/W is ready
1242  *  @val64 :  Value read from adapter status register.
1243  *  @flag : indicates if the adapter enable bit was ever written once
1244  *  before.
1245  *  Description: Returns whether the H/W is ready to go or not. Depending
1246  *  on whether adapter enable bit was written or not the comparison
1247  *  differs and the calling function passes the input argument flag to
1248  *  indicate this.
1249  *  Return: 1 If xena is quiescence
1250  *          0 If Xena is not quiescence
1251  */
1252
1253 static int verify_xena_quiescence(u64 val64, int flag)
1254 {
1255         int ret = 0;
1256         u64 tmp64 = ~((u64) val64);
1257
1258         if (!
1259             (tmp64 &
1260              (ADAPTER_STATUS_TDMA_READY | ADAPTER_STATUS_RDMA_READY |
1261               ADAPTER_STATUS_PFC_READY | ADAPTER_STATUS_TMAC_BUF_EMPTY |
1262               ADAPTER_STATUS_PIC_QUIESCENT | ADAPTER_STATUS_MC_DRAM_READY |
1263               ADAPTER_STATUS_MC_QUEUES_READY | ADAPTER_STATUS_M_PLL_LOCK |
1264               ADAPTER_STATUS_P_PLL_LOCK))) {
1265                 if (flag == FALSE) {
1266                         if (!(val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1267                             ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1268                              ADAPTER_STATUS_RC_PRC_QUIESCENT)) {
1269
1270                                 ret = 1;
1271
1272                         }
1273                 } else {
1274                         if (((val64 & ADAPTER_STATUS_RMAC_PCC_IDLE) ==
1275                              ADAPTER_STATUS_RMAC_PCC_IDLE) &&
1276                             (!(val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ||
1277                              ((val64 & ADAPTER_STATUS_RC_PRC_QUIESCENT) ==
1278                               ADAPTER_STATUS_RC_PRC_QUIESCENT))) {
1279
1280                                 ret = 1;
1281
1282                         }
1283                 }
1284         }
1285
1286         return ret;
1287 }
1288
1289 /**
1290  * fix_mac_address -  Fix for Mac addr problem on Alpha platforms
1291  * @sp: Pointer to device specifc structure
1292  * Description :
1293  * New procedure to clear mac address reading  problems on Alpha platforms
1294  *
1295  */
1296
1297 static void fix_mac_address(nic_t * sp)
1298 {
1299         XENA_dev_config_t __iomem *bar0 = sp->bar0;
1300         u64 val64;
1301         int i = 0;
1302
1303         while (fix_mac[i] != END_SIGN) {
1304                 writeq(fix_mac[i++], &bar0->gpio_control);
1305                 val64 = readq(&bar0->gpio_control);
1306         }
1307 }
1308
1309 /**
1310  *  start_nic - Turns the device on
1311  *  @nic : device private variable.
1312  *  Description:
1313  *  This function actually turns the device on. Before this  function is
1314  *  called,all Registers are configured from their reset states
1315  *  and shared memory is allocated but the NIC is still quiescent. On
1316  *  calling this function, the device interrupts are cleared and the NIC is
1317  *  literally switched on by writing into the adapter control register.
1318  *  Return Value:
1319  *  SUCCESS on success and -1 on failure.
1320  */
1321
1322 static int start_nic(struct s2io_nic *nic)
1323 {
1324         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1325         struct net_device *dev = nic->dev;
1326         register u64 val64 = 0;
1327         u16 interruptible, i;
1328         u16 subid;
1329         mac_info_t *mac_control;
1330         struct config_param *config;
1331
1332         mac_control = &nic->mac_control;
1333         config = &nic->config;
1334
1335         /*  PRC Initialization and configuration */
1336         for (i = 0; i < config->rx_ring_num; i++) {
1337                 writeq((u64) nic->rx_blocks[i][0].block_dma_addr,
1338                        &bar0->prc_rxd0_n[i]);
1339
1340                 val64 = readq(&bar0->prc_ctrl_n[i]);
1341 #ifndef CONFIG_2BUFF_MODE
1342                 val64 |= PRC_CTRL_RC_ENABLED;
1343 #else
1344                 val64 |= PRC_CTRL_RC_ENABLED | PRC_CTRL_RING_MODE_3;
1345 #endif
1346                 writeq(val64, &bar0->prc_ctrl_n[i]);
1347         }
1348
1349 #ifdef CONFIG_2BUFF_MODE
1350         /* Enabling 2 buffer mode by writing into Rx_pa_cfg reg. */
1351         val64 = readq(&bar0->rx_pa_cfg);
1352         val64 |= RX_PA_CFG_IGNORE_L2_ERR;
1353         writeq(val64, &bar0->rx_pa_cfg);
1354 #endif
1355
1356         /*
1357          * Enabling MC-RLDRAM. After enabling the device, we timeout
1358          * for around 100ms, which is approximately the time required
1359          * for the device to be ready for operation.
1360          */
1361         val64 = readq(&bar0->mc_rldram_mrs);
1362         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE | MC_RLDRAM_MRS_ENABLE;
1363         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
1364         val64 = readq(&bar0->mc_rldram_mrs);
1365
1366         msleep(100);                    /* Delay by around 100 ms. */
1367
1368         /* Enabling ECC Protection. */
1369         val64 = readq(&bar0->adapter_control);
1370         val64 &= ~ADAPTER_ECC_EN;
1371         writeq(val64, &bar0->adapter_control);
1372
1373         /*
1374          * Clearing any possible Link state change interrupts that
1375          * could have popped up just before Enabling the card.
1376          */
1377         val64 = readq(&bar0->mac_rmac_err_reg);
1378         if (val64)
1379                 writeq(val64, &bar0->mac_rmac_err_reg);
1380
1381         /*
1382          * Verify if the device is ready to be enabled, if so enable
1383          * it.
1384          */
1385         val64 = readq(&bar0->adapter_status);
1386         if (!verify_xena_quiescence(val64, nic->device_enabled_once)) {
1387                 DBG_PRINT(ERR_DBG, "%s: device is not ready, ", dev->name);
1388                 DBG_PRINT(ERR_DBG, "Adapter status reads: 0x%llx\n",
1389                           (unsigned long long) val64);
1390                 return FAILURE;
1391         }
1392
1393         /*  Enable select interrupts */
1394         interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR | TX_MAC_INTR |
1395             RX_MAC_INTR;
1396         en_dis_able_nic_intrs(nic, interruptible, ENABLE_INTRS);
1397
1398         /*
1399          * With some switches, link might be already up at this point.
1400          * Because of this weird behavior, when we enable laser,
1401          * we may not get link. We need to handle this. We cannot
1402          * figure out which switch is misbehaving. So we are forced to
1403          * make a global change.
1404          */
1405
1406         /* Enabling Laser. */
1407         val64 = readq(&bar0->adapter_control);
1408         val64 |= ADAPTER_EOI_TX_ON;
1409         writeq(val64, &bar0->adapter_control);
1410
1411         /* SXE-002: Initialize link and activity LED */
1412         subid = nic->pdev->subsystem_device;
1413         if ((subid & 0xFF) >= 0x07) {
1414                 val64 = readq(&bar0->gpio_control);
1415                 val64 |= 0x0000800000000000ULL;
1416                 writeq(val64, &bar0->gpio_control);
1417                 val64 = 0x0411040400000000ULL;
1418                 writeq(val64, (void __iomem *) bar0 + 0x2700);
1419         }
1420
1421         /*
1422          * Don't see link state interrupts on certain switches, so
1423          * directly scheduling a link state task from here.
1424          */
1425         schedule_work(&nic->set_link_task);
1426
1427         /*
1428          * Here we are performing soft reset on XGXS to
1429          * force link down. Since link is already up, we will get
1430          * link state change interrupt after this reset
1431          */
1432         SPECIAL_REG_WRITE(0x80010515001E0000ULL, &bar0->dtx_control, UF);
1433         val64 = readq(&bar0->dtx_control);
1434         udelay(50);
1435         SPECIAL_REG_WRITE(0x80010515001E00E0ULL, &bar0->dtx_control, UF);
1436         val64 = readq(&bar0->dtx_control);
1437         udelay(50);
1438         SPECIAL_REG_WRITE(0x80070515001F00E4ULL, &bar0->dtx_control, UF);
1439         val64 = readq(&bar0->dtx_control);
1440         udelay(50);
1441
1442         return SUCCESS;
1443 }
1444
1445 /**
1446  *  free_tx_buffers - Free all queued Tx buffers
1447  *  @nic : device private variable.
1448  *  Description:
1449  *  Free all queued Tx buffers.
1450  *  Return Value: void
1451 */
1452
1453 static void free_tx_buffers(struct s2io_nic *nic)
1454 {
1455         struct net_device *dev = nic->dev;
1456         struct sk_buff *skb;
1457         TxD_t *txdp;
1458         int i, j;
1459         mac_info_t *mac_control;
1460         struct config_param *config;
1461         int cnt = 0;
1462
1463         mac_control = &nic->mac_control;
1464         config = &nic->config;
1465
1466         for (i = 0; i < config->tx_fifo_num; i++) {
1467                 for (j = 0; j < config->tx_cfg[i].fifo_len - 1; j++) {
1468                         txdp = (TxD_t *) nic->list_info[i][j].
1469                             list_virt_addr;
1470                         skb =
1471                             (struct sk_buff *) ((unsigned long) txdp->
1472                                                 Host_Control);
1473                         if (skb == NULL) {
1474                                 memset(txdp, 0, sizeof(TxD_t));
1475                                 continue;
1476                         }
1477                         dev_kfree_skb(skb);
1478                         memset(txdp, 0, sizeof(TxD_t));
1479                         cnt++;
1480                 }
1481                 DBG_PRINT(INTR_DBG,
1482                           "%s:forcibly freeing %d skbs on FIFO%d\n",
1483                           dev->name, cnt, i);
1484                 mac_control->tx_curr_get_info[i].offset = 0;
1485                 mac_control->tx_curr_put_info[i].offset = 0;
1486         }
1487 }
1488
1489 /**
1490  *   stop_nic -  To stop the nic
1491  *   @nic ; device private variable.
1492  *   Description:
1493  *   This function does exactly the opposite of what the start_nic()
1494  *   function does. This function is called to stop the device.
1495  *   Return Value:
1496  *   void.
1497  */
1498
1499 static void stop_nic(struct s2io_nic *nic)
1500 {
1501         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1502         register u64 val64 = 0;
1503         u16 interruptible, i;
1504         mac_info_t *mac_control;
1505         struct config_param *config;
1506
1507         mac_control = &nic->mac_control;
1508         config = &nic->config;
1509
1510         /*  Disable all interrupts */
1511         interruptible = TX_TRAFFIC_INTR | RX_TRAFFIC_INTR | TX_MAC_INTR |
1512             RX_MAC_INTR;
1513         en_dis_able_nic_intrs(nic, interruptible, DISABLE_INTRS);
1514
1515         /*  Disable PRCs */
1516         for (i = 0; i < config->rx_ring_num; i++) {
1517                 val64 = readq(&bar0->prc_ctrl_n[i]);
1518                 val64 &= ~((u64) PRC_CTRL_RC_ENABLED);
1519                 writeq(val64, &bar0->prc_ctrl_n[i]);
1520         }
1521 }
1522
1523 /**
1524  *  fill_rx_buffers - Allocates the Rx side skbs
1525  *  @nic:  device private variable
1526  *  @ring_no: ring number
1527  *  Description:
1528  *  The function allocates Rx side skbs and puts the physical
1529  *  address of these buffers into the RxD buffer pointers, so that the NIC
1530  *  can DMA the received frame into these locations.
1531  *  The NIC supports 3 receive modes, viz
1532  *  1. single buffer,
1533  *  2. three buffer and
1534  *  3. Five buffer modes.
1535  *  Each mode defines how many fragments the received frame will be split
1536  *  up into by the NIC. The frame is split into L3 header, L4 Header,
1537  *  L4 payload in three buffer mode and in 5 buffer mode, L4 payload itself
1538  *  is split into 3 fragments. As of now only single buffer mode is
1539  *  supported.
1540  *   Return Value:
1541  *  SUCCESS on success or an appropriate -ve value on failure.
1542  */
1543
1544 static int fill_rx_buffers(struct s2io_nic *nic, int ring_no)
1545 {
1546         struct net_device *dev = nic->dev;
1547         struct sk_buff *skb;
1548         RxD_t *rxdp;
1549         int off, off1, size, block_no, block_no1;
1550         int offset, offset1;
1551         u32 alloc_tab = 0;
1552         u32 alloc_cnt = nic->pkt_cnt[ring_no] -
1553             atomic_read(&nic->rx_bufs_left[ring_no]);
1554         mac_info_t *mac_control;
1555         struct config_param *config;
1556 #ifdef CONFIG_2BUFF_MODE
1557         RxD_t *rxdpnext;
1558         int nextblk;
1559         unsigned long tmp;
1560         buffAdd_t *ba;
1561         dma_addr_t rxdpphys;
1562 #endif
1563 #ifndef CONFIG_S2IO_NAPI
1564         unsigned long flags;
1565 #endif
1566
1567         mac_control = &nic->mac_control;
1568         config = &nic->config;
1569
1570         size = dev->mtu + HEADER_ETHERNET_II_802_3_SIZE +
1571             HEADER_802_2_SIZE + HEADER_SNAP_SIZE;
1572
1573         while (alloc_tab < alloc_cnt) {
1574                 block_no = mac_control->rx_curr_put_info[ring_no].
1575                     block_index;
1576                 block_no1 = mac_control->rx_curr_get_info[ring_no].
1577                     block_index;
1578                 off = mac_control->rx_curr_put_info[ring_no].offset;
1579                 off1 = mac_control->rx_curr_get_info[ring_no].offset;
1580 #ifndef CONFIG_2BUFF_MODE
1581                 offset = block_no * (MAX_RXDS_PER_BLOCK + 1) + off;
1582                 offset1 = block_no1 * (MAX_RXDS_PER_BLOCK + 1) + off1;
1583 #else
1584                 offset = block_no * (MAX_RXDS_PER_BLOCK) + off;
1585                 offset1 = block_no1 * (MAX_RXDS_PER_BLOCK) + off1;
1586 #endif
1587
1588                 rxdp = nic->rx_blocks[ring_no][block_no].
1589                     block_virt_addr + off;
1590                 if ((offset == offset1) && (rxdp->Host_Control)) {
1591                         DBG_PRINT(INTR_DBG, "%s: Get and Put", dev->name);
1592                         DBG_PRINT(INTR_DBG, " info equated\n");
1593                         goto end;
1594                 }
1595 #ifndef CONFIG_2BUFF_MODE
1596                 if (rxdp->Control_1 == END_OF_BLOCK) {
1597                         mac_control->rx_curr_put_info[ring_no].
1598                             block_index++;
1599                         mac_control->rx_curr_put_info[ring_no].
1600                             block_index %= nic->block_count[ring_no];
1601                         block_no = mac_control->rx_curr_put_info
1602                             [ring_no].block_index;
1603                         off++;
1604                         off %= (MAX_RXDS_PER_BLOCK + 1);
1605                         mac_control->rx_curr_put_info[ring_no].offset =
1606                             off;
1607                         rxdp = (RxD_t *) ((unsigned long) rxdp->Control_2);
1608                         DBG_PRINT(INTR_DBG, "%s: Next block at: %p\n",
1609                                   dev->name, rxdp);
1610                 }
1611 #ifndef CONFIG_S2IO_NAPI
1612                 spin_lock_irqsave(&nic->put_lock, flags);
1613                 nic->put_pos[ring_no] =
1614                     (block_no * (MAX_RXDS_PER_BLOCK + 1)) + off;
1615                 spin_unlock_irqrestore(&nic->put_lock, flags);
1616 #endif
1617 #else
1618                 if (rxdp->Host_Control == END_OF_BLOCK) {
1619                         mac_control->rx_curr_put_info[ring_no].
1620                             block_index++;
1621                         mac_control->rx_curr_put_info[ring_no].
1622                             block_index %= nic->block_count[ring_no];
1623                         block_no = mac_control->rx_curr_put_info
1624                             [ring_no].block_index;
1625                         off = 0;
1626                         DBG_PRINT(INTR_DBG, "%s: block%d at: 0x%llx\n",
1627                                   dev->name, block_no,
1628                                   (unsigned long long) rxdp->Control_1);
1629                         mac_control->rx_curr_put_info[ring_no].offset =
1630                             off;
1631                         rxdp = nic->rx_blocks[ring_no][block_no].
1632                             block_virt_addr;
1633                 }
1634 #ifndef CONFIG_S2IO_NAPI
1635                 spin_lock_irqsave(&nic->put_lock, flags);
1636                 nic->put_pos[ring_no] = (block_no *
1637                                          (MAX_RXDS_PER_BLOCK + 1)) + off;
1638                 spin_unlock_irqrestore(&nic->put_lock, flags);
1639 #endif
1640 #endif
1641
1642 #ifndef CONFIG_2BUFF_MODE
1643                 if (rxdp->Control_1 & RXD_OWN_XENA)
1644 #else
1645                 if (rxdp->Control_2 & BIT(0))
1646 #endif
1647                 {
1648                         mac_control->rx_curr_put_info[ring_no].
1649                             offset = off;
1650                         goto end;
1651                 }
1652 #ifdef  CONFIG_2BUFF_MODE
1653                 /*
1654                  * RxDs Spanning cache lines will be replenished only
1655                  * if the succeeding RxD is also owned by Host. It
1656                  * will always be the ((8*i)+3) and ((8*i)+6)
1657                  * descriptors for the 48 byte descriptor. The offending
1658                  * decsriptor is of-course the 3rd descriptor.
1659                  */
1660                 rxdpphys = nic->rx_blocks[ring_no][block_no].
1661                     block_dma_addr + (off * sizeof(RxD_t));
1662                 if (((u64) (rxdpphys)) % 128 > 80) {
1663                         rxdpnext = nic->rx_blocks[ring_no][block_no].
1664                             block_virt_addr + (off + 1);
1665                         if (rxdpnext->Host_Control == END_OF_BLOCK) {
1666                                 nextblk = (block_no + 1) %
1667                                     (nic->block_count[ring_no]);
1668                                 rxdpnext = nic->rx_blocks[ring_no]
1669                                     [nextblk].block_virt_addr;
1670                         }
1671                         if (rxdpnext->Control_2 & BIT(0))
1672                                 goto end;
1673                 }
1674 #endif
1675
1676 #ifndef CONFIG_2BUFF_MODE
1677                 skb = dev_alloc_skb(size + NET_IP_ALIGN);
1678 #else
1679                 skb = dev_alloc_skb(dev->mtu + ALIGN_SIZE + BUF0_LEN + 4);
1680 #endif
1681                 if (!skb) {
1682                         DBG_PRINT(ERR_DBG, "%s: Out of ", dev->name);
1683                         DBG_PRINT(ERR_DBG, "memory to allocate SKBs\n");
1684                         return -ENOMEM;
1685                 }
1686 #ifndef CONFIG_2BUFF_MODE
1687                 skb_reserve(skb, NET_IP_ALIGN);
1688                 memset(rxdp, 0, sizeof(RxD_t));
1689                 rxdp->Buffer0_ptr = pci_map_single
1690                     (nic->pdev, skb->data, size, PCI_DMA_FROMDEVICE);
1691                 rxdp->Control_2 &= (~MASK_BUFFER0_SIZE);
1692                 rxdp->Control_2 |= SET_BUFFER0_SIZE(size);
1693                 rxdp->Host_Control = (unsigned long) (skb);
1694                 rxdp->Control_1 |= RXD_OWN_XENA;
1695                 off++;
1696                 off %= (MAX_RXDS_PER_BLOCK + 1);
1697                 mac_control->rx_curr_put_info[ring_no].offset = off;
1698 #else
1699                 ba = &nic->ba[ring_no][block_no][off];
1700                 skb_reserve(skb, BUF0_LEN);
1701                 tmp = (unsigned long) skb->data;
1702                 tmp += ALIGN_SIZE;
1703                 tmp &= ~ALIGN_SIZE;
1704                 skb->data = (void *) tmp;
1705                 skb->tail = (void *) tmp;
1706
1707                 memset(rxdp, 0, sizeof(RxD_t));
1708                 rxdp->Buffer2_ptr = pci_map_single
1709                     (nic->pdev, skb->data, dev->mtu + BUF0_LEN + 4,
1710                      PCI_DMA_FROMDEVICE);
1711                 rxdp->Buffer0_ptr =
1712                     pci_map_single(nic->pdev, ba->ba_0, BUF0_LEN,
1713                                    PCI_DMA_FROMDEVICE);
1714                 rxdp->Buffer1_ptr =
1715                     pci_map_single(nic->pdev, ba->ba_1, BUF1_LEN,
1716                                    PCI_DMA_FROMDEVICE);
1717
1718                 rxdp->Control_2 = SET_BUFFER2_SIZE(dev->mtu + 4);
1719                 rxdp->Control_2 |= SET_BUFFER0_SIZE(BUF0_LEN);
1720                 rxdp->Control_2 |= SET_BUFFER1_SIZE(1); /* dummy. */
1721                 rxdp->Control_2 |= BIT(0);      /* Set Buffer_Empty bit. */
1722                 rxdp->Host_Control = (u64) ((unsigned long) (skb));
1723                 rxdp->Control_1 |= RXD_OWN_XENA;
1724                 off++;
1725                 mac_control->rx_curr_put_info[ring_no].offset = off;
1726 #endif
1727                 atomic_inc(&nic->rx_bufs_left[ring_no]);
1728                 alloc_tab++;
1729         }
1730
1731       end:
1732         return SUCCESS;
1733 }
1734
1735 /**
1736  *  free_rx_buffers - Frees all Rx buffers
1737  *  @sp: device private variable.
1738  *  Description:
1739  *  This function will free all Rx buffers allocated by host.
1740  *  Return Value:
1741  *  NONE.
1742  */
1743
1744 static void free_rx_buffers(struct s2io_nic *sp)
1745 {
1746         struct net_device *dev = sp->dev;
1747         int i, j, blk = 0, off, buf_cnt = 0;
1748         RxD_t *rxdp;
1749         struct sk_buff *skb;
1750         mac_info_t *mac_control;
1751         struct config_param *config;
1752 #ifdef CONFIG_2BUFF_MODE
1753         buffAdd_t *ba;
1754 #endif
1755
1756         mac_control = &sp->mac_control;
1757         config = &sp->config;
1758
1759         for (i = 0; i < config->rx_ring_num; i++) {
1760                 for (j = 0, blk = 0; j < config->rx_cfg[i].num_rxd; j++) {
1761                         off = j % (MAX_RXDS_PER_BLOCK + 1);
1762                         rxdp = sp->rx_blocks[i][blk].block_virt_addr + off;
1763
1764 #ifndef CONFIG_2BUFF_MODE
1765                         if (rxdp->Control_1 == END_OF_BLOCK) {
1766                                 rxdp =
1767                                     (RxD_t *) ((unsigned long) rxdp->
1768                                                Control_2);
1769                                 j++;
1770                                 blk++;
1771                         }
1772 #else
1773                         if (rxdp->Host_Control == END_OF_BLOCK) {
1774                                 blk++;
1775                                 continue;
1776                         }
1777 #endif
1778
1779                         if (!(rxdp->Control_1 & RXD_OWN_XENA)) {
1780                                 memset(rxdp, 0, sizeof(RxD_t));
1781                                 continue;
1782                         }
1783
1784                         skb =
1785                             (struct sk_buff *) ((unsigned long) rxdp->
1786                                                 Host_Control);
1787                         if (skb) {
1788 #ifndef CONFIG_2BUFF_MODE
1789                                 pci_unmap_single(sp->pdev, (dma_addr_t)
1790                                                  rxdp->Buffer0_ptr,
1791                                                  dev->mtu +
1792                                                  HEADER_ETHERNET_II_802_3_SIZE
1793                                                  + HEADER_802_2_SIZE +
1794                                                  HEADER_SNAP_SIZE,
1795                                                  PCI_DMA_FROMDEVICE);
1796 #else
1797                                 ba = &sp->ba[i][blk][off];
1798                                 pci_unmap_single(sp->pdev, (dma_addr_t)
1799                                                  rxdp->Buffer0_ptr,
1800                                                  BUF0_LEN,
1801                                                  PCI_DMA_FROMDEVICE);
1802                                 pci_unmap_single(sp->pdev, (dma_addr_t)
1803                                                  rxdp->Buffer1_ptr,
1804                                                  BUF1_LEN,
1805                                                  PCI_DMA_FROMDEVICE);
1806                                 pci_unmap_single(sp->pdev, (dma_addr_t)
1807                                                  rxdp->Buffer2_ptr,
1808                                                  dev->mtu + BUF0_LEN + 4,
1809                                                  PCI_DMA_FROMDEVICE);
1810 #endif
1811                                 dev_kfree_skb(skb);
1812                                 atomic_dec(&sp->rx_bufs_left[i]);
1813                                 buf_cnt++;
1814                         }
1815                         memset(rxdp, 0, sizeof(RxD_t));
1816                 }
1817                 mac_control->rx_curr_put_info[i].block_index = 0;
1818                 mac_control->rx_curr_get_info[i].block_index = 0;
1819                 mac_control->rx_curr_put_info[i].offset = 0;
1820                 mac_control->rx_curr_get_info[i].offset = 0;
1821                 atomic_set(&sp->rx_bufs_left[i], 0);
1822                 DBG_PRINT(INIT_DBG, "%s:Freed 0x%x Rx Buffers on ring%d\n",
1823                           dev->name, buf_cnt, i);
1824         }
1825 }
1826
1827 /**
1828  * s2io_poll - Rx interrupt handler for NAPI support
1829  * @dev : pointer to the device structure.
1830  * @budget : The number of packets that were budgeted to be processed
1831  * during  one pass through the 'Poll" function.
1832  * Description:
1833  * Comes into picture only if NAPI support has been incorporated. It does
1834  * the same thing that rx_intr_handler does, but not in a interrupt context
1835  * also It will process only a given number of packets.
1836  * Return value:
1837  * 0 on success and 1 if there are No Rx packets to be processed.
1838  */
1839
1840 #ifdef CONFIG_S2IO_NAPI
1841 static int s2io_poll(struct net_device *dev, int *budget)
1842 {
1843         nic_t *nic = dev->priv;
1844         XENA_dev_config_t __iomem *bar0 = nic->bar0;
1845         int pkts_to_process = *budget, pkt_cnt = 0;
1846         register u64 val64 = 0;
1847         rx_curr_get_info_t get_info, put_info;
1848         int i, get_block, put_block, get_offset, put_offset, ring_bufs;
1849 #ifndef CONFIG_2BUFF_MODE
1850         u16 val16, cksum;
1851 #endif
1852         struct sk_buff *skb;
1853         RxD_t *rxdp;
1854         mac_info_t *mac_control;
1855         struct config_param *config;
1856 #ifdef CONFIG_2BUFF_MODE
1857         buffAdd_t *ba;
1858 #endif
1859
1860         mac_control = &nic->mac_control;
1861         config = &nic->config;
1862
1863         if (pkts_to_process > dev->quota)
1864                 pkts_to_process = dev->quota;
1865
1866         val64 = readq(&bar0->rx_traffic_int);
1867         writeq(val64, &bar0->rx_traffic_int);
1868
1869         for (i = 0; i < config->rx_ring_num; i++) {
1870                 get_info = mac_control->rx_curr_get_info[i];
1871                 get_block = get_info.block_index;
1872                 put_info = mac_control->rx_curr_put_info[i];
1873                 put_block = put_info.block_index;
1874                 ring_bufs = config->rx_cfg[i].num_rxd;
1875                 rxdp = nic->rx_blocks[i][get_block].block_virt_addr +
1876                     get_info.offset;
1877 #ifndef CONFIG_2BUFF_MODE
1878                 get_offset = (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
1879                     get_info.offset;
1880                 put_offset = (put_block * (MAX_RXDS_PER_BLOCK + 1)) +
1881                     put_info.offset;
1882                 while ((!(rxdp->Control_1 & RXD_OWN_XENA)) &&
1883                        (((get_offset + 1) % ring_bufs) != put_offset)) {
1884                         if (--pkts_to_process < 0) {
1885                                 goto no_rx;
1886                         }
1887                         if (rxdp->Control_1 == END_OF_BLOCK) {
1888                                 rxdp =
1889                                     (RxD_t *) ((unsigned long) rxdp->
1890                                                Control_2);
1891                                 get_info.offset++;
1892                                 get_info.offset %=
1893                                     (MAX_RXDS_PER_BLOCK + 1);
1894                                 get_block++;
1895                                 get_block %= nic->block_count[i];
1896                                 mac_control->rx_curr_get_info[i].
1897                                     offset = get_info.offset;
1898                                 mac_control->rx_curr_get_info[i].
1899                                     block_index = get_block;
1900                                 continue;
1901                         }
1902                         get_offset =
1903                             (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
1904                             get_info.offset;
1905                         skb =
1906                             (struct sk_buff *) ((unsigned long) rxdp->
1907                                                 Host_Control);
1908                         if (skb == NULL) {
1909                                 DBG_PRINT(ERR_DBG, "%s: The skb is ",
1910                                           dev->name);
1911                                 DBG_PRINT(ERR_DBG, "Null in Rx Intr\n");
1912                                 goto no_rx;
1913                         }
1914                         val64 = RXD_GET_BUFFER0_SIZE(rxdp->Control_2);
1915                         val16 = (u16) (val64 >> 48);
1916                         cksum = RXD_GET_L4_CKSUM(rxdp->Control_1);
1917                         pci_unmap_single(nic->pdev, (dma_addr_t)
1918                                          rxdp->Buffer0_ptr,
1919                                          dev->mtu +
1920                                          HEADER_ETHERNET_II_802_3_SIZE +
1921                                          HEADER_802_2_SIZE +
1922                                          HEADER_SNAP_SIZE,
1923                                          PCI_DMA_FROMDEVICE);
1924                         rx_osm_handler(nic, val16, rxdp, i);
1925                         pkt_cnt++;
1926                         get_info.offset++;
1927                         get_info.offset %= (MAX_RXDS_PER_BLOCK + 1);
1928                         rxdp =
1929                             nic->rx_blocks[i][get_block].block_virt_addr +
1930                             get_info.offset;
1931                         mac_control->rx_curr_get_info[i].offset =
1932                             get_info.offset;
1933                 }
1934 #else
1935                 get_offset = (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
1936                     get_info.offset;
1937                 put_offset = (put_block * (MAX_RXDS_PER_BLOCK + 1)) +
1938                     put_info.offset;
1939                 while (((!(rxdp->Control_1 & RXD_OWN_XENA)) &&
1940                         !(rxdp->Control_2 & BIT(0))) &&
1941                        (((get_offset + 1) % ring_bufs) != put_offset)) {
1942                         if (--pkts_to_process < 0) {
1943                                 goto no_rx;
1944                         }
1945                         skb = (struct sk_buff *) ((unsigned long)
1946                                                   rxdp->Host_Control);
1947                         if (skb == NULL) {
1948                                 DBG_PRINT(ERR_DBG, "%s: The skb is ",
1949                                           dev->name);
1950                                 DBG_PRINT(ERR_DBG, "Null in Rx Intr\n");
1951                                 goto no_rx;
1952                         }
1953
1954                         pci_unmap_single(nic->pdev, (dma_addr_t)
1955                                          rxdp->Buffer0_ptr,
1956                                          BUF0_LEN, PCI_DMA_FROMDEVICE);
1957                         pci_unmap_single(nic->pdev, (dma_addr_t)
1958                                          rxdp->Buffer1_ptr,
1959                                          BUF1_LEN, PCI_DMA_FROMDEVICE);
1960                         pci_unmap_single(nic->pdev, (dma_addr_t)
1961                                          rxdp->Buffer2_ptr,
1962                                          dev->mtu + BUF0_LEN + 4,
1963                                          PCI_DMA_FROMDEVICE);
1964                         ba = &nic->ba[i][get_block][get_info.offset];
1965
1966                         rx_osm_handler(nic, rxdp, i, ba);
1967
1968                         get_info.offset++;
1969                         mac_control->rx_curr_get_info[i].offset =
1970                             get_info.offset;
1971                         rxdp =
1972                             nic->rx_blocks[i][get_block].block_virt_addr +
1973                             get_info.offset;
1974
1975                         if (get_info.offset &&
1976                             (!(get_info.offset % MAX_RXDS_PER_BLOCK))) {
1977                                 get_info.offset = 0;
1978                                 mac_control->rx_curr_get_info[i].
1979                                     offset = get_info.offset;
1980                                 get_block++;
1981                                 get_block %= nic->block_count[i];
1982                                 mac_control->rx_curr_get_info[i].
1983                                     block_index = get_block;
1984                                 rxdp =
1985                                     nic->rx_blocks[i][get_block].
1986                                     block_virt_addr;
1987                         }
1988                         get_offset =
1989                             (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
1990                             get_info.offset;
1991                         pkt_cnt++;
1992                 }
1993 #endif
1994         }
1995         if (!pkt_cnt)
1996                 pkt_cnt = 1;
1997
1998         dev->quota -= pkt_cnt;
1999         *budget -= pkt_cnt;
2000         netif_rx_complete(dev);
2001
2002         for (i = 0; i < config->rx_ring_num; i++) {
2003                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2004                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2005                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2006                         break;
2007                 }
2008         }
2009         /* Re enable the Rx interrupts. */
2010         en_dis_able_nic_intrs(nic, RX_TRAFFIC_INTR, ENABLE_INTRS);
2011         return 0;
2012
2013       no_rx:
2014         dev->quota -= pkt_cnt;
2015         *budget -= pkt_cnt;
2016
2017         for (i = 0; i < config->rx_ring_num; i++) {
2018                 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2019                         DBG_PRINT(ERR_DBG, "%s:Out of memory", dev->name);
2020                         DBG_PRINT(ERR_DBG, " in Rx Poll!!\n");
2021                         break;
2022                 }
2023         }
2024         return 1;
2025 }
2026 #else
2027 /**
2028  *  rx_intr_handler - Rx interrupt handler
2029  *  @nic: device private variable.
2030  *  Description:
2031  *  If the interrupt is because of a received frame or if the
2032  *  receive ring contains fresh as yet un-processed frames,this function is
2033  *  called. It picks out the RxD at which place the last Rx processing had
2034  *  stopped and sends the skb to the OSM's Rx handler and then increments
2035  *  the offset.
2036  *  Return Value:
2037  *  NONE.
2038  */
2039
2040 static void rx_intr_handler(struct s2io_nic *nic)
2041 {
2042         struct net_device *dev = (struct net_device *) nic->dev;
2043         XENA_dev_config_t *bar0 = (XENA_dev_config_t *) nic->bar0;
2044         rx_curr_get_info_t get_info, put_info;
2045         RxD_t *rxdp;
2046         struct sk_buff *skb;
2047 #ifndef CONFIG_2BUFF_MODE
2048         u16 val16, cksum;
2049 #endif
2050         register u64 val64 = 0;
2051         int get_block, get_offset, put_block, put_offset, ring_bufs;
2052         int i, pkt_cnt = 0;
2053         mac_info_t *mac_control;
2054         struct config_param *config;
2055 #ifdef CONFIG_2BUFF_MODE
2056         buffAdd_t *ba;
2057 #endif
2058
2059         mac_control = &nic->mac_control;
2060         config = &nic->config;
2061
2062         /*
2063          * rx_traffic_int reg is an R1 register, hence we read and write back
2064          * the samevalue in the register to clear it.
2065          */
2066         val64 = readq(&bar0->rx_traffic_int);
2067         writeq(val64, &bar0->rx_traffic_int);
2068
2069         for (i = 0; i < config->rx_ring_num; i++) {
2070                 get_info = mac_control->rx_curr_get_info[i];
2071                 get_block = get_info.block_index;
2072                 put_info = mac_control->rx_curr_put_info[i];
2073                 put_block = put_info.block_index;
2074                 ring_bufs = config->rx_cfg[i].num_rxd;
2075                 rxdp = nic->rx_blocks[i][get_block].block_virt_addr +
2076                     get_info.offset;
2077 #ifndef CONFIG_2BUFF_MODE
2078                 get_offset = (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
2079                     get_info.offset;
2080                 spin_lock(&nic->put_lock);
2081                 put_offset = nic->put_pos[i];
2082                 spin_unlock(&nic->put_lock);
2083                 while ((!(rxdp->Control_1 & RXD_OWN_XENA)) &&
2084                        (((get_offset + 1) % ring_bufs) != put_offset)) {
2085                         if (rxdp->Control_1 == END_OF_BLOCK) {
2086                                 rxdp = (RxD_t *) ((unsigned long)
2087                                                   rxdp->Control_2);
2088                                 get_info.offset++;
2089                                 get_info.offset %=
2090                                     (MAX_RXDS_PER_BLOCK + 1);
2091                                 get_block++;
2092                                 get_block %= nic->block_count[i];
2093                                 mac_control->rx_curr_get_info[i].
2094                                     offset = get_info.offset;
2095                                 mac_control->rx_curr_get_info[i].
2096                                     block_index = get_block;
2097                                 continue;
2098                         }
2099                         get_offset =
2100                             (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
2101                             get_info.offset;
2102                         skb = (struct sk_buff *) ((unsigned long)
2103                                                   rxdp->Host_Control);
2104                         if (skb == NULL) {
2105                                 DBG_PRINT(ERR_DBG, "%s: The skb is ",
2106                                           dev->name);
2107                                 DBG_PRINT(ERR_DBG, "Null in Rx Intr\n");
2108                                 return;
2109                         }
2110                         val64 = RXD_GET_BUFFER0_SIZE(rxdp->Control_2);
2111                         val16 = (u16) (val64 >> 48);
2112                         cksum = RXD_GET_L4_CKSUM(rxdp->Control_1);
2113                         pci_unmap_single(nic->pdev, (dma_addr_t)
2114                                          rxdp->Buffer0_ptr,
2115                                          dev->mtu +
2116                                          HEADER_ETHERNET_II_802_3_SIZE +
2117                                          HEADER_802_2_SIZE +
2118                                          HEADER_SNAP_SIZE,
2119                                          PCI_DMA_FROMDEVICE);
2120                         rx_osm_handler(nic, val16, rxdp, i);
2121                         get_info.offset++;
2122                         get_info.offset %= (MAX_RXDS_PER_BLOCK + 1);
2123                         rxdp =
2124                             nic->rx_blocks[i][get_block].block_virt_addr +
2125                             get_info.offset;
2126                         mac_control->rx_curr_get_info[i].offset =
2127                             get_info.offset;
2128                         pkt_cnt++;
2129                         if ((indicate_max_pkts)
2130                             && (pkt_cnt > indicate_max_pkts))
2131                                 break;
2132                 }
2133 #else
2134                 get_offset = (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
2135                     get_info.offset;
2136                 spin_lock(&nic->put_lock);
2137                 put_offset = nic->put_pos[i];
2138                 spin_unlock(&nic->put_lock);
2139                 while (((!(rxdp->Control_1 & RXD_OWN_XENA)) &&
2140                         !(rxdp->Control_2 & BIT(0))) &&
2141                        (((get_offset + 1) % ring_bufs) != put_offset)) {
2142                         skb = (struct sk_buff *) ((unsigned long)
2143                                                   rxdp->Host_Control);
2144                         if (skb == NULL) {
2145                                 DBG_PRINT(ERR_DBG, "%s: The skb is ",
2146                                           dev->name);
2147                                 DBG_PRINT(ERR_DBG, "Null in Rx Intr\n");
2148                                 return;
2149                         }
2150
2151                         pci_unmap_single(nic->pdev, (dma_addr_t)
2152                                          rxdp->Buffer0_ptr,
2153                                          BUF0_LEN, PCI_DMA_FROMDEVICE);
2154                         pci_unmap_single(nic->pdev, (dma_addr_t)
2155                                          rxdp->Buffer1_ptr,
2156                                          BUF1_LEN, PCI_DMA_FROMDEVICE);
2157                         pci_unmap_single(nic->pdev, (dma_addr_t)
2158                                          rxdp->Buffer2_ptr,
2159                                          dev->mtu + BUF0_LEN + 4,
2160                                          PCI_DMA_FROMDEVICE);
2161                         ba = &nic->ba[i][get_block][get_info.offset];
2162
2163                         rx_osm_handler(nic, rxdp, i, ba);
2164
2165                         get_info.offset++;
2166                         mac_control->rx_curr_get_info[i].offset =
2167                             get_info.offset;
2168                         rxdp =
2169                             nic->rx_blocks[i][get_block].block_virt_addr +
2170                             get_info.offset;
2171
2172                         if (get_info.offset &&
2173                             (!(get_info.offset % MAX_RXDS_PER_BLOCK))) {
2174                                 get_info.offset = 0;
2175                                 mac_control->rx_curr_get_info[i].
2176                                     offset = get_info.offset;
2177                                 get_block++;
2178                                 get_block %= nic->block_count[i];
2179                                 mac_control->rx_curr_get_info[i].
2180                                     block_index = get_block;
2181                                 rxdp =
2182                                     nic->rx_blocks[i][get_block].
2183                                     block_virt_addr;
2184                         }
2185                         get_offset =
2186                             (get_block * (MAX_RXDS_PER_BLOCK + 1)) +
2187                             get_info.offset;
2188                         pkt_cnt++;
2189                         if ((indicate_max_pkts)
2190                             && (pkt_cnt > indicate_max_pkts))
2191                                 break;
2192                 }
2193 #endif
2194                 if ((indicate_max_pkts) && (pkt_cnt > indicate_max_pkts))
2195                         break;
2196         }
2197 }
2198 #endif
2199 /**
2200  *  tx_intr_handler - Transmit interrupt handler
2201  *  @nic : device private variable
2202  *  Description:
2203  *  If an interrupt was raised to indicate DMA complete of the
2204  *  Tx packet, this function is called. It identifies the last TxD
2205  *  whose buffer was freed and frees all skbs whose data have already
2206  *  DMA'ed into the NICs internal memory.
2207  *  Return Value:
2208  *  NONE
2209  */
2210
2211 static void tx_intr_handler(struct s2io_nic *nic)
2212 {
2213         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2214         struct net_device *dev = (struct net_device *) nic->dev;
2215         tx_curr_get_info_t get_info, put_info;
2216         struct sk_buff *skb;
2217         TxD_t *txdlp;
2218         register u64 val64 = 0;
2219         int i;
2220         u16 j, frg_cnt;
2221         mac_info_t *mac_control;
2222         struct config_param *config;
2223
2224         mac_control = &nic->mac_control;
2225         config = &nic->config;
2226
2227         /*
2228          * tx_traffic_int reg is an R1 register, hence we read and write
2229          * back the samevalue in the register to clear it.
2230          */
2231         val64 = readq(&bar0->tx_traffic_int);
2232         writeq(val64, &bar0->tx_traffic_int);
2233
2234         for (i = 0; i < config->tx_fifo_num; i++) {
2235                 get_info = mac_control->tx_curr_get_info[i];
2236                 put_info = mac_control->tx_curr_put_info[i];
2237                 txdlp = (TxD_t *) nic->list_info[i][get_info.offset].
2238                     list_virt_addr;
2239                 while ((!(txdlp->Control_1 & TXD_LIST_OWN_XENA)) &&
2240                        (get_info.offset != put_info.offset) &&
2241                        (txdlp->Host_Control)) {
2242                         /* Check for TxD errors */
2243                         if (txdlp->Control_1 & TXD_T_CODE) {
2244                                 unsigned long long err;
2245                                 err = txdlp->Control_1 & TXD_T_CODE;
2246                                 DBG_PRINT(ERR_DBG, "***TxD error %llx\n",
2247                                           err);
2248                         }
2249
2250                         skb = (struct sk_buff *) ((unsigned long)
2251                                                   txdlp->Host_Control);
2252                         if (skb == NULL) {
2253                                 DBG_PRINT(ERR_DBG, "%s: Null skb ",
2254                                           dev->name);
2255                                 DBG_PRINT(ERR_DBG, "in Tx Free Intr\n");
2256                                 return;
2257                         }
2258                         nic->tx_pkt_count++;
2259
2260                         frg_cnt = skb_shinfo(skb)->nr_frags;
2261
2262                         /*  For unfragmented skb */
2263                         pci_unmap_single(nic->pdev, (dma_addr_t)
2264                                          txdlp->Buffer_Pointer,
2265                                          skb->len - skb->data_len,
2266                                          PCI_DMA_TODEVICE);
2267                         if (frg_cnt) {
2268                                 TxD_t *temp = txdlp;
2269                                 txdlp++;
2270                                 for (j = 0; j < frg_cnt; j++, txdlp++) {
2271                                         skb_frag_t *frag =
2272                                             &skb_shinfo(skb)->frags[j];
2273                                         pci_unmap_page(nic->pdev,
2274                                                        (dma_addr_t)
2275                                                        txdlp->
2276                                                        Buffer_Pointer,
2277                                                        frag->size,
2278                                                        PCI_DMA_TODEVICE);
2279                                 }
2280                                 txdlp = temp;
2281                         }
2282                         memset(txdlp, 0,
2283                                (sizeof(TxD_t) * config->max_txds));
2284
2285                         /* Updating the statistics block */
2286                         nic->stats.tx_packets++;
2287                         nic->stats.tx_bytes += skb->len;
2288                         dev_kfree_skb_irq(skb);
2289
2290                         get_info.offset++;
2291                         get_info.offset %= get_info.fifo_len + 1;
2292                         txdlp = (TxD_t *) nic->list_info[i]
2293                             [get_info.offset].list_virt_addr;
2294                         mac_control->tx_curr_get_info[i].offset =
2295                             get_info.offset;
2296                 }
2297         }
2298
2299         spin_lock(&nic->tx_lock);
2300         if (netif_queue_stopped(dev))
2301                 netif_wake_queue(dev);
2302         spin_unlock(&nic->tx_lock);
2303 }
2304
2305 /**
2306  *  alarm_intr_handler - Alarm Interrrupt handler
2307  *  @nic: device private variable
2308  *  Description: If the interrupt was neither because of Rx packet or Tx
2309  *  complete, this function is called. If the interrupt was to indicate
2310  *  a loss of link, the OSM link status handler is invoked for any other
2311  *  alarm interrupt the block that raised the interrupt is displayed
2312  *  and a H/W reset is issued.
2313  *  Return Value:
2314  *  NONE
2315 */
2316
2317 static void alarm_intr_handler(struct s2io_nic *nic)
2318 {
2319         struct net_device *dev = (struct net_device *) nic->dev;
2320         XENA_dev_config_t __iomem *bar0 = nic->bar0;
2321         register u64 val64 = 0, err_reg = 0;
2322
2323         /* Handling link status change error Intr */
2324         err_reg = readq(&bar0->mac_rmac_err_reg);
2325         writeq(err_reg, &bar0->mac_rmac_err_reg);
2326         if (err_reg & RMAC_LINK_STATE_CHANGE_INT) {
2327                 schedule_work(&nic->set_link_task);
2328         }
2329
2330         /* In case of a serious error, the device will be Reset. */
2331         val64 = readq(&bar0->serr_source);
2332         if (val64 & SERR_SOURCE_ANY) {
2333                 DBG_PRINT(ERR_DBG, "%s: Device indicates ", dev->name);
2334                 DBG_PRINT(ERR_DBG, "serious error!!\n");
2335                 netif_stop_queue(dev);
2336                 schedule_work(&nic->rst_timer_task);
2337         }
2338
2339         /*
2340          * Also as mentioned in the latest Errata sheets if the PCC_FB_ECC
2341          * Error occurs, the adapter will be recycled by disabling the
2342          * adapter enable bit and enabling it again after the device
2343          * becomes Quiescent.
2344          */
2345         val64 = readq(&bar0->pcc_err_reg);
2346         writeq(val64, &bar0->pcc_err_reg);
2347         if (val64 & PCC_FB_ECC_DB_ERR) {
2348                 u64 ac = readq(&bar0->adapter_control);
2349                 ac &= ~(ADAPTER_CNTL_EN);
2350                 writeq(ac, &bar0->adapter_control);
2351                 ac = readq(&bar0->adapter_control);
2352                 schedule_work(&nic->set_link_task);
2353         }
2354
2355         /* Other type of interrupts are not being handled now,  TODO */
2356 }
2357
2358 /**
2359  *  wait_for_cmd_complete - waits for a command to complete.
2360  *  @sp : private member of the device structure, which is a pointer to the
2361  *  s2io_nic structure.
2362  *  Description: Function that waits for a command to Write into RMAC
2363  *  ADDR DATA registers to be completed and returns either success or
2364  *  error depending on whether the command was complete or not.
2365  *  Return value:
2366  *   SUCCESS on success and FAILURE on failure.
2367  */
2368
2369 static int wait_for_cmd_complete(nic_t * sp)
2370 {
2371         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2372         int ret = FAILURE, cnt = 0;
2373         u64 val64;
2374
2375         while (TRUE) {
2376                 val64 = readq(&bar0->rmac_addr_cmd_mem);
2377                 if (!(val64 & RMAC_ADDR_CMD_MEM_STROBE_CMD_EXECUTING)) {
2378                         ret = SUCCESS;
2379                         break;
2380                 }
2381                 msleep(50);
2382                 if (cnt++ > 10)
2383                         break;
2384         }
2385
2386         return ret;
2387 }
2388
2389 /**
2390  *  s2io_reset - Resets the card.
2391  *  @sp : private member of the device structure.
2392  *  Description: Function to Reset the card. This function then also
2393  *  restores the previously saved PCI configuration space registers as
2394  *  the card reset also resets the configuration space.
2395  *  Return value:
2396  *  void.
2397  */
2398
2399 static void s2io_reset(nic_t * sp)
2400 {
2401         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2402         u64 val64;
2403         u16 subid;
2404
2405         val64 = SW_RESET_ALL;
2406         writeq(val64, &bar0->sw_reset);
2407
2408         /*
2409          * At this stage, if the PCI write is indeed completed, the
2410          * card is reset and so is the PCI Config space of the device.
2411          * So a read cannot be issued at this stage on any of the
2412          * registers to ensure the write into "sw_reset" register
2413          * has gone through.
2414          * Question: Is there any system call that will explicitly force
2415          * all the write commands still pending on the bus to be pushed
2416          * through?
2417          * As of now I'am just giving a 250ms delay and hoping that the
2418          * PCI write to sw_reset register is done by this time.
2419          */
2420         msleep(250);
2421
2422         /* Restore the PCI state saved during initializarion. */
2423         pci_restore_state(sp->pdev);
2424         s2io_init_pci(sp);
2425
2426         msleep(250);
2427
2428         /* SXE-002: Configure link and activity LED to turn it off */
2429         subid = sp->pdev->subsystem_device;
2430         if ((subid & 0xFF) >= 0x07) {
2431                 val64 = readq(&bar0->gpio_control);
2432                 val64 |= 0x0000800000000000ULL;
2433                 writeq(val64, &bar0->gpio_control);
2434                 val64 = 0x0411040400000000ULL;
2435                 writeq(val64, (void __iomem *) bar0 + 0x2700);
2436         }
2437
2438         sp->device_enabled_once = FALSE;
2439 }
2440
2441 /**
2442  *  s2io_set_swapper - to set the swapper controle on the card
2443  *  @sp : private member of the device structure,
2444  *  pointer to the s2io_nic structure.
2445  *  Description: Function to set the swapper control on the card
2446  *  correctly depending on the 'endianness' of the system.
2447  *  Return value:
2448  *  SUCCESS on success and FAILURE on failure.
2449  */
2450
2451 static int s2io_set_swapper(nic_t * sp)
2452 {
2453         struct net_device *dev = sp->dev;
2454         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2455         u64 val64, valt, valr;
2456
2457         /*
2458          * Set proper endian settings and verify the same by reading
2459          * the PIF Feed-back register.
2460          */
2461
2462         val64 = readq(&bar0->pif_rd_swapper_fb);
2463         if (val64 != 0x0123456789ABCDEFULL) {
2464                 int i = 0;
2465                 u64 value[] = { 0xC30000C3C30000C3ULL,   /* FE=1, SE=1 */
2466                                 0x8100008181000081ULL,  /* FE=1, SE=0 */
2467                                 0x4200004242000042ULL,  /* FE=0, SE=1 */
2468                                 0};                     /* FE=0, SE=0 */
2469
2470                 while(i<4) {
2471                         writeq(value[i], &bar0->swapper_ctrl);
2472                         val64 = readq(&bar0->pif_rd_swapper_fb);
2473                         if (val64 == 0x0123456789ABCDEFULL)
2474                                 break;
2475                         i++;
2476                 }
2477                 if (i == 4) {
2478                         DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
2479                                 dev->name);
2480                         DBG_PRINT(ERR_DBG, "feedback read %llx\n",
2481                                 (unsigned long long) val64);
2482                         return FAILURE;
2483                 }
2484                 valr = value[i];
2485         } else {
2486                 valr = readq(&bar0->swapper_ctrl);
2487         }
2488
2489         valt = 0x0123456789ABCDEFULL;
2490         writeq(valt, &bar0->xmsi_address);
2491         val64 = readq(&bar0->xmsi_address);
2492
2493         if(val64 != valt) {
2494                 int i = 0;
2495                 u64 value[] = { 0x00C3C30000C3C300ULL,  /* FE=1, SE=1 */
2496                                 0x0081810000818100ULL,  /* FE=1, SE=0 */
2497                                 0x0042420000424200ULL,  /* FE=0, SE=1 */
2498                                 0};                     /* FE=0, SE=0 */
2499
2500                 while(i<4) {
2501                         writeq((value[i] | valr), &bar0->swapper_ctrl);
2502                         writeq(valt, &bar0->xmsi_address);
2503                         val64 = readq(&bar0->xmsi_address);
2504                         if(val64 == valt)
2505                                 break;
2506                         i++;
2507                 }
2508                 if(i == 4) {
2509                         DBG_PRINT(ERR_DBG, "Write failed, Xmsi_addr ");
2510                         DBG_PRINT(ERR_DBG, "reads:0x%llx\n",val64);
2511                         return FAILURE;
2512                 }
2513         }
2514         val64 = readq(&bar0->swapper_ctrl);
2515         val64 &= 0xFFFF000000000000ULL;
2516
2517 #ifdef  __BIG_ENDIAN
2518         /*
2519          * The device by default set to a big endian format, so a
2520          * big endian driver need not set anything.
2521          */
2522         val64 |= (SWAPPER_CTRL_TXP_FE |
2523                  SWAPPER_CTRL_TXP_SE |
2524                  SWAPPER_CTRL_TXD_R_FE |
2525                  SWAPPER_CTRL_TXD_W_FE |
2526                  SWAPPER_CTRL_TXF_R_FE |
2527                  SWAPPER_CTRL_RXD_R_FE |
2528                  SWAPPER_CTRL_RXD_W_FE |
2529                  SWAPPER_CTRL_RXF_W_FE |
2530                  SWAPPER_CTRL_XMSI_FE |
2531                  SWAPPER_CTRL_XMSI_SE |
2532                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
2533         writeq(val64, &bar0->swapper_ctrl);
2534 #else
2535         /*
2536          * Initially we enable all bits to make it accessible by the
2537          * driver, then we selectively enable only those bits that
2538          * we want to set.
2539          */
2540         val64 |= (SWAPPER_CTRL_TXP_FE |
2541                  SWAPPER_CTRL_TXP_SE |
2542                  SWAPPER_CTRL_TXD_R_FE |
2543                  SWAPPER_CTRL_TXD_R_SE |
2544                  SWAPPER_CTRL_TXD_W_FE |
2545                  SWAPPER_CTRL_TXD_W_SE |
2546                  SWAPPER_CTRL_TXF_R_FE |
2547                  SWAPPER_CTRL_RXD_R_FE |
2548                  SWAPPER_CTRL_RXD_R_SE |
2549                  SWAPPER_CTRL_RXD_W_FE |
2550                  SWAPPER_CTRL_RXD_W_SE |
2551                  SWAPPER_CTRL_RXF_W_FE |
2552                  SWAPPER_CTRL_XMSI_FE |
2553                  SWAPPER_CTRL_XMSI_SE |
2554                  SWAPPER_CTRL_STATS_FE | SWAPPER_CTRL_STATS_SE);
2555         writeq(val64, &bar0->swapper_ctrl);
2556 #endif
2557         val64 = readq(&bar0->swapper_ctrl);
2558
2559         /*
2560          * Verifying if endian settings are accurate by reading a
2561          * feedback register.
2562          */
2563         val64 = readq(&bar0->pif_rd_swapper_fb);
2564         if (val64 != 0x0123456789ABCDEFULL) {
2565                 /* Endian settings are incorrect, calls for another dekko. */
2566                 DBG_PRINT(ERR_DBG, "%s: Endian settings are wrong, ",
2567                           dev->name);
2568                 DBG_PRINT(ERR_DBG, "feedback read %llx\n",
2569                           (unsigned long long) val64);
2570                 return FAILURE;
2571         }
2572
2573         return SUCCESS;
2574 }
2575
2576 /* ********************************************************* *
2577  * Functions defined below concern the OS part of the driver *
2578  * ********************************************************* */
2579
2580 /**
2581  *  s2io_open - open entry point of the driver
2582  *  @dev : pointer to the device structure.
2583  *  Description:
2584  *  This function is the open entry point of the driver. It mainly calls a
2585  *  function to allocate Rx buffers and inserts them into the buffer
2586  *  descriptors and then enables the Rx part of the NIC.
2587  *  Return value:
2588  *  0 on success and an appropriate (-)ve integer as defined in errno.h
2589  *   file on failure.
2590  */
2591
2592 static int s2io_open(struct net_device *dev)
2593 {
2594         nic_t *sp = dev->priv;
2595         int err = 0;
2596
2597         /*
2598          * Make sure you have link off by default every time
2599          * Nic is initialized
2600          */
2601         netif_carrier_off(dev);
2602         sp->last_link_state = LINK_DOWN;
2603
2604         /* Initialize H/W and enable interrupts */
2605         if (s2io_card_up(sp)) {
2606                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
2607                           dev->name);
2608                 return -ENODEV;
2609         }
2610
2611         /* After proper initialization of H/W, register ISR */
2612         err = request_irq((int) sp->irq, s2io_isr, SA_SHIRQ,
2613                           sp->name, dev);
2614         if (err) {
2615                 s2io_reset(sp);
2616                 DBG_PRINT(ERR_DBG, "%s: ISR registration failed\n",
2617                           dev->name);
2618                 return err;
2619         }
2620
2621         if (s2io_set_mac_addr(dev, dev->dev_addr) == FAILURE) {
2622                 DBG_PRINT(ERR_DBG, "Set Mac Address Failed\n");
2623                 s2io_reset(sp);
2624                 return -ENODEV;
2625         }
2626
2627         netif_start_queue(dev);
2628         return 0;
2629 }
2630
2631 /**
2632  *  s2io_close -close entry point of the driver
2633  *  @dev : device pointer.
2634  *  Description:
2635  *  This is the stop entry point of the driver. It needs to undo exactly
2636  *  whatever was done by the open entry point,thus it's usually referred to
2637  *  as the close function.Among other things this function mainly stops the
2638  *  Rx side of the NIC and frees all the Rx buffers in the Rx rings.
2639  *  Return value:
2640  *  0 on success and an appropriate (-)ve integer as defined in errno.h
2641  *  file on failure.
2642  */
2643
2644 static int s2io_close(struct net_device *dev)
2645 {
2646         nic_t *sp = dev->priv;
2647
2648         flush_scheduled_work();
2649         netif_stop_queue(dev);
2650         /* Reset card, kill tasklet and free Tx and Rx buffers. */
2651         s2io_card_down(sp);
2652
2653         free_irq(dev->irq, dev);
2654         sp->device_close_flag = TRUE;   /* Device is shut down. */
2655         return 0;
2656 }
2657
2658 /**
2659  *  s2io_xmit - Tx entry point of te driver
2660  *  @skb : the socket buffer containing the Tx data.
2661  *  @dev : device pointer.
2662  *  Description :
2663  *  This function is the Tx entry point of the driver. S2IO NIC supports
2664  *  certain protocol assist features on Tx side, namely  CSO, S/G, LSO.
2665  *  NOTE: when device cant queue the pkt,just the trans_start variable will
2666  *  not be upadted.
2667  *  Return value:
2668  *  0 on success & 1 on failure.
2669  */
2670
2671 static int s2io_xmit(struct sk_buff *skb, struct net_device *dev)
2672 {
2673         nic_t *sp = dev->priv;
2674         u16 frg_cnt, frg_len, i, queue, queue_len, put_off, get_off;
2675         register u64 val64;
2676         TxD_t *txdp;
2677         TxFIFO_element_t __iomem *tx_fifo;
2678         unsigned long flags;
2679 #ifdef NETIF_F_TSO
2680         int mss;
2681 #endif
2682         mac_info_t *mac_control;
2683         struct config_param *config;
2684         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2685
2686         mac_control = &sp->mac_control;
2687         config = &sp->config;
2688
2689         DBG_PRINT(TX_DBG, "%s: In S2IO Tx routine\n", dev->name);
2690         spin_lock_irqsave(&sp->tx_lock, flags);
2691
2692         if (atomic_read(&sp->card_state) == CARD_DOWN) {
2693                 DBG_PRINT(ERR_DBG, "%s: Card going down for reset\n",
2694                           dev->name);
2695                 spin_unlock_irqrestore(&sp->tx_lock, flags);
2696                 return 1;
2697         }
2698
2699         queue = 0;
2700         put_off = (u16) mac_control->tx_curr_put_info[queue].offset;
2701         get_off = (u16) mac_control->tx_curr_get_info[queue].offset;
2702         txdp = (TxD_t *) sp->list_info[queue][put_off].list_virt_addr;
2703
2704         queue_len = mac_control->tx_curr_put_info[queue].fifo_len + 1;
2705         /* Avoid "put" pointer going beyond "get" pointer */
2706         if (txdp->Host_Control || (((put_off + 1) % queue_len) == get_off)) {
2707                 DBG_PRINT(ERR_DBG, "Error in xmit, No free TXDs.\n");
2708                 netif_stop_queue(dev);
2709                 dev_kfree_skb(skb);
2710                 spin_unlock_irqrestore(&sp->tx_lock, flags);
2711                 return 0;
2712         }
2713 #ifdef NETIF_F_TSO
2714         mss = skb_shinfo(skb)->tso_size;
2715         if (mss) {
2716                 txdp->Control_1 |= TXD_TCP_LSO_EN;
2717                 txdp->Control_1 |= TXD_TCP_LSO_MSS(mss);
2718         }
2719 #endif
2720
2721         frg_cnt = skb_shinfo(skb)->nr_frags;
2722         frg_len = skb->len - skb->data_len;
2723
2724         txdp->Host_Control = (unsigned long) skb;
2725         txdp->Buffer_Pointer = pci_map_single
2726             (sp->pdev, skb->data, frg_len, PCI_DMA_TODEVICE);
2727         if (skb->ip_summed == CHECKSUM_HW) {
2728                 txdp->Control_2 |=
2729                     (TXD_TX_CKO_IPV4_EN | TXD_TX_CKO_TCP_EN |
2730                      TXD_TX_CKO_UDP_EN);
2731         }
2732
2733         txdp->Control_2 |= config->tx_intr_type;
2734
2735         txdp->Control_1 |= (TXD_BUFFER0_SIZE(frg_len) |
2736                             TXD_GATHER_CODE_FIRST);
2737         txdp->Control_1 |= TXD_LIST_OWN_XENA;
2738
2739         /* For fragmented SKB. */
2740         for (i = 0; i < frg_cnt; i++) {
2741                 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2742                 txdp++;
2743                 txdp->Buffer_Pointer = (u64) pci_map_page
2744                     (sp->pdev, frag->page, frag->page_offset,
2745                      frag->size, PCI_DMA_TODEVICE);
2746                 txdp->Control_1 |= TXD_BUFFER0_SIZE(frag->size);
2747         }
2748         txdp->Control_1 |= TXD_GATHER_CODE_LAST;
2749
2750         tx_fifo = mac_control->tx_FIFO_start[queue];
2751         val64 = sp->list_info[queue][put_off].list_phy_addr;
2752         writeq(val64, &tx_fifo->TxDL_Pointer);
2753
2754         val64 = (TX_FIFO_LAST_TXD_NUM(frg_cnt) | TX_FIFO_FIRST_LIST |
2755                  TX_FIFO_LAST_LIST);
2756 #ifdef NETIF_F_TSO
2757         if (mss)
2758                 val64 |= TX_FIFO_SPECIAL_FUNC;
2759 #endif
2760         writeq(val64, &tx_fifo->List_Control);
2761
2762         /* Perform a PCI read to flush previous writes */
2763         val64 = readq(&bar0->general_int_status);
2764
2765         put_off++;
2766         put_off %= mac_control->tx_curr_put_info[queue].fifo_len + 1;
2767         mac_control->tx_curr_put_info[queue].offset = put_off;
2768
2769         /* Avoid "put" pointer going beyond "get" pointer */
2770         if (((put_off + 1) % queue_len) == get_off) {
2771                 DBG_PRINT(TX_DBG,
2772                           "No free TxDs for xmit, Put: 0x%x Get:0x%x\n",
2773                           put_off, get_off);
2774                 netif_stop_queue(dev);
2775         }
2776
2777         dev->trans_start = jiffies;
2778         spin_unlock_irqrestore(&sp->tx_lock, flags);
2779
2780         return 0;
2781 }
2782
2783 /**
2784  *  s2io_isr - ISR handler of the device .
2785  *  @irq: the irq of the device.
2786  *  @dev_id: a void pointer to the dev structure of the NIC.
2787  *  @pt_regs: pointer to the registers pushed on the stack.
2788  *  Description:  This function is the ISR handler of the device. It
2789  *  identifies the reason for the interrupt and calls the relevant
2790  *  service routines. As a contongency measure, this ISR allocates the
2791  *  recv buffers, if their numbers are below the panic value which is
2792  *  presently set to 25% of the original number of rcv buffers allocated.
2793  *  Return value:
2794  *   IRQ_HANDLED: will be returned if IRQ was handled by this routine
2795  *   IRQ_NONE: will be returned if interrupt is not from our device
2796  */
2797 static irqreturn_t s2io_isr(int irq, void *dev_id, struct pt_regs *regs)
2798 {
2799         struct net_device *dev = (struct net_device *) dev_id;
2800         nic_t *sp = dev->priv;
2801         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2802 #ifndef CONFIG_S2IO_NAPI
2803         int i, ret;
2804 #endif
2805         u64 reason = 0;
2806         mac_info_t *mac_control;
2807         struct config_param *config;
2808
2809         mac_control = &sp->mac_control;
2810         config = &sp->config;
2811
2812         /*
2813          * Identify the cause for interrupt and call the appropriate
2814          * interrupt handler. Causes for the interrupt could be;
2815          * 1. Rx of packet.
2816          * 2. Tx complete.
2817          * 3. Link down.
2818          * 4. Error in any functional blocks of the NIC.
2819          */
2820         reason = readq(&bar0->general_int_status);
2821
2822         if (!reason) {
2823                 /* The interrupt was not raised by Xena. */
2824                 return IRQ_NONE;
2825         }
2826
2827         /* If Intr is because of Tx Traffic */
2828         if (reason & GEN_INTR_TXTRAFFIC) {
2829                 tx_intr_handler(sp);
2830         }
2831
2832         /* If Intr is because of an error */
2833         if (reason & (GEN_ERROR_INTR))
2834                 alarm_intr_handler(sp);
2835
2836 #ifdef CONFIG_S2IO_NAPI
2837         if (reason & GEN_INTR_RXTRAFFIC) {
2838                 if (netif_rx_schedule_prep(dev)) {
2839                         en_dis_able_nic_intrs(sp, RX_TRAFFIC_INTR,
2840                                               DISABLE_INTRS);
2841                         __netif_rx_schedule(dev);
2842                 }
2843         }
2844 #else
2845         /* If Intr is because of Rx Traffic */
2846         if (reason & GEN_INTR_RXTRAFFIC) {
2847                 rx_intr_handler(sp);
2848         }
2849 #endif
2850
2851         /*
2852          * If the Rx buffer count is below the panic threshold then
2853          * reallocate the buffers from the interrupt handler itself,
2854          * else schedule a tasklet to reallocate the buffers.
2855          */
2856 #ifndef CONFIG_S2IO_NAPI
2857         for (i = 0; i < config->rx_ring_num; i++) {
2858                 int rxb_size = atomic_read(&sp->rx_bufs_left[i]);
2859                 int level = rx_buffer_level(sp, rxb_size, i);
2860
2861                 if ((level == PANIC) && (!TASKLET_IN_USE)) {
2862                         DBG_PRINT(INTR_DBG, "%s: Rx BD hit ", dev->name);
2863                         DBG_PRINT(INTR_DBG, "PANIC levels\n");
2864                         if ((ret = fill_rx_buffers(sp, i)) == -ENOMEM) {
2865                                 DBG_PRINT(ERR_DBG, "%s:Out of memory",
2866                                           dev->name);
2867                                 DBG_PRINT(ERR_DBG, " in ISR!!\n");
2868                                 clear_bit(0, (&sp->tasklet_status));
2869                                 return IRQ_HANDLED;
2870                         }
2871                         clear_bit(0, (&sp->tasklet_status));
2872                 } else if (level == LOW) {
2873                         tasklet_schedule(&sp->task);
2874                 }
2875         }
2876 #endif
2877
2878         return IRQ_HANDLED;
2879 }
2880
2881 /**
2882  *  s2io_get_stats - Updates the device statistics structure.
2883  *  @dev : pointer to the device structure.
2884  *  Description:
2885  *  This function updates the device statistics structure in the s2io_nic
2886  *  structure and returns a pointer to the same.
2887  *  Return value:
2888  *  pointer to the updated net_device_stats structure.
2889  */
2890
2891 static struct net_device_stats *s2io_get_stats(struct net_device *dev)
2892 {
2893         nic_t *sp = dev->priv;
2894         mac_info_t *mac_control;
2895         struct config_param *config;
2896
2897         mac_control = &sp->mac_control;
2898         config = &sp->config;
2899
2900         sp->stats.tx_errors = mac_control->stats_info->tmac_any_err_frms;
2901         sp->stats.rx_errors = mac_control->stats_info->rmac_drop_frms;
2902         sp->stats.multicast = mac_control->stats_info->rmac_vld_mcst_frms;
2903         sp->stats.rx_length_errors =
2904             mac_control->stats_info->rmac_long_frms;
2905
2906         return (&sp->stats);
2907 }
2908
2909 /**
2910  *  s2io_set_multicast - entry point for multicast address enable/disable.
2911  *  @dev : pointer to the device structure
2912  *  Description:
2913  *  This function is a driver entry point which gets called by the kernel
2914  *  whenever multicast addresses must be enabled/disabled. This also gets
2915  *  called to set/reset promiscuous mode. Depending on the deivce flag, we
2916  *  determine, if multicast address must be enabled or if promiscuous mode
2917  *  is to be disabled etc.
2918  *  Return value:
2919  *  void.
2920  */
2921
2922 static void s2io_set_multicast(struct net_device *dev)
2923 {
2924         int i, j, prev_cnt;
2925         struct dev_mc_list *mclist;
2926         nic_t *sp = dev->priv;
2927         XENA_dev_config_t __iomem *bar0 = sp->bar0;
2928         u64 val64 = 0, multi_mac = 0x010203040506ULL, mask =
2929             0xfeffffffffffULL;
2930         u64 dis_addr = 0xffffffffffffULL, mac_addr = 0;
2931         void __iomem *add;
2932
2933         if ((dev->flags & IFF_ALLMULTI) && (!sp->m_cast_flg)) {
2934                 /*  Enable all Multicast addresses */
2935                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(multi_mac),
2936                        &bar0->rmac_addr_data0_mem);
2937                 writeq(RMAC_ADDR_DATA1_MEM_MASK(mask),
2938                        &bar0->rmac_addr_data1_mem);
2939                 val64 = RMAC_ADDR_CMD_MEM_WE |
2940                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
2941                     RMAC_ADDR_CMD_MEM_OFFSET(MAC_MC_ALL_MC_ADDR_OFFSET);
2942                 writeq(val64, &bar0->rmac_addr_cmd_mem);
2943                 /* Wait till command completes */
2944                 wait_for_cmd_complete(sp);
2945
2946                 sp->m_cast_flg = 1;
2947                 sp->all_multi_pos = MAC_MC_ALL_MC_ADDR_OFFSET;
2948         } else if ((dev->flags & IFF_ALLMULTI) && (sp->m_cast_flg)) {
2949                 /*  Disable all Multicast addresses */
2950                 writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
2951                        &bar0->rmac_addr_data0_mem);
2952                 val64 = RMAC_ADDR_CMD_MEM_WE |
2953                     RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
2954                     RMAC_ADDR_CMD_MEM_OFFSET(sp->all_multi_pos);
2955                 writeq(val64, &bar0->rmac_addr_cmd_mem);
2956                 /* Wait till command completes */
2957                 wait_for_cmd_complete(sp);
2958
2959                 sp->m_cast_flg = 0;
2960                 sp->all_multi_pos = 0;
2961         }
2962
2963         if ((dev->flags & IFF_PROMISC) && (!sp->promisc_flg)) {
2964                 /*  Put the NIC into promiscuous mode */
2965                 add = &bar0->mac_cfg;
2966                 val64 = readq(&bar0->mac_cfg);
2967                 val64 |= MAC_CFG_RMAC_PROM_ENABLE;
2968
2969                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
2970                 writel((u32) val64, add);
2971                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
2972                 writel((u32) (val64 >> 32), (add + 4));
2973
2974                 val64 = readq(&bar0->mac_cfg);
2975                 sp->promisc_flg = 1;
2976                 DBG_PRINT(ERR_DBG, "%s: entered promiscuous mode\n",
2977                           dev->name);
2978         } else if (!(dev->flags & IFF_PROMISC) && (sp->promisc_flg)) {
2979                 /*  Remove the NIC from promiscuous mode */
2980                 add = &bar0->mac_cfg;
2981                 val64 = readq(&bar0->mac_cfg);
2982                 val64 &= ~MAC_CFG_RMAC_PROM_ENABLE;
2983
2984                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
2985                 writel((u32) val64, add);
2986                 writeq(RMAC_CFG_KEY(0x4C0D), &bar0->rmac_cfg_key);
2987                 writel((u32) (val64 >> 32), (add + 4));
2988
2989                 val64 = readq(&bar0->mac_cfg);
2990                 sp->promisc_flg = 0;
2991                 DBG_PRINT(ERR_DBG, "%s: left promiscuous mode\n",
2992                           dev->name);
2993         }
2994
2995         /*  Update individual M_CAST address list */
2996         if ((!sp->m_cast_flg) && dev->mc_count) {
2997                 if (dev->mc_count >
2998                     (MAX_ADDRS_SUPPORTED - MAC_MC_ADDR_START_OFFSET - 1)) {
2999                         DBG_PRINT(ERR_DBG, "%s: No more Rx filters ",
3000                                   dev->name);
3001                         DBG_PRINT(ERR_DBG, "can be added, please enable ");
3002                         DBG_PRINT(ERR_DBG, "ALL_MULTI instead\n");
3003                         return;
3004                 }
3005
3006                 prev_cnt = sp->mc_addr_count;
3007                 sp->mc_addr_count = dev->mc_count;
3008
3009                 /* Clear out the previous list of Mc in the H/W. */
3010                 for (i = 0; i < prev_cnt; i++) {
3011                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(dis_addr),
3012                                &bar0->rmac_addr_data0_mem);
3013                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
3014                                 &bar0->rmac_addr_data1_mem);
3015                         val64 = RMAC_ADDR_CMD_MEM_WE |
3016                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
3017                             RMAC_ADDR_CMD_MEM_OFFSET
3018                             (MAC_MC_ADDR_START_OFFSET + i);
3019                         writeq(val64, &bar0->rmac_addr_cmd_mem);
3020
3021                         /* Wait for command completes */
3022                         if (wait_for_cmd_complete(sp)) {
3023                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
3024                                           dev->name);
3025                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
3026                                 return;
3027                         }
3028                 }
3029
3030                 /* Create the new Rx filter list and update the same in H/W. */
3031                 for (i = 0, mclist = dev->mc_list; i < dev->mc_count;
3032                      i++, mclist = mclist->next) {
3033                         memcpy(sp->usr_addrs[i].addr, mclist->dmi_addr,
3034                                ETH_ALEN);
3035                         for (j = 0; j < ETH_ALEN; j++) {
3036                                 mac_addr |= mclist->dmi_addr[j];
3037                                 mac_addr <<= 8;
3038                         }
3039                         mac_addr >>= 8;
3040                         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
3041                                &bar0->rmac_addr_data0_mem);
3042                         writeq(RMAC_ADDR_DATA1_MEM_MASK(0ULL),
3043                                 &bar0->rmac_addr_data1_mem);
3044
3045                         val64 = RMAC_ADDR_CMD_MEM_WE |
3046                             RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
3047                             RMAC_ADDR_CMD_MEM_OFFSET
3048                             (i + MAC_MC_ADDR_START_OFFSET);
3049                         writeq(val64, &bar0->rmac_addr_cmd_mem);
3050
3051                         /* Wait for command completes */
3052                         if (wait_for_cmd_complete(sp)) {
3053                                 DBG_PRINT(ERR_DBG, "%s: Adding ",
3054                                           dev->name);
3055                                 DBG_PRINT(ERR_DBG, "Multicasts failed\n");
3056                                 return;
3057                         }
3058                 }
3059         }
3060 }
3061
3062 /**
3063  *  s2io_set_mac_addr - Programs the Xframe mac address
3064  *  @dev : pointer to the device structure.
3065  *  @addr: a uchar pointer to the new mac address which is to be set.
3066  *  Description : This procedure will program the Xframe to receive
3067  *  frames with new Mac Address
3068  *  Return value: SUCCESS on success and an appropriate (-)ve integer
3069  *  as defined in errno.h file on failure.
3070  */
3071
3072 int s2io_set_mac_addr(struct net_device *dev, u8 * addr)
3073 {
3074         nic_t *sp = dev->priv;
3075         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3076         register u64 val64, mac_addr = 0;
3077         int i;
3078
3079         /*
3080          * Set the new MAC address as the new unicast filter and reflect this
3081          * change on the device address registered with the OS. It will be
3082          * at offset 0.
3083          */
3084         for (i = 0; i < ETH_ALEN; i++) {
3085                 mac_addr <<= 8;
3086                 mac_addr |= addr[i];
3087         }
3088
3089         writeq(RMAC_ADDR_DATA0_MEM_ADDR(mac_addr),
3090                &bar0->rmac_addr_data0_mem);
3091
3092         val64 =
3093             RMAC_ADDR_CMD_MEM_WE | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
3094             RMAC_ADDR_CMD_MEM_OFFSET(0);
3095         writeq(val64, &bar0->rmac_addr_cmd_mem);
3096         /* Wait till command completes */
3097         if (wait_for_cmd_complete(sp)) {
3098                 DBG_PRINT(ERR_DBG, "%s: set_mac_addr failed\n", dev->name);
3099                 return FAILURE;
3100         }
3101
3102         return SUCCESS;
3103 }
3104
3105 /**
3106  * s2io_ethtool_sset - Sets different link parameters.
3107  * @sp : private member of the device structure, which is a pointer to the  * s2io_nic structure.
3108  * @info: pointer to the structure with parameters given by ethtool to set
3109  * link information.
3110  * Description:
3111  * The function sets different link parameters provided by the user onto
3112  * the NIC.
3113  * Return value:
3114  * 0 on success.
3115 */
3116
3117 static int s2io_ethtool_sset(struct net_device *dev,
3118                              struct ethtool_cmd *info)
3119 {
3120         nic_t *sp = dev->priv;
3121         if ((info->autoneg == AUTONEG_ENABLE) ||
3122             (info->speed != SPEED_10000) || (info->duplex != DUPLEX_FULL))
3123                 return -EINVAL;
3124         else {
3125                 s2io_close(sp->dev);
3126                 s2io_open(sp->dev);
3127         }
3128
3129         return 0;
3130 }
3131
3132 /**
3133  * s2io_ethtol_gset - Return link specific information.
3134  * @sp : private member of the device structure, pointer to the
3135  *      s2io_nic structure.
3136  * @info : pointer to the structure with parameters given by ethtool
3137  * to return link information.
3138  * Description:
3139  * Returns link specific information like speed, duplex etc.. to ethtool.
3140  * Return value :
3141  * return 0 on success.
3142  */
3143
3144 static int s2io_ethtool_gset(struct net_device *dev, struct ethtool_cmd *info)
3145 {
3146         nic_t *sp = dev->priv;
3147         info->supported = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
3148         info->advertising = (SUPPORTED_10000baseT_Full | SUPPORTED_FIBRE);
3149         info->port = PORT_FIBRE;
3150         /* info->transceiver?? TODO */
3151
3152         if (netif_carrier_ok(sp->dev)) {
3153                 info->speed = 10000;
3154                 info->duplex = DUPLEX_FULL;
3155         } else {
3156                 info->speed = -1;
3157                 info->duplex = -1;
3158         }
3159
3160         info->autoneg = AUTONEG_DISABLE;
3161         return 0;
3162 }
3163
3164 /**
3165  * s2io_ethtool_gdrvinfo - Returns driver specific information.
3166  * @sp : private member of the device structure, which is a pointer to the
3167  * s2io_nic structure.
3168  * @info : pointer to the structure with parameters given by ethtool to
3169  * return driver information.
3170  * Description:
3171  * Returns driver specefic information like name, version etc.. to ethtool.
3172  * Return value:
3173  *  void
3174  */
3175
3176 static void s2io_ethtool_gdrvinfo(struct net_device *dev,
3177                                   struct ethtool_drvinfo *info)
3178 {
3179         nic_t *sp = dev->priv;
3180
3181         strncpy(info->driver, s2io_driver_name, sizeof(s2io_driver_name));
3182         strncpy(info->version, s2io_driver_version,
3183                 sizeof(s2io_driver_version));
3184         strncpy(info->fw_version, "", 32);
3185         strncpy(info->bus_info, pci_name(sp->pdev), 32);
3186         info->regdump_len = XENA_REG_SPACE;
3187         info->eedump_len = XENA_EEPROM_SPACE;
3188         info->testinfo_len = S2IO_TEST_LEN;
3189         info->n_stats = S2IO_STAT_LEN;
3190 }
3191
3192 /**
3193  *  s2io_ethtool_gregs - dumps the entire space of Xfame into the buffer.
3194  *  @sp: private member of the device structure, which is a pointer to the
3195  *  s2io_nic structure.
3196  *  @regs : pointer to the structure with parameters given by ethtool for
3197  *  dumping the registers.
3198  *  @reg_space: The input argumnet into which all the registers are dumped.
3199  *  Description:
3200  *  Dumps the entire register space of xFrame NIC into the user given
3201  *  buffer area.
3202  * Return value :
3203  * void .
3204 */
3205
3206 static void s2io_ethtool_gregs(struct net_device *dev,
3207                                struct ethtool_regs *regs, void *space)
3208 {
3209         int i;
3210         u64 reg;
3211         u8 *reg_space = (u8 *) space;
3212         nic_t *sp = dev->priv;
3213
3214         regs->len = XENA_REG_SPACE;
3215         regs->version = sp->pdev->subsystem_device;
3216
3217         for (i = 0; i < regs->len; i += 8) {
3218                 reg = readq(sp->bar0 + i);
3219                 memcpy((reg_space + i), &reg, 8);
3220         }
3221 }
3222
3223 /**
3224  *  s2io_phy_id  - timer function that alternates adapter LED.
3225  *  @data : address of the private member of the device structure, which
3226  *  is a pointer to the s2io_nic structure, provided as an u32.
3227  * Description: This is actually the timer function that alternates the
3228  * adapter LED bit of the adapter control bit to set/reset every time on
3229  * invocation. The timer is set for 1/2 a second, hence tha NIC blinks
3230  *  once every second.
3231 */
3232 static void s2io_phy_id(unsigned long data)
3233 {
3234         nic_t *sp = (nic_t *) data;
3235         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3236         u64 val64 = 0;
3237         u16 subid;
3238
3239         subid = sp->pdev->subsystem_device;
3240         if ((subid & 0xFF) >= 0x07) {
3241                 val64 = readq(&bar0->gpio_control);
3242                 val64 ^= GPIO_CTRL_GPIO_0;
3243                 writeq(val64, &bar0->gpio_control);
3244         } else {
3245                 val64 = readq(&bar0->adapter_control);
3246                 val64 ^= ADAPTER_LED_ON;
3247                 writeq(val64, &bar0->adapter_control);
3248         }
3249
3250         mod_timer(&sp->id_timer, jiffies + HZ / 2);
3251 }
3252
3253 /**
3254  * s2io_ethtool_idnic - To physically identify the nic on the system.
3255  * @sp : private member of the device structure, which is a pointer to the
3256  * s2io_nic structure.
3257  * @id : pointer to the structure with identification parameters given by
3258  * ethtool.
3259  * Description: Used to physically identify the NIC on the system.
3260  * The Link LED will blink for a time specified by the user for
3261  * identification.
3262  * NOTE: The Link has to be Up to be able to blink the LED. Hence
3263  * identification is possible only if it's link is up.
3264  * Return value:
3265  * int , returns 0 on success
3266  */
3267
3268 static int s2io_ethtool_idnic(struct net_device *dev, u32 data)
3269 {
3270         u64 val64 = 0, last_gpio_ctrl_val;
3271         nic_t *sp = dev->priv;
3272         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3273         u16 subid;
3274
3275         subid = sp->pdev->subsystem_device;
3276         last_gpio_ctrl_val = readq(&bar0->gpio_control);
3277         if ((subid & 0xFF) < 0x07) {
3278                 val64 = readq(&bar0->adapter_control);
3279                 if (!(val64 & ADAPTER_CNTL_EN)) {
3280                         printk(KERN_ERR
3281                                "Adapter Link down, cannot blink LED\n");
3282                         return -EFAULT;
3283                 }
3284         }
3285         if (sp->id_timer.function == NULL) {
3286                 init_timer(&sp->id_timer);
3287                 sp->id_timer.function = s2io_phy_id;
3288                 sp->id_timer.data = (unsigned long) sp;
3289         }
3290         mod_timer(&sp->id_timer, jiffies);
3291         if (data)
3292                 msleep(data * 1000);
3293         else
3294                 msleep(0xFFFFFFFF);
3295         del_timer_sync(&sp->id_timer);
3296
3297         if (CARDS_WITH_FAULTY_LINK_INDICATORS(subid)) {
3298                 writeq(last_gpio_ctrl_val, &bar0->gpio_control);
3299                 last_gpio_ctrl_val = readq(&bar0->gpio_control);
3300         }
3301
3302         return 0;
3303 }
3304
3305 /**
3306  * s2io_ethtool_getpause_data -Pause frame frame generation and reception.
3307  * @sp : private member of the device structure, which is a pointer to the  * s2io_nic structure.
3308  * @ep : pointer to the structure with pause parameters given by ethtool.
3309  * Description:
3310  * Returns the Pause frame generation and reception capability of the NIC.
3311  * Return value:
3312  *  void
3313  */
3314 static void s2io_ethtool_getpause_data(struct net_device *dev,
3315                                        struct ethtool_pauseparam *ep)
3316 {
3317         u64 val64;
3318         nic_t *sp = dev->priv;
3319         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3320
3321         val64 = readq(&bar0->rmac_pause_cfg);
3322         if (val64 & RMAC_PAUSE_GEN_ENABLE)
3323                 ep->tx_pause = TRUE;
3324         if (val64 & RMAC_PAUSE_RX_ENABLE)
3325                 ep->rx_pause = TRUE;
3326         ep->autoneg = FALSE;
3327 }
3328
3329 /**
3330  * s2io_ethtool_setpause_data -  set/reset pause frame generation.
3331  * @sp : private member of the device structure, which is a pointer to the
3332  *      s2io_nic structure.
3333  * @ep : pointer to the structure with pause parameters given by ethtool.
3334  * Description:
3335  * It can be used to set or reset Pause frame generation or reception
3336  * support of the NIC.
3337  * Return value:
3338  * int, returns 0 on Success
3339  */
3340
3341 static int s2io_ethtool_setpause_data(struct net_device *dev,
3342                                       struct ethtool_pauseparam *ep)
3343 {
3344         u64 val64;
3345         nic_t *sp = dev->priv;
3346         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3347
3348         val64 = readq(&bar0->rmac_pause_cfg);
3349         if (ep->tx_pause)
3350                 val64 |= RMAC_PAUSE_GEN_ENABLE;
3351         else
3352                 val64 &= ~RMAC_PAUSE_GEN_ENABLE;
3353         if (ep->rx_pause)
3354                 val64 |= RMAC_PAUSE_RX_ENABLE;
3355         else
3356                 val64 &= ~RMAC_PAUSE_RX_ENABLE;
3357         writeq(val64, &bar0->rmac_pause_cfg);
3358         return 0;
3359 }
3360
3361 /**
3362  * read_eeprom - reads 4 bytes of data from user given offset.
3363  * @sp : private member of the device structure, which is a pointer to the
3364  *      s2io_nic structure.
3365  * @off : offset at which the data must be written
3366  * @data : Its an output parameter where the data read at the given
3367  *      offset is stored.
3368  * Description:
3369  * Will read 4 bytes of data from the user given offset and return the
3370  * read data.
3371  * NOTE: Will allow to read only part of the EEPROM visible through the
3372  *   I2C bus.
3373  * Return value:
3374  *  -1 on failure and 0 on success.
3375  */
3376
3377 #define S2IO_DEV_ID             5
3378 static int read_eeprom(nic_t * sp, int off, u32 * data)
3379 {
3380         int ret = -1;
3381         u32 exit_cnt = 0;
3382         u64 val64;
3383         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3384
3385         val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
3386             I2C_CONTROL_BYTE_CNT(0x3) | I2C_CONTROL_READ |
3387             I2C_CONTROL_CNTL_START;
3388         SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
3389
3390         while (exit_cnt < 5) {
3391                 val64 = readq(&bar0->i2c_control);
3392                 if (I2C_CONTROL_CNTL_END(val64)) {
3393                         *data = I2C_CONTROL_GET_DATA(val64);
3394                         ret = 0;
3395                         break;
3396                 }
3397                 msleep(50);
3398                 exit_cnt++;
3399         }
3400
3401         return ret;
3402 }
3403
3404 /**
3405  *  write_eeprom - actually writes the relevant part of the data value.
3406  *  @sp : private member of the device structure, which is a pointer to the
3407  *       s2io_nic structure.
3408  *  @off : offset at which the data must be written
3409  *  @data : The data that is to be written
3410  *  @cnt : Number of bytes of the data that are actually to be written into
3411  *  the Eeprom. (max of 3)
3412  * Description:
3413  *  Actually writes the relevant part of the data value into the Eeprom
3414  *  through the I2C bus.
3415  * Return value:
3416  *  0 on success, -1 on failure.
3417  */
3418
3419 static int write_eeprom(nic_t * sp, int off, u32 data, int cnt)
3420 {
3421         int exit_cnt = 0, ret = -1;
3422         u64 val64;
3423         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3424
3425         val64 = I2C_CONTROL_DEV_ID(S2IO_DEV_ID) | I2C_CONTROL_ADDR(off) |
3426             I2C_CONTROL_BYTE_CNT(cnt) | I2C_CONTROL_SET_DATA(data) |
3427             I2C_CONTROL_CNTL_START;
3428         SPECIAL_REG_WRITE(val64, &bar0->i2c_control, LF);
3429
3430         while (exit_cnt < 5) {
3431                 val64 = readq(&bar0->i2c_control);
3432                 if (I2C_CONTROL_CNTL_END(val64)) {
3433                         if (!(val64 & I2C_CONTROL_NACK))
3434                                 ret = 0;
3435                         break;
3436                 }
3437                 msleep(50);
3438                 exit_cnt++;
3439         }
3440
3441         return ret;
3442 }
3443
3444 /**
3445  *  s2io_ethtool_geeprom  - reads the value stored in the Eeprom.
3446  *  @sp : private member of the device structure, which is a pointer to the *       s2io_nic structure.
3447  *  @eeprom : pointer to the user level structure provided by ethtool,
3448  *  containing all relevant information.
3449  *  @data_buf : user defined value to be written into Eeprom.
3450  *  Description: Reads the values stored in the Eeprom at given offset
3451  *  for a given length. Stores these values int the input argument data
3452  *  buffer 'data_buf' and returns these to the caller (ethtool.)
3453  *  Return value:
3454  *  int  0 on success
3455  */
3456
3457 static int s2io_ethtool_geeprom(struct net_device *dev,
3458                                 struct ethtool_eeprom *eeprom, u8 * data_buf)
3459 {
3460         u32 data, i, valid;
3461         nic_t *sp = dev->priv;
3462
3463         eeprom->magic = sp->pdev->vendor | (sp->pdev->device << 16);
3464
3465         if ((eeprom->offset + eeprom->len) > (XENA_EEPROM_SPACE))
3466                 eeprom->len = XENA_EEPROM_SPACE - eeprom->offset;
3467
3468         for (i = 0; i < eeprom->len; i += 4) {
3469                 if (read_eeprom(sp, (eeprom->offset + i), &data)) {
3470                         DBG_PRINT(ERR_DBG, "Read of EEPROM failed\n");
3471                         return -EFAULT;
3472                 }
3473                 valid = INV(data);
3474                 memcpy((data_buf + i), &valid, 4);
3475         }
3476         return 0;
3477 }
3478
3479 /**
3480  *  s2io_ethtool_seeprom - tries to write the user provided value in Eeprom
3481  *  @sp : private member of the device structure, which is a pointer to the
3482  *  s2io_nic structure.
3483  *  @eeprom : pointer to the user level structure provided by ethtool,
3484  *  containing all relevant information.
3485  *  @data_buf ; user defined value to be written into Eeprom.
3486  *  Description:
3487  *  Tries to write the user provided value in the Eeprom, at the offset
3488  *  given by the user.
3489  *  Return value:
3490  *  0 on success, -EFAULT on failure.
3491  */
3492
3493 static int s2io_ethtool_seeprom(struct net_device *dev,
3494                                 struct ethtool_eeprom *eeprom,
3495                                 u8 * data_buf)
3496 {
3497         int len = eeprom->len, cnt = 0;
3498         u32 valid = 0, data;
3499         nic_t *sp = dev->priv;
3500
3501         if (eeprom->magic != (sp->pdev->vendor | (sp->pdev->device << 16))) {
3502                 DBG_PRINT(ERR_DBG,
3503                           "ETHTOOL_WRITE_EEPROM Err: Magic value ");
3504                 DBG_PRINT(ERR_DBG, "is wrong, Its not 0x%x\n",
3505                           eeprom->magic);
3506                 return -EFAULT;
3507         }
3508
3509         while (len) {
3510                 data = (u32) data_buf[cnt] & 0x000000FF;
3511                 if (data) {
3512                         valid = (u32) (data << 24);
3513                 } else
3514                         valid = data;
3515
3516                 if (write_eeprom(sp, (eeprom->offset + cnt), valid, 0)) {
3517                         DBG_PRINT(ERR_DBG,
3518                                   "ETHTOOL_WRITE_EEPROM Err: Cannot ");
3519                         DBG_PRINT(ERR_DBG,
3520                                   "write into the specified offset\n");
3521                         return -EFAULT;
3522                 }
3523                 cnt++;
3524                 len--;
3525         }
3526
3527         return 0;
3528 }
3529
3530 /**
3531  * s2io_register_test - reads and writes into all clock domains.
3532  * @sp : private member of the device structure, which is a pointer to the
3533  * s2io_nic structure.
3534  * @data : variable that returns the result of each of the test conducted b
3535  * by the driver.
3536  * Description:
3537  * Read and write into all clock domains. The NIC has 3 clock domains,
3538  * see that registers in all the three regions are accessible.
3539  * Return value:
3540  * 0 on success.
3541  */
3542
3543 static int s2io_register_test(nic_t * sp, uint64_t * data)
3544 {
3545         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3546         u64 val64 = 0;
3547         int fail = 0;
3548
3549         val64 = readq(&bar0->pcc_enable);
3550         if (val64 != 0xff00000000000000ULL) {
3551                 fail = 1;
3552                 DBG_PRINT(INFO_DBG, "Read Test level 1 fails\n");
3553         }
3554
3555         val64 = readq(&bar0->rmac_pause_cfg);
3556         if (val64 != 0xc000ffff00000000ULL) {
3557                 fail = 1;
3558                 DBG_PRINT(INFO_DBG, "Read Test level 2 fails\n");
3559         }
3560
3561         val64 = readq(&bar0->rx_queue_cfg);
3562         if (val64 != 0x0808080808080808ULL) {
3563                 fail = 1;
3564                 DBG_PRINT(INFO_DBG, "Read Test level 3 fails\n");
3565         }
3566
3567         val64 = readq(&bar0->xgxs_efifo_cfg);
3568         if (val64 != 0x000000001923141EULL) {
3569                 fail = 1;
3570                 DBG_PRINT(INFO_DBG, "Read Test level 4 fails\n");
3571         }
3572
3573         val64 = 0x5A5A5A5A5A5A5A5AULL;
3574         writeq(val64, &bar0->xmsi_data);
3575         val64 = readq(&bar0->xmsi_data);
3576         if (val64 != 0x5A5A5A5A5A5A5A5AULL) {
3577                 fail = 1;
3578                 DBG_PRINT(ERR_DBG, "Write Test level 1 fails\n");
3579         }
3580
3581         val64 = 0xA5A5A5A5A5A5A5A5ULL;
3582         writeq(val64, &bar0->xmsi_data);
3583         val64 = readq(&bar0->xmsi_data);
3584         if (val64 != 0xA5A5A5A5A5A5A5A5ULL) {
3585                 fail = 1;
3586                 DBG_PRINT(ERR_DBG, "Write Test level 2 fails\n");
3587         }
3588
3589         *data = fail;
3590         return 0;
3591 }
3592
3593 /**
3594  * s2io_eeprom_test - to verify that EEprom in the xena can be programmed.
3595  * @sp : private member of the device structure, which is a pointer to the
3596  * s2io_nic structure.
3597  * @data:variable that returns the result of each of the test conducted by
3598  * the driver.
3599  * Description:
3600  * Verify that EEPROM in the xena can be programmed using I2C_CONTROL
3601  * register.
3602  * Return value:
3603  * 0 on success.
3604  */
3605
3606 static int s2io_eeprom_test(nic_t * sp, uint64_t * data)
3607 {
3608         int fail = 0;
3609         u32 ret_data;
3610
3611         /* Test Write Error at offset 0 */
3612         if (!write_eeprom(sp, 0, 0, 3))
3613                 fail = 1;
3614
3615         /* Test Write at offset 4f0 */
3616         if (write_eeprom(sp, 0x4F0, 0x01234567, 3))
3617                 fail = 1;
3618         if (read_eeprom(sp, 0x4F0, &ret_data))
3619                 fail = 1;
3620
3621         if (ret_data != 0x01234567)
3622                 fail = 1;
3623
3624         /* Reset the EEPROM data go FFFF */
3625         write_eeprom(sp, 0x4F0, 0xFFFFFFFF, 3);
3626
3627         /* Test Write Request Error at offset 0x7c */
3628         if (!write_eeprom(sp, 0x07C, 0, 3))
3629                 fail = 1;
3630
3631         /* Test Write Request at offset 0x7fc */
3632         if (write_eeprom(sp, 0x7FC, 0x01234567, 3))
3633                 fail = 1;
3634         if (read_eeprom(sp, 0x7FC, &ret_data))
3635                 fail = 1;
3636
3637         if (ret_data != 0x01234567)
3638                 fail = 1;
3639
3640         /* Reset the EEPROM data go FFFF */
3641         write_eeprom(sp, 0x7FC, 0xFFFFFFFF, 3);
3642
3643         /* Test Write Error at offset 0x80 */
3644         if (!write_eeprom(sp, 0x080, 0, 3))
3645                 fail = 1;
3646
3647         /* Test Write Error at offset 0xfc */
3648         if (!write_eeprom(sp, 0x0FC, 0, 3))
3649                 fail = 1;
3650
3651         /* Test Write Error at offset 0x100 */
3652         if (!write_eeprom(sp, 0x100, 0, 3))
3653                 fail = 1;
3654
3655         /* Test Write Error at offset 4ec */
3656         if (!write_eeprom(sp, 0x4EC, 0, 3))
3657                 fail = 1;
3658
3659         *data = fail;
3660         return 0;
3661 }
3662
3663 /**
3664  * s2io_bist_test - invokes the MemBist test of the card .
3665  * @sp : private member of the device structure, which is a pointer to the
3666  * s2io_nic structure.
3667  * @data:variable that returns the result of each of the test conducted by
3668  * the driver.
3669  * Description:
3670  * This invokes the MemBist test of the card. We give around
3671  * 2 secs time for the Test to complete. If it's still not complete
3672  * within this peiod, we consider that the test failed.
3673  * Return value:
3674  * 0 on success and -1 on failure.
3675  */
3676
3677 static int s2io_bist_test(nic_t * sp, uint64_t * data)
3678 {
3679         u8 bist = 0;
3680         int cnt = 0, ret = -1;
3681
3682         pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
3683         bist |= PCI_BIST_START;
3684         pci_write_config_word(sp->pdev, PCI_BIST, bist);
3685
3686         while (cnt < 20) {
3687                 pci_read_config_byte(sp->pdev, PCI_BIST, &bist);
3688                 if (!(bist & PCI_BIST_START)) {
3689                         *data = (bist & PCI_BIST_CODE_MASK);
3690                         ret = 0;
3691                         break;
3692                 }
3693                 msleep(100);
3694                 cnt++;
3695         }
3696
3697         return ret;
3698 }
3699
3700 /**
3701  * s2io-link_test - verifies the link state of the nic
3702  * @sp ; private member of the device structure, which is a pointer to the
3703  * s2io_nic structure.
3704  * @data: variable that returns the result of each of the test conducted by
3705  * the driver.
3706  * Description:
3707  * The function verifies the link state of the NIC and updates the input
3708  * argument 'data' appropriately.
3709  * Return value:
3710  * 0 on success.
3711  */
3712
3713 static int s2io_link_test(nic_t * sp, uint64_t * data)
3714 {
3715         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3716         u64 val64;
3717
3718         val64 = readq(&bar0->adapter_status);
3719         if (val64 & ADAPTER_STATUS_RMAC_LOCAL_FAULT)
3720                 *data = 1;
3721
3722         return 0;
3723 }
3724
3725 /**
3726  * s2io_rldram_test - offline test for access to the RldRam chip on the NIC
3727  * @sp - private member of the device structure, which is a pointer to the
3728  * s2io_nic structure.
3729  * @data - variable that returns the result of each of the test
3730  * conducted by the driver.
3731  * Description:
3732  *  This is one of the offline test that tests the read and write
3733  *  access to the RldRam chip on the NIC.
3734  * Return value:
3735  *  0 on success.
3736  */
3737
3738 static int s2io_rldram_test(nic_t * sp, uint64_t * data)
3739 {
3740         XENA_dev_config_t __iomem *bar0 = sp->bar0;
3741         u64 val64;
3742         int cnt, iteration = 0, test_pass = 0;
3743
3744         val64 = readq(&bar0->adapter_control);
3745         val64 &= ~ADAPTER_ECC_EN;
3746         writeq(val64, &bar0->adapter_control);
3747
3748         val64 = readq(&bar0->mc_rldram_test_ctrl);
3749         val64 |= MC_RLDRAM_TEST_MODE;
3750         writeq(val64, &bar0->mc_rldram_test_ctrl);
3751
3752         val64 = readq(&bar0->mc_rldram_mrs);
3753         val64 |= MC_RLDRAM_QUEUE_SIZE_ENABLE;
3754         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
3755
3756         val64 |= MC_RLDRAM_MRS_ENABLE;
3757         SPECIAL_REG_WRITE(val64, &bar0->mc_rldram_mrs, UF);
3758
3759         while (iteration < 2) {
3760                 val64 = 0x55555555aaaa0000ULL;
3761                 if (iteration == 1) {
3762                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
3763                 }
3764                 writeq(val64, &bar0->mc_rldram_test_d0);
3765
3766                 val64 = 0xaaaa5a5555550000ULL;
3767                 if (iteration == 1) {
3768                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
3769                 }
3770                 writeq(val64, &bar0->mc_rldram_test_d1);
3771
3772                 val64 = 0x55aaaaaaaa5a0000ULL;
3773                 if (iteration == 1) {
3774                         val64 ^= 0xFFFFFFFFFFFF0000ULL;
3775                 }
3776                 writeq(val64, &bar0->mc_rldram_test_d2);
3777
3778                 val64 = (u64) (0x0000003fffff0000ULL);
3779                 writeq(val64, &bar0->mc_rldram_test_add);
3780
3781
3782                 val64 = MC_RLDRAM_TEST_MODE;
3783                 writeq(val64, &bar0->mc_rldram_test_ctrl);
3784
3785                 val64 |=
3786                     MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_WRITE |
3787                     MC_RLDRAM_TEST_GO;
3788                 writeq(val64, &bar0->mc_rldram_test_ctrl);
3789
3790                 for (cnt = 0; cnt < 5; cnt++) {
3791                         val64 = readq(&bar0->mc_rldram_test_ctrl);
3792                         if (val64 & MC_RLDRAM_TEST_DONE)
3793                                 break;
3794                         msleep(200);
3795                 }
3796
3797                 if (cnt == 5)
3798                         break;
3799
3800                 val64 = MC_RLDRAM_TEST_MODE;
3801                 writeq(val64, &bar0->mc_rldram_test_ctrl);
3802
3803                 val64 |= MC_RLDRAM_TEST_MODE | MC_RLDRAM_TEST_GO;
3804                 writeq(val64, &bar0->mc_rldram_test_ctrl);
3805
3806                 for (cnt = 0; cnt < 5; cnt++) {
3807                         val64 = readq(&bar0->mc_rldram_test_ctrl);
3808                         if (val64 & MC_RLDRAM_TEST_DONE)
3809                                 break;
3810                         msleep(500);
3811                 }
3812
3813                 if (cnt == 5)
3814                         break;
3815
3816                 val64 = readq(&bar0->mc_rldram_test_ctrl);
3817                 if (val64 & MC_RLDRAM_TEST_PASS)
3818                         test_pass = 1;
3819
3820                 iteration++;
3821         }
3822
3823         if (!test_pass)
3824                 *data = 1;
3825         else
3826                 *data = 0;
3827
3828         return 0;
3829 }
3830
3831 /**
3832  *  s2io_ethtool_test - conducts 6 tsets to determine the health of card.
3833  *  @sp : private member of the device structure, which is a pointer to the
3834  *  s2io_nic structure.
3835  *  @ethtest : pointer to a ethtool command specific structure that will be
3836  *  returned to the user.
3837  *  @data : variable that returns the result of each of the test
3838  * conducted by the driver.
3839  * Description:
3840  *  This function conducts 6 tests ( 4 offline and 2 online) to determine
3841  *  the health of the card.
3842  * Return value:
3843  *  void
3844  */
3845
3846 static void s2io_ethtool_test(struct net_device *dev,
3847                               struct ethtool_test *ethtest,
3848                               uint64_t * data)
3849 {
3850         nic_t *sp = dev->priv;
3851         int orig_state = netif_running(sp->dev);
3852
3853         if (ethtest->flags == ETH_TEST_FL_OFFLINE) {
3854                 /* Offline Tests. */
3855                 if (orig_state) {
3856                         s2io_close(sp->dev);
3857                         s2io_set_swapper(sp);
3858                 } else
3859                         s2io_set_swapper(sp);
3860
3861                 if (s2io_register_test(sp, &data[0]))
3862                         ethtest->flags |= ETH_TEST_FL_FAILED;
3863
3864                 s2io_reset(sp);
3865                 s2io_set_swapper(sp);
3866
3867                 if (s2io_rldram_test(sp, &data[3]))
3868                         ethtest->flags |= ETH_TEST_FL_FAILED;
3869
3870                 s2io_reset(sp);
3871                 s2io_set_swapper(sp);
3872
3873                 if (s2io_eeprom_test(sp, &data[1]))
3874                         ethtest->flags |= ETH_TEST_FL_FAILED;
3875
3876                 if (s2io_bist_test(sp, &data[4]))
3877                         ethtest->flags |= ETH_TEST_FL_FAILED;
3878
3879                 if (orig_state)
3880                         s2io_open(sp->dev);
3881
3882                 data[2] = 0;
3883         } else {
3884                 /* Online Tests. */
3885                 if (!orig_state) {
3886                         DBG_PRINT(ERR_DBG,
3887                                   "%s: is not up, cannot run test\n",
3888                                   dev->name);
3889                         data[0] = -1;
3890                         data[1] = -1;
3891                         data[2] = -1;
3892                         data[3] = -1;
3893                         data[4] = -1;
3894                 }
3895
3896                 if (s2io_link_test(sp, &data[2]))
3897                         ethtest->flags |= ETH_TEST_FL_FAILED;
3898
3899                 data[0] = 0;
3900                 data[1] = 0;
3901                 data[3] = 0;
3902                 data[4] = 0;
3903         }
3904 }
3905
3906 static void s2io_get_ethtool_stats(struct net_device *dev,
3907                                    struct ethtool_stats *estats,
3908                                    u64 * tmp_stats)
3909 {
3910         int i = 0;
3911         nic_t *sp = dev->priv;
3912         StatInfo_t *stat_info = sp->mac_control.stats_info;
3913
3914         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_frms);
3915         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_data_octets);
3916         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_drop_frms);
3917         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_mcst_frms);
3918         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_bcst_frms);
3919         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_pause_ctrl_frms);
3920         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_any_err_frms);
3921         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_vld_ip_octets);
3922         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_vld_ip);
3923         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_drop_ip);
3924         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_icmp);
3925         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_rst_tcp);
3926         tmp_stats[i++] = le64_to_cpu(stat_info->tmac_tcp);
3927         tmp_stats[i++] = le32_to_cpu(stat_info->tmac_udp);
3928         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_vld_frms);
3929         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_data_octets);
3930         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_fcs_err_frms);
3931         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_drop_frms);
3932         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_vld_mcst_frms);
3933         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_vld_bcst_frms);
3934         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_in_rng_len_err_frms);
3935         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_long_frms);
3936         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_pause_ctrl_frms);
3937         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_discarded_frms);
3938         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_usized_frms);
3939         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_osized_frms);
3940         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_frag_frms);
3941         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_jabber_frms);
3942         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_ip);
3943         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_ip_octets);
3944         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_hdr_err_ip);
3945         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_drop_ip);
3946         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_icmp);
3947         tmp_stats[i++] = le64_to_cpu(stat_info->rmac_tcp);
3948         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_udp);
3949         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_err_drp_udp);
3950         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_pause_cnt);
3951         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_accepted_ip);
3952         tmp_stats[i++] = le32_to_cpu(stat_info->rmac_err_tcp);
3953 }
3954
3955 static int s2io_ethtool_get_regs_len(struct net_device *dev)
3956 {
3957         return (XENA_REG_SPACE);
3958 }
3959
3960
3961 static u32 s2io_ethtool_get_rx_csum(struct net_device * dev)
3962 {
3963         nic_t *sp = dev->priv;
3964
3965         return (sp->rx_csum);
3966 }
3967
3968 static int s2io_ethtool_set_rx_csum(struct net_device *dev, u32 data)
3969 {
3970         nic_t *sp = dev->priv;
3971
3972         if (data)
3973                 sp->rx_csum = 1;
3974         else
3975                 sp->rx_csum = 0;
3976
3977         return 0;
3978 }
3979
3980 static int s2io_get_eeprom_len(struct net_device *dev)
3981 {
3982         return (XENA_EEPROM_SPACE);
3983 }
3984
3985 static int s2io_ethtool_self_test_count(struct net_device *dev)
3986 {
3987         return (S2IO_TEST_LEN);
3988 }
3989
3990 static void s2io_ethtool_get_strings(struct net_device *dev,
3991                                      u32 stringset, u8 * data)
3992 {
3993         switch (stringset) {
3994         case ETH_SS_TEST:
3995                 memcpy(data, s2io_gstrings, S2IO_STRINGS_LEN);
3996                 break;
3997         case ETH_SS_STATS:
3998                 memcpy(data, &ethtool_stats_keys,
3999                        sizeof(ethtool_stats_keys));
4000         }
4001 }
4002
4003 static int s2io_ethtool_get_stats_count(struct net_device *dev)
4004 {
4005         return (S2IO_STAT_LEN);
4006 }
4007
4008 static int s2io_ethtool_op_set_tx_csum(struct net_device *dev, u32 data)
4009 {
4010         if (data)
4011                 dev->features |= NETIF_F_IP_CSUM;
4012         else
4013                 dev->features &= ~NETIF_F_IP_CSUM;
4014
4015         return 0;
4016 }
4017
4018
4019 static struct ethtool_ops netdev_ethtool_ops = {
4020         .get_settings = s2io_ethtool_gset,
4021         .set_settings = s2io_ethtool_sset,
4022         .get_drvinfo = s2io_ethtool_gdrvinfo,
4023         .get_regs_len = s2io_ethtool_get_regs_len,
4024         .get_regs = s2io_ethtool_gregs,
4025         .get_link = ethtool_op_get_link,
4026         .get_eeprom_len = s2io_get_eeprom_len,
4027         .get_eeprom = s2io_ethtool_geeprom,
4028         .set_eeprom = s2io_ethtool_seeprom,
4029         .get_pauseparam = s2io_ethtool_getpause_data,
4030         .set_pauseparam = s2io_ethtool_setpause_data,
4031         .get_rx_csum = s2io_ethtool_get_rx_csum,
4032         .set_rx_csum = s2io_ethtool_set_rx_csum,
4033         .get_tx_csum = ethtool_op_get_tx_csum,
4034         .set_tx_csum = s2io_ethtool_op_set_tx_csum,
4035         .get_sg = ethtool_op_get_sg,
4036         .set_sg = ethtool_op_set_sg,
4037 #ifdef NETIF_F_TSO
4038         .get_tso = ethtool_op_get_tso,
4039         .set_tso = ethtool_op_set_tso,
4040 #endif
4041         .self_test_count = s2io_ethtool_self_test_count,
4042         .self_test = s2io_ethtool_test,
4043         .get_strings = s2io_ethtool_get_strings,
4044         .phys_id = s2io_ethtool_idnic,
4045         .get_stats_count = s2io_ethtool_get_stats_count,
4046         .get_ethtool_stats = s2io_get_ethtool_stats
4047 };
4048
4049 /**
4050  *  s2io_ioctl - Entry point for the Ioctl
4051  *  @dev :  Device pointer.
4052  *  @ifr :  An IOCTL specefic structure, that can contain a pointer to
4053  *  a proprietary structure used to pass information to the driver.
4054  *  @cmd :  This is used to distinguish between the different commands that
4055  *  can be passed to the IOCTL functions.
4056  *  Description:
4057  *  This function has support for ethtool, adding multiple MAC addresses on
4058  *  the NIC and some DBG commands for the util tool.
4059  *  Return value:
4060  *  Currently the IOCTL supports no operations, hence by default this
4061  *  function returns OP NOT SUPPORTED value.
4062  */
4063
4064 static int s2io_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
4065 {
4066         return -EOPNOTSUPP;
4067 }
4068
4069 /**
4070  *  s2io_change_mtu - entry point to change MTU size for the device.
4071  *   @dev : device pointer.
4072  *   @new_mtu : the new MTU size for the device.
4073  *   Description: A driver entry point to change MTU size for the device.
4074  *   Before changing the MTU the device must be stopped.
4075  *  Return value:
4076  *   0 on success and an appropriate (-)ve integer as defined in errno.h
4077  *   file on failure.
4078  */
4079
4080 static int s2io_change_mtu(struct net_device *dev, int new_mtu)
4081 {
4082         nic_t *sp = dev->priv;
4083         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4084         register u64 val64;
4085
4086         if (netif_running(dev)) {
4087                 DBG_PRINT(ERR_DBG, "%s: Must be stopped to ", dev->name);
4088                 DBG_PRINT(ERR_DBG, "change its MTU \n");
4089                 return -EBUSY;
4090         }
4091
4092         if ((new_mtu < MIN_MTU) || (new_mtu > S2IO_JUMBO_SIZE)) {
4093                 DBG_PRINT(ERR_DBG, "%s: MTU size is invalid.\n",
4094                           dev->name);
4095                 return -EPERM;
4096         }
4097
4098         /* Set the new MTU into the PYLD register of the NIC */
4099         val64 = new_mtu;
4100         writeq(vBIT(val64, 2, 14), &bar0->rmac_max_pyld_len);
4101
4102         dev->mtu = new_mtu;
4103
4104         return 0;
4105 }
4106
4107 /**
4108  *  s2io_tasklet - Bottom half of the ISR.
4109  *  @dev_adr : address of the device structure in dma_addr_t format.
4110  *  Description:
4111  *  This is the tasklet or the bottom half of the ISR. This is
4112  *  an extension of the ISR which is scheduled by the scheduler to be run
4113  *  when the load on the CPU is low. All low priority tasks of the ISR can
4114  *  be pushed into the tasklet. For now the tasklet is used only to
4115  *  replenish the Rx buffers in the Rx buffer descriptors.
4116  *  Return value:
4117  *  void.
4118  */
4119
4120 static void s2io_tasklet(unsigned long dev_addr)
4121 {
4122         struct net_device *dev = (struct net_device *) dev_addr;
4123         nic_t *sp = dev->priv;
4124         int i, ret;
4125         mac_info_t *mac_control;
4126         struct config_param *config;
4127
4128         mac_control = &sp->mac_control;
4129         config = &sp->config;
4130
4131         if (!TASKLET_IN_USE) {
4132                 for (i = 0; i < config->rx_ring_num; i++) {
4133                         ret = fill_rx_buffers(sp, i);
4134                         if (ret == -ENOMEM) {
4135                                 DBG_PRINT(ERR_DBG, "%s: Out of ",
4136                                           dev->name);
4137                                 DBG_PRINT(ERR_DBG, "memory in tasklet\n");
4138                                 break;
4139                         } else if (ret == -EFILL) {
4140                                 DBG_PRINT(ERR_DBG,
4141                                           "%s: Rx Ring %d is full\n",
4142                                           dev->name, i);
4143                                 break;
4144                         }
4145                 }
4146                 clear_bit(0, (&sp->tasklet_status));
4147         }
4148 }
4149
4150 /**
4151  * s2io_set_link - Set the LInk status
4152  * @data: long pointer to device private structue
4153  * Description: Sets the link status for the adapter
4154  */
4155
4156 static void s2io_set_link(unsigned long data)
4157 {
4158         nic_t *nic = (nic_t *) data;
4159         struct net_device *dev = nic->dev;
4160         XENA_dev_config_t __iomem *bar0 = nic->bar0;
4161         register u64 val64;
4162         u16 subid;
4163
4164         if (test_and_set_bit(0, &(nic->link_state))) {
4165                 /* The card is being reset, no point doing anything */
4166                 return;
4167         }
4168
4169         subid = nic->pdev->subsystem_device;
4170         /*
4171          * Allow a small delay for the NICs self initiated
4172          * cleanup to complete.
4173          */
4174         msleep(100);
4175
4176         val64 = readq(&bar0->adapter_status);
4177         if (verify_xena_quiescence(val64, nic->device_enabled_once)) {
4178                 if (LINK_IS_UP(val64)) {
4179                         val64 = readq(&bar0->adapter_control);
4180                         val64 |= ADAPTER_CNTL_EN;
4181                         writeq(val64, &bar0->adapter_control);
4182                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(subid)) {
4183                                 val64 = readq(&bar0->gpio_control);
4184                                 val64 |= GPIO_CTRL_GPIO_0;
4185                                 writeq(val64, &bar0->gpio_control);
4186                                 val64 = readq(&bar0->gpio_control);
4187                         } else {
4188                                 val64 |= ADAPTER_LED_ON;
4189                                 writeq(val64, &bar0->adapter_control);
4190                         }
4191                         val64 = readq(&bar0->adapter_status);
4192                         if (!LINK_IS_UP(val64)) {
4193                                 DBG_PRINT(ERR_DBG, "%s:", dev->name);
4194                                 DBG_PRINT(ERR_DBG, " Link down");
4195                                 DBG_PRINT(ERR_DBG, "after ");
4196                                 DBG_PRINT(ERR_DBG, "enabling ");
4197                                 DBG_PRINT(ERR_DBG, "device \n");
4198                         }
4199                         if (nic->device_enabled_once == FALSE) {
4200                                 nic->device_enabled_once = TRUE;
4201                         }
4202                         s2io_link(nic, LINK_UP);
4203                 } else {
4204                         if (CARDS_WITH_FAULTY_LINK_INDICATORS(subid)) {
4205                                 val64 = readq(&bar0->gpio_control);
4206                                 val64 &= ~GPIO_CTRL_GPIO_0;
4207                                 writeq(val64, &bar0->gpio_control);
4208                                 val64 = readq(&bar0->gpio_control);
4209                         }
4210                         s2io_link(nic, LINK_DOWN);
4211                 }
4212         } else {                /* NIC is not Quiescent. */
4213                 DBG_PRINT(ERR_DBG, "%s: Error: ", dev->name);
4214                 DBG_PRINT(ERR_DBG, "device is not Quiescent\n");
4215                 netif_stop_queue(dev);
4216         }
4217         clear_bit(0, &(nic->link_state));
4218 }
4219
4220 static void s2io_card_down(nic_t * sp)
4221 {
4222         int cnt = 0;
4223         XENA_dev_config_t __iomem *bar0 = sp->bar0;
4224         unsigned long flags;
4225         register u64 val64 = 0;
4226
4227         /* If s2io_set_link task is executing, wait till it completes. */
4228         while (test_and_set_bit(0, &(sp->link_state)))
4229                 msleep(50);
4230         atomic_set(&sp->card_state, CARD_DOWN);
4231
4232         /* disable Tx and Rx traffic on the NIC */
4233         stop_nic(sp);
4234
4235         /* Kill tasklet. */
4236         tasklet_kill(&sp->task);
4237
4238         /* Check if the device is Quiescent and then Reset the NIC */
4239         do {
4240                 val64 = readq(&bar0->adapter_status);
4241                 if (verify_xena_quiescence(val64, sp->device_enabled_once)) {
4242                         break;
4243                 }
4244
4245                 msleep(50);
4246                 cnt++;
4247                 if (cnt == 10) {
4248                         DBG_PRINT(ERR_DBG,
4249                                   "s2io_close:Device not Quiescent ");
4250                         DBG_PRINT(ERR_DBG, "adaper status reads 0x%llx\n",
4251                                   (unsigned long long) val64);
4252                         break;
4253                 }
4254         } while (1);
4255         spin_lock_irqsave(&sp->tx_lock, flags);
4256         s2io_reset(sp);
4257
4258         /* Free all unused Tx and Rx buffers */
4259         free_tx_buffers(sp);
4260         free_rx_buffers(sp);
4261
4262         spin_unlock_irqrestore(&sp->tx_lock, flags);
4263         clear_bit(0, &(sp->link_state));
4264 }
4265
4266 static int s2io_card_up(nic_t * sp)
4267 {
4268         int i, ret;
4269         mac_info_t *mac_control;
4270         struct config_param *config;
4271         struct net_device *dev = (struct net_device *) sp->dev;
4272
4273         /* Initialize the H/W I/O registers */
4274         if (init_nic(sp) != 0) {
4275                 DBG_PRINT(ERR_DBG, "%s: H/W initialization failed\n",
4276                           dev->name);
4277                 return -ENODEV;
4278         }
4279
4280         /*
4281          * Initializing the Rx buffers. For now we are considering only 1
4282          * Rx ring and initializing buffers into 30 Rx blocks
4283          */
4284         mac_control = &sp->mac_control;
4285         config = &sp->config;
4286
4287         for (i = 0; i < config->rx_ring_num; i++) {
4288                 if ((ret = fill_rx_buffers(sp, i))) {
4289                         DBG_PRINT(ERR_DBG, "%s: Out of memory in Open\n",
4290                                   dev->name);
4291                         s2io_reset(sp);
4292                         free_rx_buffers(sp);
4293                         return -ENOMEM;
4294                 }
4295                 DBG_PRINT(INFO_DBG, "Buf in ring:%d is %d:\n", i,
4296                           atomic_read(&sp->rx_bufs_left[i]));
4297         }
4298
4299         /* Setting its receive mode */
4300         s2io_set_multicast(dev);
4301
4302         /* Enable tasklet for the device */
4303         tasklet_init(&sp->task, s2io_tasklet, (unsigned long) dev);
4304
4305         /* Enable Rx Traffic and interrupts on the NIC */
4306         if (start_nic(sp)) {
4307                 DBG_PRINT(ERR_DBG, "%s: Starting NIC failed\n", dev->name);
4308                 tasklet_kill(&sp->task);
4309                 s2io_reset(sp);
4310                 free_irq(dev->irq, dev);
4311                 free_rx_buffers(sp);
4312                 return -ENODEV;
4313         }
4314
4315         atomic_set(&sp->card_state, CARD_UP);
4316         return 0;
4317 }
4318
4319 /**
4320  * s2io_restart_nic - Resets the NIC.
4321  * @data : long pointer to the device private structure
4322  * Description:
4323  * This function is scheduled to be run by the s2io_tx_watchdog
4324  * function after 0.5 secs to reset the NIC. The idea is to reduce
4325  * the run time of the watch dog routine which is run holding a
4326  * spin lock.
4327  */
4328
4329 static void s2io_restart_nic(unsigned long data)
4330 {
4331         struct net_device *dev = (struct net_device *) data;
4332         nic_t *sp = dev->priv;
4333
4334         s2io_card_down(sp);
4335         if (s2io_card_up(sp)) {
4336                 DBG_PRINT(ERR_DBG, "%s: Device bring up failed\n",
4337                           dev->name);
4338         }
4339         netif_wake_queue(dev);
4340         DBG_PRINT(ERR_DBG, "%s: was reset by Tx watchdog timer\n",
4341                   dev->name);
4342 }
4343
4344 /**
4345  *  s2io_tx_watchdog - Watchdog for transmit side.
4346  *  @dev : Pointer to net device structure
4347  *  Description:
4348  *  This function is triggered if the Tx Queue is stopped
4349  *  for a pre-defined amount of time when the Interface is still up.
4350  *  If the Interface is jammed in such a situation, the hardware is
4351  *  reset (by s2io_close) and restarted again (by s2io_open) to
4352  *  overcome any problem that might have been caused in the hardware.
4353  *  Return value:
4354  *  void
4355  */
4356
4357 static void s2io_tx_watchdog(struct net_device *dev)
4358 {
4359         nic_t *sp = dev->priv;
4360
4361         if (netif_carrier_ok(dev)) {
4362                 schedule_work(&sp->rst_timer_task);
4363         }
4364 }
4365
4366 /**
4367  *   rx_osm_handler - To perform some OS related operations on SKB.
4368  *   @sp: private member of the device structure,pointer to s2io_nic structure.
4369  *   @skb : the socket buffer pointer.
4370  *   @len : length of the packet
4371  *   @cksum : FCS checksum of the frame.
4372  *   @ring_no : the ring from which this RxD was extracted.
4373  *   Description:
4374  *   This function is called by the Tx interrupt serivce routine to perform
4375  *   some OS related operations on the SKB before passing it to the upper
4376  *   layers. It mainly checks if the checksum is OK, if so adds it to the
4377  *   SKBs cksum variable, increments the Rx packet count and passes the SKB
4378  *   to the upper layer. If the checksum is wrong, it increments the Rx
4379  *   packet error count, frees the SKB and returns error.
4380  *   Return value:
4381  *   SUCCESS on success and -1 on failure.
4382  */
4383 #ifndef CONFIG_2BUFF_MODE
4384 static int rx_osm_handler(nic_t * sp, u16 len, RxD_t * rxdp, int ring_no)
4385 #else
4386 static int rx_osm_handler(nic_t * sp, RxD_t * rxdp, int ring_no,
4387                           buffAdd_t * ba)
4388 #endif
4389 {
4390         struct net_device *dev = (struct net_device *) sp->dev;
4391         struct sk_buff *skb =
4392             (struct sk_buff *) ((unsigned long) rxdp->Host_Control);
4393         u16 l3_csum, l4_csum;
4394 #ifdef CONFIG_2BUFF_MODE
4395         int buf0_len, buf2_len;
4396         unsigned char *buff;
4397 #endif
4398
4399         l3_csum = RXD_GET_L3_CKSUM(rxdp->Control_1);
4400         if ((rxdp->Control_1 & TCP_OR_UDP_FRAME) && (sp->rx_csum)) {
4401                 l4_csum = RXD_GET_L4_CKSUM(rxdp->Control_1);
4402                 if ((l3_csum == L3_CKSUM_OK) && (l4_csum == L4_CKSUM_OK)) {
4403                         /*
4404                          * NIC verifies if the Checksum of the received
4405                          * frame is Ok or not and accordingly returns
4406                          * a flag in the RxD.
4407                          */
4408                         skb->ip_summed = CHECKSUM_UNNECESSARY;
4409                 } else {
4410                         /*
4411                          * Packet with erroneous checksum, let the
4412                          * upper layers deal with it.
4413                          */
4414                         skb->ip_summed = CHECKSUM_NONE;
4415                 }
4416         } else {
4417                 skb->ip_summed = CHECKSUM_NONE;
4418         }
4419
4420         if (rxdp->Control_1 & RXD_T_CODE) {
4421                 unsigned long long err = rxdp->Control_1 & RXD_T_CODE;
4422                 DBG_PRINT(ERR_DBG, "%s: Rx error Value: 0x%llx\n",
4423                           dev->name, err);
4424         }
4425 #ifdef CONFIG_2BUFF_MODE
4426         buf0_len = RXD_GET_BUFFER0_SIZE(rxdp->Control_2);
4427         buf2_len = RXD_GET_BUFFER2_SIZE(rxdp->Control_2);
4428 #endif
4429
4430         skb->dev = dev;
4431 #ifndef CONFIG_2BUFF_MODE
4432         skb_put(skb, len);
4433         skb->protocol = eth_type_trans(skb, dev);
4434 #else
4435         buff = skb_push(skb, buf0_len);
4436         memcpy(buff, ba->ba_0, buf0_len);
4437         skb_put(skb, buf2_len);
4438         skb->protocol = eth_type_trans(skb, dev);
4439 #endif
4440
4441 #ifdef CONFIG_S2IO_NAPI
4442         netif_receive_skb(skb);
4443 #else
4444         netif_rx(skb);
4445 #endif
4446
4447         dev->last_rx = jiffies;
4448         sp->rx_pkt_count++;
4449         sp->stats.rx_packets++;
4450 #ifndef CONFIG_2BUFF_MODE
4451         sp->stats.rx_bytes += len;
4452 #else
4453         sp->stats.rx_bytes += buf0_len + buf2_len;
4454 #endif
4455
4456         atomic_dec(&sp->rx_bufs_left[ring_no]);
4457         rxdp->Host_Control = 0;
4458         return SUCCESS;
4459 }
4460
4461 /**
4462  *  s2io_link - stops/starts the Tx queue.
4463  *  @sp : private member of the device structure, which is a pointer to the
4464  *  s2io_nic structure.
4465  *  @link : inidicates whether link is UP/DOWN.
4466  *  Description:
4467  *  This function stops/starts the Tx queue depending on whether the link
4468  *  status of the NIC is is down or up. This is called by the Alarm
4469  *  interrupt handler whenever a link change interrupt comes up.
4470  *  Return value:
4471  *  void.
4472  */
4473
4474 static void s2io_link(nic_t * sp, int link)
4475 {
4476         struct net_device *dev = (struct net_device *) sp->dev;
4477
4478         if (link != sp->last_link_state) {
4479                 if (link == LINK_DOWN) {
4480                         DBG_PRINT(ERR_DBG, "%s: Link down\n", dev->name);
4481                         netif_carrier_off(dev);
4482                 } else {
4483                         DBG_PRINT(ERR_DBG, "%s: Link Up\n", dev->name);
4484                         netif_carrier_on(dev);
4485                 }
4486         }
4487         sp->last_link_state = link;
4488 }
4489
4490 /**
4491  *  s2io_init_pci -Initialization of PCI and PCI-X configuration registers .
4492  *  @sp : private member of the device structure, which is a pointer to the
4493  *  s2io_nic structure.
4494  *  Description:
4495  *  This function initializes a few of the PCI and PCI-X configuration registers
4496  *  with recommended values.
4497  *  Return value:
4498  *  void
4499  */
4500
4501 static void s2io_init_pci(nic_t * sp)
4502 {
4503         u16 pci_cmd = 0;
4504
4505         /* Enable Data Parity Error Recovery in PCI-X command register. */
4506         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4507                              &(sp->pcix_cmd));
4508         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4509                               (sp->pcix_cmd | 1));
4510         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4511                              &(sp->pcix_cmd));
4512
4513         /* Set the PErr Response bit in PCI command register. */
4514         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
4515         pci_write_config_word(sp->pdev, PCI_COMMAND,
4516                               (pci_cmd | PCI_COMMAND_PARITY));
4517         pci_read_config_word(sp->pdev, PCI_COMMAND, &pci_cmd);
4518
4519         /* Set MMRB count to 1024 in PCI-X Command register. */
4520         sp->pcix_cmd &= 0xFFF3;
4521         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER, (sp->pcix_cmd | (0x1 << 2)));    /* MMRBC 1K */
4522         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4523                              &(sp->pcix_cmd));
4524
4525         /*  Setting Maximum outstanding splits based on system type. */
4526         sp->pcix_cmd &= 0xFF8F;
4527
4528         sp->pcix_cmd |= XENA_MAX_OUTSTANDING_SPLITS(0x1);       /* 2 splits. */
4529         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4530                               sp->pcix_cmd);
4531         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4532                              &(sp->pcix_cmd));
4533         /* Forcibly disabling relaxed ordering capability of the card. */
4534         sp->pcix_cmd &= 0xfffd;
4535         pci_write_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4536                               sp->pcix_cmd);
4537         pci_read_config_word(sp->pdev, PCIX_COMMAND_REGISTER,
4538                              &(sp->pcix_cmd));
4539 }
4540
4541 MODULE_AUTHOR("Raghavendra Koushik <raghavendra.koushik@neterion.com>");
4542 MODULE_LICENSE("GPL");
4543 module_param(tx_fifo_num, int, 0);
4544 module_param_array(tx_fifo_len, int, NULL, 0);
4545 module_param(rx_ring_num, int, 0);
4546 module_param_array(rx_ring_sz, int, NULL, 0);
4547 module_param(Stats_refresh_time, int, 0);
4548 module_param(rmac_pause_time, int, 0);
4549 module_param(mc_pause_threshold_q0q3, int, 0);
4550 module_param(mc_pause_threshold_q4q7, int, 0);
4551 module_param(shared_splits, int, 0);
4552 module_param(tmac_util_period, int, 0);
4553 module_param(rmac_util_period, int, 0);
4554 #ifndef CONFIG_S2IO_NAPI
4555 module_param(indicate_max_pkts, int, 0);
4556 #endif
4557 /**
4558  *  s2io_init_nic - Initialization of the adapter .
4559  *  @pdev : structure containing the PCI related information of the device.
4560  *  @pre: List of PCI devices supported by the driver listed in s2io_tbl.
4561  *  Description:
4562  *  The function initializes an adapter identified by the pci_dec structure.
4563  *  All OS related initialization including memory and device structure and
4564  *  initlaization of the device private variable is done. Also the swapper
4565  *  control register is initialized to enable read and write into the I/O
4566  *  registers of the device.
4567  *  Return value:
4568  *  returns 0 on success and negative on failure.
4569  */
4570
4571 static int __devinit
4572 s2io_init_nic(struct pci_dev *pdev, const struct pci_device_id *pre)
4573 {
4574         nic_t *sp;
4575         struct net_device *dev;
4576         char *dev_name = "S2IO 10GE NIC";
4577         int i, j, ret;
4578         int dma_flag = FALSE;
4579         u32 mac_up, mac_down;
4580         u64 val64 = 0, tmp64 = 0;
4581         XENA_dev_config_t __iomem *bar0 = NULL;
4582         u16 subid;
4583         mac_info_t *mac_control;
4584         struct config_param *config;
4585
4586
4587         DBG_PRINT(ERR_DBG, "Loading S2IO driver with %s\n",
4588                 s2io_driver_version);
4589
4590         if ((ret = pci_enable_device(pdev))) {
4591                 DBG_PRINT(ERR_DBG,
4592                           "s2io_init_nic: pci_enable_device failed\n");
4593                 return ret;
4594         }
4595
4596         if (!pci_set_dma_mask(pdev, 0xffffffffffffffffULL)) {
4597                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 64bit DMA\n");
4598                 dma_flag = TRUE;
4599
4600                 if (pci_set_consistent_dma_mask
4601                     (pdev, 0xffffffffffffffffULL)) {
4602                         DBG_PRINT(ERR_DBG,
4603                                   "Unable to obtain 64bit DMA for \
4604                                         consistent allocations\n");
4605                         pci_disable_device(pdev);
4606                         return -ENOMEM;
4607                 }
4608         } else if (!pci_set_dma_mask(pdev, 0xffffffffUL)) {
4609                 DBG_PRINT(INIT_DBG, "s2io_init_nic: Using 32bit DMA\n");
4610         } else {
4611                 pci_disable_device(pdev);
4612                 return -ENOMEM;
4613         }
4614
4615         if (pci_request_regions(pdev, s2io_driver_name)) {
4616                 DBG_PRINT(ERR_DBG, "Request Regions failed\n"),
4617                     pci_disable_device(pdev);
4618                 return -ENODEV;
4619         }
4620
4621         dev = alloc_etherdev(sizeof(nic_t));
4622         if (dev == NULL) {
4623                 DBG_PRINT(ERR_DBG, "Device allocation failed\n");
4624                 pci_disable_device(pdev);
4625                 pci_release_regions(pdev);
4626                 return -ENODEV;
4627         }
4628
4629         pci_set_master(pdev);
4630         pci_set_drvdata(pdev, dev);
4631         SET_MODULE_OWNER(dev);
4632         SET_NETDEV_DEV(dev, &pdev->dev);
4633
4634         /*  Private member variable initialized to s2io NIC structure */
4635         sp = dev->priv;
4636         memset(sp, 0, sizeof(nic_t));
4637         sp->dev = dev;
4638         sp->pdev = pdev;
4639         sp->vendor_id = pdev->vendor;
4640         sp->device_id = pdev->device;
4641         sp->high_dma_flag = dma_flag;
4642         sp->irq = pdev->irq;
4643         sp->device_enabled_once = FALSE;
4644         strcpy(sp->name, dev_name);
4645
4646         /* Initialize some PCI/PCI-X fields of the NIC. */
4647         s2io_init_pci(sp);
4648
4649         /*
4650          * Setting the device configuration parameters.
4651          * Most of these parameters can be specified by the user during
4652          * module insertion as they are module loadable parameters. If
4653          * these parameters are not not specified during load time, they
4654          * are initialized with default values.
4655          */
4656         mac_control = &sp->mac_control;
4657         config = &sp->config;
4658
4659         /* Tx side parameters. */
4660         tx_fifo_len[0] = DEFAULT_FIFO_LEN;      /* Default value. */
4661         config->tx_fifo_num = tx_fifo_num;
4662         for (i = 0; i < MAX_TX_FIFOS; i++) {
4663                 config->tx_cfg[i].fifo_len = tx_fifo_len[i];
4664                 config->tx_cfg[i].fifo_priority = i;
4665         }
4666
4667         config->tx_intr_type = TXD_INT_TYPE_UTILZ;
4668         for (i = 0; i < config->tx_fifo_num; i++) {
4669                 config->tx_cfg[i].f_no_snoop =
4670                     (NO_SNOOP_TXD | NO_SNOOP_TXD_BUFFER);
4671                 if (config->tx_cfg[i].fifo_len < 65) {
4672                         config->tx_intr_type = TXD_INT_TYPE_PER_LIST;
4673                         break;
4674                 }
4675         }
4676         config->max_txds = MAX_SKB_FRAGS;
4677
4678         /* Rx side parameters. */
4679         rx_ring_sz[0] = SMALL_BLK_CNT;  /* Default value. */
4680         config->rx_ring_num = rx_ring_num;
4681         for (i = 0; i < MAX_RX_RINGS; i++) {
4682                 config->rx_cfg[i].num_rxd = rx_ring_sz[i] *
4683                     (MAX_RXDS_PER_BLOCK + 1);
4684                 config->rx_cfg[i].ring_priority = i;
4685         }
4686
4687         for (i = 0; i < rx_ring_num; i++) {
4688                 config->rx_cfg[i].ring_org = RING_ORG_BUFF1;
4689                 config->rx_cfg[i].f_no_snoop =
4690                     (NO_SNOOP_RXD | NO_SNOOP_RXD_BUFFER);
4691         }
4692
4693         /*  Setting Mac Control parameters */
4694         mac_control->rmac_pause_time = rmac_pause_time;
4695         mac_control->mc_pause_threshold_q0q3 = mc_pause_threshold_q0q3;
4696         mac_control->mc_pause_threshold_q4q7 = mc_pause_threshold_q4q7;
4697
4698
4699         /* Initialize Ring buffer parameters. */
4700         for (i = 0; i < config->rx_ring_num; i++)
4701                 atomic_set(&sp->rx_bufs_left[i], 0);
4702
4703         /*  initialize the shared memory used by the NIC and the host */
4704         if (init_shared_mem(sp)) {
4705                 DBG_PRINT(ERR_DBG, "%s: Memory allocation failed\n",
4706                           dev->name);
4707                 ret = -ENOMEM;
4708                 goto mem_alloc_failed;
4709         }
4710
4711         sp->bar0 = ioremap(pci_resource_start(pdev, 0),
4712                                      pci_resource_len(pdev, 0));
4713         if (!sp->bar0) {
4714                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem1\n",
4715                           dev->name);
4716                 ret = -ENOMEM;
4717                 goto bar0_remap_failed;
4718         }
4719
4720         sp->bar1 = ioremap(pci_resource_start(pdev, 2),
4721                                      pci_resource_len(pdev, 2));
4722         if (!sp->bar1) {
4723                 DBG_PRINT(ERR_DBG, "%s: S2IO: cannot remap io mem2\n",
4724                           dev->name);
4725                 ret = -ENOMEM;
4726                 goto bar1_remap_failed;
4727         }
4728
4729         dev->irq = pdev->irq;
4730         dev->base_addr = (unsigned long) sp->bar0;
4731
4732         /* Initializing the BAR1 address as the start of the FIFO pointer. */
4733         for (j = 0; j < MAX_TX_FIFOS; j++) {
4734                 mac_control->tx_FIFO_start[j] = (TxFIFO_element_t __iomem *)
4735                     (sp->bar1 + (j * 0x00020000));
4736         }
4737
4738         /*  Driver entry points */
4739         dev->open = &s2io_open;
4740         dev->stop = &s2io_close;
4741         dev->hard_start_xmit = &s2io_xmit;
4742         dev->get_stats = &s2io_get_stats;
4743         dev->set_multicast_list = &s2io_set_multicast;
4744         dev->do_ioctl = &s2io_ioctl;
4745         dev->change_mtu = &s2io_change_mtu;
4746         SET_ETHTOOL_OPS(dev, &netdev_ethtool_ops);
4747         /*
4748          * will use eth_mac_addr() for  dev->set_mac_address
4749          * mac address will be set every time dev->open() is called
4750          */
4751 #ifdef CONFIG_S2IO_NAPI
4752         dev->poll = s2io_poll;
4753         dev->weight = 90;
4754 #endif
4755
4756         dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM;
4757         if (sp->high_dma_flag == TRUE)
4758                 dev->features |= NETIF_F_HIGHDMA;
4759 #ifdef NETIF_F_TSO
4760         dev->features |= NETIF_F_TSO;
4761 #endif
4762
4763         dev->tx_timeout = &s2io_tx_watchdog;
4764         dev->watchdog_timeo = WATCH_DOG_TIMEOUT;
4765         INIT_WORK(&sp->rst_timer_task,
4766                   (void (*)(void *)) s2io_restart_nic, dev);
4767         INIT_WORK(&sp->set_link_task,
4768                   (void (*)(void *)) s2io_set_link, sp);
4769
4770         pci_save_state(sp->pdev);
4771
4772         /* Setting swapper control on the NIC, for proper reset operation */
4773         if (s2io_set_swapper(sp)) {
4774                 DBG_PRINT(ERR_DBG, "%s:swapper settings are wrong\n",
4775                           dev->name);
4776                 ret = -EAGAIN;
4777                 goto set_swap_failed;
4778         }
4779
4780         /* Fix for all "FFs" MAC address problems observed on Alpha platforms */
4781         fix_mac_address(sp);
4782         s2io_reset(sp);
4783
4784         /*
4785          * Setting swapper control on the NIC, so the MAC address can be read.
4786          */
4787         if (s2io_set_swapper(sp)) {
4788                 DBG_PRINT(ERR_DBG,
4789                           "%s: S2IO: swapper settings are wrong\n",
4790                           dev->name);
4791                 ret = -EAGAIN;
4792                 goto set_swap_failed;
4793         }
4794
4795         /*
4796          * MAC address initialization.
4797          * For now only one mac address will be read and used.
4798          */
4799         bar0 = sp->bar0;
4800         val64 = RMAC_ADDR_CMD_MEM_RD | RMAC_ADDR_CMD_MEM_STROBE_NEW_CMD |
4801             RMAC_ADDR_CMD_MEM_OFFSET(0 + MAC_MAC_ADDR_START_OFFSET);
4802         writeq(val64, &bar0->rmac_addr_cmd_mem);
4803         wait_for_cmd_complete(sp);
4804
4805         tmp64 = readq(&bar0->rmac_addr_data0_mem);
4806         mac_down = (u32) tmp64;
4807         mac_up = (u32) (tmp64 >> 32);
4808
4809         memset(sp->def_mac_addr[0].mac_addr, 0, sizeof(ETH_ALEN));
4810
4811         sp->def_mac_addr[0].mac_addr[3] = (u8) (mac_up);
4812         sp->def_mac_addr[0].mac_addr[2] = (u8) (mac_up >> 8);
4813         sp->def_mac_addr[0].mac_addr[1] = (u8) (mac_up >> 16);
4814         sp->def_mac_addr[0].mac_addr[0] = (u8) (mac_up >> 24);
4815         sp->def_mac_addr[0].mac_addr[5] = (u8) (mac_down >> 16);
4816         sp->def_mac_addr[0].mac_addr[4] = (u8) (mac_down >> 24);
4817
4818         DBG_PRINT(INIT_DBG,
4819                   "DEFAULT MAC ADDR:0x%02x-%02x-%02x-%02x-%02x-%02x\n",
4820                   sp->def_mac_addr[0].mac_addr[0],
4821                   sp->def_mac_addr[0].mac_addr[1],
4822                   sp->def_mac_addr[0].mac_addr[2],
4823                   sp->def_mac_addr[0].mac_addr[3],
4824                   sp->def_mac_addr[0].mac_addr[4],
4825                   sp->def_mac_addr[0].mac_addr[5]);
4826
4827         /*  Set the factory defined MAC address initially   */
4828         dev->addr_len = ETH_ALEN;
4829         memcpy(dev->dev_addr, sp->def_mac_addr, ETH_ALEN);
4830
4831         /*
4832          * Initialize the tasklet status and link state flags
4833          * and the card statte parameter
4834          */
4835         atomic_set(&(sp->card_state), 0);
4836         sp->tasklet_status = 0;
4837         sp->link_state = 0;
4838
4839
4840         /* Initialize spinlocks */
4841         spin_lock_init(&sp->tx_lock);
4842 #ifndef CONFIG_S2IO_NAPI
4843         spin_lock_init(&sp->put_lock);
4844 #endif
4845
4846         /*
4847          * SXE-002: Configure link and activity LED to init state
4848          * on driver load.
4849          */
4850         subid = sp->pdev->subsystem_device;
4851         if ((subid & 0xFF) >= 0x07) {
4852                 val64 = readq(&bar0->gpio_control);
4853                 val64 |= 0x0000800000000000ULL;
4854                 writeq(val64, &bar0->gpio_control);
4855                 val64 = 0x0411040400000000ULL;
4856                 writeq(val64, (void __iomem *) bar0 + 0x2700);
4857                 val64 = readq(&bar0->gpio_control);
4858         }
4859
4860         sp->rx_csum = 1;        /* Rx chksum verify enabled by default */
4861
4862         if (register_netdev(dev)) {
4863                 DBG_PRINT(ERR_DBG, "Device registration failed\n");
4864                 ret = -ENODEV;
4865                 goto register_failed;
4866         }
4867
4868         /*
4869          * Make Link state as off at this point, when the Link change
4870          * interrupt comes the state will be automatically changed to
4871          * the right state.
4872          */
4873         netif_carrier_off(dev);
4874         sp->last_link_state = LINK_DOWN;
4875
4876         return 0;
4877
4878       register_failed:
4879       set_swap_failed:
4880         iounmap(sp->bar1);
4881       bar1_remap_failed:
4882         iounmap(sp->bar0);
4883       bar0_remap_failed:
4884       mem_alloc_failed:
4885         free_shared_mem(sp);
4886         pci_disable_device(pdev);
4887         pci_release_regions(pdev);
4888         pci_set_drvdata(pdev, NULL);
4889         free_netdev(dev);
4890
4891         return ret;
4892 }
4893
4894 /**
4895  * s2io_rem_nic - Free the PCI device
4896  * @pdev: structure containing the PCI related information of the device.
4897  * Description: This function is called by the Pci subsystem to release a
4898  * PCI device and free up all resource held up by the device. This could
4899  * be in response to a Hot plug event or when the driver is to be removed
4900  * from memory.
4901  */
4902
4903 static void __devexit s2io_rem_nic(struct pci_dev *pdev)
4904 {
4905         struct net_device *dev =
4906             (struct net_device *) pci_get_drvdata(pdev);
4907         nic_t *sp;
4908
4909         if (dev == NULL) {
4910                 DBG_PRINT(ERR_DBG, "Driver Data is NULL!!\n");
4911                 return;
4912         }
4913
4914         sp = dev->priv;
4915         unregister_netdev(dev);
4916
4917         free_shared_mem(sp);
4918         iounmap(sp->bar0);
4919         iounmap(sp->bar1);
4920         pci_disable_device(pdev);
4921         pci_release_regions(pdev);
4922         pci_set_drvdata(pdev, NULL);
4923
4924         free_netdev(dev);
4925 }
4926
4927 /**
4928  * s2io_starter - Entry point for the driver
4929  * Description: This function is the entry point for the driver. It verifies
4930  * the module loadable parameters and initializes PCI configuration space.
4931  */
4932
4933 int __init s2io_starter(void)
4934 {
4935         return pci_module_init(&s2io_driver);
4936 }
4937
4938 /**
4939  * s2io_closer - Cleanup routine for the driver
4940  * Description: This function is the cleanup routine for the driver. It unregist * ers the driver.
4941  */
4942
4943 static void s2io_closer(void)
4944 {
4945         pci_unregister_driver(&s2io_driver);
4946         DBG_PRINT(INIT_DBG, "cleanup done\n");
4947 }
4948
4949 module_init(s2io_starter);
4950 module_exit(s2io_closer);