drivers/net/ethernet/emulex/benet/be_main.c

   1 /*
   2  * Copyright (C) 2005 - 2014 Emulex
   3  * All rights reserved.
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License version 2
   7  * as published by the Free Software Foundation.  The full GNU General
   8  * Public License is included in this distribution in the file called COPYING.
   9  *
  10  * Contact Information:
  11  * linux-drivers@emulex.com
  12  *
  13  * Emulex
  14  * 3333 Susan Street
  15  * Costa Mesa, CA 92626
  16  */
  17
  18 #include <linux/prefetch.h>
  19 #include <linux/module.h>
  20 #include "be.h"
  21 #include "be_cmds.h"
  22 #include <asm/div64.h>
  23 #include <linux/aer.h>
  24 #include <linux/if_bridge.h>
  25 #include <net/busy_poll.h>
  26 #include <net/vxlan.h>
  27
  28 MODULE_VERSION(DRV_VER);
  29 MODULE_DESCRIPTION(DRV_DESC " " DRV_VER);
  30 MODULE_AUTHOR("Emulex Corporation");
  31 MODULE_LICENSE("GPL");
  32
  33 /* num_vfs module param is obsolete.
  34  * Use sysfs method to enable/disable VFs.
  35  */
  36 static unsigned int num_vfs;
  37 module_param(num_vfs, uint, S_IRUGO);
  38 MODULE_PARM_DESC(num_vfs, "Number of PCI VFs to initialize");
  39
  40 static ushort rx_frag_size = 2048;
  41 module_param(rx_frag_size, ushort, S_IRUGO);
  42 MODULE_PARM_DESC(rx_frag_size, "Size of a fragment that holds rcvd data.");
  43
  44 static const struct pci_device_id be_dev_ids[] = {
  45         { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID1) },
  46         { PCI_DEVICE(BE_VENDOR_ID, BE_DEVICE_ID2) },
  47         { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID1) },
  48         { PCI_DEVICE(BE_VENDOR_ID, OC_DEVICE_ID2) },
  49         { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID3)},
  50         { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID4)},
  51         { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID5)},
  52         { PCI_DEVICE(EMULEX_VENDOR_ID, OC_DEVICE_ID6)},
  53         { 0 }
  54 };
  55 MODULE_DEVICE_TABLE(pci, be_dev_ids);
  56 /* UE Status Low CSR */
  57 static const char * const ue_status_low_desc[] = {
  58         "CEV",
  59         "CTX",
  60         "DBUF",
  61         "ERX",
  62         "Host",
  63         "MPU",
  64         "NDMA",
  65         "PTC ",
  66         "RDMA ",
  67         "RXF ",
  68         "RXIPS ",
  69         "RXULP0 ",
  70         "RXULP1 ",
  71         "RXULP2 ",
  72         "TIM ",
  73         "TPOST ",
  74         "TPRE ",
  75         "TXIPS ",
  76         "TXULP0 ",
  77         "TXULP1 ",
  78         "UC ",
  79         "WDMA ",
  80         "TXULP2 ",
  81         "HOST1 ",
  82         "P0_OB_LINK ",
  83         "P1_OB_LINK ",
  84         "HOST_GPIO ",
  85         "MBOX ",
  86         "ERX2 ",
  87         "SPARE ",
  88         "JTAG ",
  89         "MPU_INTPEND "
  90 };
  91
  92 /* UE Status High CSR */
  93 static const char * const ue_status_hi_desc[] = {
  94         "LPCMEMHOST",
  95         "MGMT_MAC",
  96         "PCS0ONLINE",
  97         "MPU_IRAM",
  98         "PCS1ONLINE",
  99         "PCTL0",
 100         "PCTL1",
 101         "PMEM",
 102         "RR",
 103         "TXPB",
 104         "RXPP",
 105         "XAUI",
 106         "TXP",
 107         "ARM",
 108         "IPC",
 109         "HOST2",
 110         "HOST3",
 111         "HOST4",
 112         "HOST5",
 113         "HOST6",
 114         "HOST7",
 115         "ECRC",
 116         "Poison TLP",
 117         "NETC",
 118         "PERIPH",
 119         "LLTXULP",
 120         "D2P",
 121         "RCON",
 122         "LDMA",
 123         "LLTXP",
 124         "LLTXPB",
 125         "Unknown"
 126 };
 127
 128 static void be_queue_free(struct be_adapter *adapter, struct be_queue_info *q)
 129 {
 130         struct be_dma_mem *mem = &q->dma_mem;
 131
 132         if (mem->va) {
 133                 dma_free_coherent(&adapter->pdev->dev, mem->size, mem->va,
 134                                   mem->dma);
 135                 mem->va = NULL;
 136         }
 137 }
 138
 139 static int be_queue_alloc(struct be_adapter *adapter, struct be_queue_info *q,
 140                           u16 len, u16 entry_size)
 141 {
 142         struct be_dma_mem *mem = &q->dma_mem;
 143
 144         memset(q, 0, sizeof(*q));
 145         q->len = len;
 146         q->entry_size = entry_size;
 147         mem->size = len * entry_size;
 148         mem->va = dma_zalloc_coherent(&adapter->pdev->dev, mem->size, &mem->dma,
 149                                       GFP_KERNEL);
 150         if (!mem->va)
 151                 return -ENOMEM;
 152         return 0;
 153 }
 154
 155 static void be_reg_intr_set(struct be_adapter *adapter, bool enable)
 156 {
 157         u32 reg, enabled;
 158
 159         pci_read_config_dword(adapter->pdev, PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET,
 160                               &reg);
 161         enabled = reg & MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
 162
 163         if (!enabled && enable)
 164                 reg |= MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
 165         else if (enabled && !enable)
 166                 reg &= ~MEMBAR_CTRL_INT_CTRL_HOSTINTR_MASK;
 167         else
 168                 return;
 169
 170         pci_write_config_dword(adapter->pdev,
 171                                PCICFG_MEMBAR_CTRL_INT_CTRL_OFFSET, reg);
 172 }
 173
 174 static void be_intr_set(struct be_adapter *adapter, bool enable)
 175 {
 176         int status = 0;
 177
 178         /* On lancer interrupts can't be controlled via this register */
 179         if (lancer_chip(adapter))
 180                 return;
 181
 182         if (adapter->eeh_error)
 183                 return;
 184
 185         status = be_cmd_intr_set(adapter, enable);
 186         if (status)
 187                 be_reg_intr_set(adapter, enable);
 188 }
 189
 190 static void be_rxq_notify(struct be_adapter *adapter, u16 qid, u16 posted)
 191 {
 192         u32 val = 0;
 193
 194         val |= qid & DB_RQ_RING_ID_MASK;
 195         val |= posted << DB_RQ_NUM_POSTED_SHIFT;
 196
 197         wmb();
 198         iowrite32(val, adapter->db + DB_RQ_OFFSET);
 199 }
 200
 201 static void be_txq_notify(struct be_adapter *adapter, struct be_tx_obj *txo,
 202                           u16 posted)
 203 {
 204         u32 val = 0;
 205
 206         val |= txo->q.id & DB_TXULP_RING_ID_MASK;
 207         val |= (posted & DB_TXULP_NUM_POSTED_MASK) << DB_TXULP_NUM_POSTED_SHIFT;
 208
 209         wmb();
 210         iowrite32(val, adapter->db + txo->db_offset);
 211 }
 212
 213 static void be_eq_notify(struct be_adapter *adapter, u16 qid,
 214                          bool arm, bool clear_int, u16 num_popped)
 215 {
 216         u32 val = 0;
 217
 218         val |= qid & DB_EQ_RING_ID_MASK;
 219         val |= ((qid & DB_EQ_RING_ID_EXT_MASK) << DB_EQ_RING_ID_EXT_MASK_SHIFT);
 220
 221         if (adapter->eeh_error)
 222                 return;
 223
 224         if (arm)
 225                 val |= 1 << DB_EQ_REARM_SHIFT;
 226         if (clear_int)
 227                 val |= 1 << DB_EQ_CLR_SHIFT;
 228         val |= 1 << DB_EQ_EVNT_SHIFT;
 229         val |= num_popped << DB_EQ_NUM_POPPED_SHIFT;
 230         iowrite32(val, adapter->db + DB_EQ_OFFSET);
 231 }
 232
 233 void be_cq_notify(struct be_adapter *adapter, u16 qid, bool arm, u16 num_popped)
 234 {
 235         u32 val = 0;
 236
 237         val |= qid & DB_CQ_RING_ID_MASK;
 238         val |= ((qid & DB_CQ_RING_ID_EXT_MASK) <<
 239                         DB_CQ_RING_ID_EXT_MASK_SHIFT);
 240
 241         if (adapter->eeh_error)
 242                 return;
 243
 244         if (arm)
 245                 val |= 1 << DB_CQ_REARM_SHIFT;
 246         val |= num_popped << DB_CQ_NUM_POPPED_SHIFT;
 247         iowrite32(val, adapter->db + DB_CQ_OFFSET);
 248 }
 249
 250 static int be_mac_addr_set(struct net_device *netdev, void *p)
 251 {
 252         struct be_adapter *adapter = netdev_priv(netdev);
 253         struct device *dev = &adapter->pdev->dev;
 254         struct sockaddr *addr = p;
 255         int status;
 256         u8 mac[ETH_ALEN];
 257         u32 old_pmac_id = adapter->pmac_id[0], curr_pmac_id = 0;
 258
 259         if (!is_valid_ether_addr(addr->sa_data))
 260                 return -EADDRNOTAVAIL;
 261
 262         /* Proceed further only if, User provided MAC is different
 263          * from active MAC
 264          */
 265         if (ether_addr_equal(addr->sa_data, netdev->dev_addr))
 266                 return 0;
 267
 268         /* The PMAC_ADD cmd may fail if the VF doesn't have FILTMGMT
 269          * privilege or if PF did not provision the new MAC address.
 270          * On BE3, this cmd will always fail if the VF doesn't have the
 271          * FILTMGMT privilege. This failure is OK, only if the PF programmed
 272          * the MAC for the VF.
 273          */
 274         status = be_cmd_pmac_add(adapter, (u8 *)addr->sa_data,
 275                                  adapter->if_handle, &adapter->pmac_id[0], 0);
 276         if (!status) {
 277                 curr_pmac_id = adapter->pmac_id[0];
 278
 279                 /* Delete the old programmed MAC. This call may fail if the
 280                  * old MAC was already deleted by the PF driver.
 281                  */
 282                 if (adapter->pmac_id[0] != old_pmac_id)
 283                         be_cmd_pmac_del(adapter, adapter->if_handle,
 284                                         old_pmac_id, 0);
 285         }
 286
 287         /* Decide if the new MAC is successfully activated only after
 288          * querying the FW
 289          */
 290         status = be_cmd_get_active_mac(adapter, curr_pmac_id, mac,
 291                                        adapter->if_handle, true, 0);
 292         if (status)
 293                 goto err;
 294
 295         /* The MAC change did not happen, either due to lack of privilege
 296          * or PF didn't pre-provision.
 297          */
 298         if (!ether_addr_equal(addr->sa_data, mac)) {
 299                 status = -EPERM;
 300                 goto err;
 301         }
 302
 303         memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
 304         dev_info(dev, "MAC address changed to %pM\n", mac);
 305         return 0;
 306 err:
 307         dev_warn(dev, "MAC address change to %pM failed\n", addr->sa_data);
 308         return status;
 309 }
 310
 311 /* BE2 supports only v0 cmd */
 312 static void *hw_stats_from_cmd(struct be_adapter *adapter)
 313 {
 314         if (BE2_chip(adapter)) {
 315                 struct be_cmd_resp_get_stats_v0 *cmd = adapter->stats_cmd.va;
 316
 317                 return &cmd->hw_stats;
 318         } else if (BE3_chip(adapter)) {
 319                 struct be_cmd_resp_get_stats_v1 *cmd = adapter->stats_cmd.va;
 320
 321                 return &cmd->hw_stats;
 322         } else {
 323                 struct be_cmd_resp_get_stats_v2 *cmd = adapter->stats_cmd.va;
 324
 325                 return &cmd->hw_stats;
 326         }
 327 }
 328
 329 /* BE2 supports only v0 cmd */
 330 static void *be_erx_stats_from_cmd(struct be_adapter *adapter)
 331 {
 332         if (BE2_chip(adapter)) {
 333                 struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
 334
 335                 return &hw_stats->erx;
 336         } else if (BE3_chip(adapter)) {
 337                 struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
 338
 339                 return &hw_stats->erx;
 340         } else {
 341                 struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
 342
 343                 return &hw_stats->erx;
 344         }
 345 }
 346
 347 static void populate_be_v0_stats(struct be_adapter *adapter)
 348 {
 349         struct be_hw_stats_v0 *hw_stats = hw_stats_from_cmd(adapter);
 350         struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
 351         struct be_rxf_stats_v0 *rxf_stats = &hw_stats->rxf;
 352         struct be_port_rxf_stats_v0 *port_stats =
 353                                         &rxf_stats->port[adapter->port_num];
 354         struct be_drv_stats *drvs = &adapter->drv_stats;
 355
 356         be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
 357         drvs->rx_pause_frames = port_stats->rx_pause_frames;
 358         drvs->rx_crc_errors = port_stats->rx_crc_errors;
 359         drvs->rx_control_frames = port_stats->rx_control_frames;
 360         drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
 361         drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
 362         drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
 363         drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
 364         drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
 365         drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
 366         drvs->rxpp_fifo_overflow_drop = port_stats->rx_fifo_overflow;
 367         drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
 368         drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
 369         drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
 370         drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
 371         drvs->rx_input_fifo_overflow_drop = port_stats->rx_input_fifo_overflow;
 372         drvs->rx_dropped_header_too_small =
 373                 port_stats->rx_dropped_header_too_small;
 374         drvs->rx_address_filtered =
 375                                         port_stats->rx_address_filtered +
 376                                         port_stats->rx_vlan_filtered;
 377         drvs->rx_alignment_symbol_errors =
 378                 port_stats->rx_alignment_symbol_errors;
 379
 380         drvs->tx_pauseframes = port_stats->tx_pauseframes;
 381         drvs->tx_controlframes = port_stats->tx_controlframes;
 382
 383         if (adapter->port_num)
 384                 drvs->jabber_events = rxf_stats->port1_jabber_events;
 385         else
 386                 drvs->jabber_events = rxf_stats->port0_jabber_events;
 387         drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
 388         drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
 389         drvs->forwarded_packets = rxf_stats->forwarded_packets;
 390         drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
 391         drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
 392         drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
 393         adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
 394 }
 395
 396 static void populate_be_v1_stats(struct be_adapter *adapter)
 397 {
 398         struct be_hw_stats_v1 *hw_stats = hw_stats_from_cmd(adapter);
 399         struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
 400         struct be_rxf_stats_v1 *rxf_stats = &hw_stats->rxf;
 401         struct be_port_rxf_stats_v1 *port_stats =
 402                                         &rxf_stats->port[adapter->port_num];
 403         struct be_drv_stats *drvs = &adapter->drv_stats;
 404
 405         be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
 406         drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
 407         drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
 408         drvs->rx_pause_frames = port_stats->rx_pause_frames;
 409         drvs->rx_crc_errors = port_stats->rx_crc_errors;
 410         drvs->rx_control_frames = port_stats->rx_control_frames;
 411         drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
 412         drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
 413         drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
 414         drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
 415         drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
 416         drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
 417         drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
 418         drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
 419         drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
 420         drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
 421         drvs->rx_dropped_header_too_small =
 422                 port_stats->rx_dropped_header_too_small;
 423         drvs->rx_input_fifo_overflow_drop =
 424                 port_stats->rx_input_fifo_overflow_drop;
 425         drvs->rx_address_filtered = port_stats->rx_address_filtered;
 426         drvs->rx_alignment_symbol_errors =
 427                 port_stats->rx_alignment_symbol_errors;
 428         drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
 429         drvs->tx_pauseframes = port_stats->tx_pauseframes;
 430         drvs->tx_controlframes = port_stats->tx_controlframes;
 431         drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
 432         drvs->jabber_events = port_stats->jabber_events;
 433         drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
 434         drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
 435         drvs->forwarded_packets = rxf_stats->forwarded_packets;
 436         drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
 437         drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
 438         drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
 439         adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
 440 }
 441
 442 static void populate_be_v2_stats(struct be_adapter *adapter)
 443 {
 444         struct be_hw_stats_v2 *hw_stats = hw_stats_from_cmd(adapter);
 445         struct be_pmem_stats *pmem_sts = &hw_stats->pmem;
 446         struct be_rxf_stats_v2 *rxf_stats = &hw_stats->rxf;
 447         struct be_port_rxf_stats_v2 *port_stats =
 448                                         &rxf_stats->port[adapter->port_num];
 449         struct be_drv_stats *drvs = &adapter->drv_stats;
 450
 451         be_dws_le_to_cpu(hw_stats, sizeof(*hw_stats));
 452         drvs->pmem_fifo_overflow_drop = port_stats->pmem_fifo_overflow_drop;
 453         drvs->rx_priority_pause_frames = port_stats->rx_priority_pause_frames;
 454         drvs->rx_pause_frames = port_stats->rx_pause_frames;
 455         drvs->rx_crc_errors = port_stats->rx_crc_errors;
 456         drvs->rx_control_frames = port_stats->rx_control_frames;
 457         drvs->rx_in_range_errors = port_stats->rx_in_range_errors;
 458         drvs->rx_frame_too_long = port_stats->rx_frame_too_long;
 459         drvs->rx_dropped_runt = port_stats->rx_dropped_runt;
 460         drvs->rx_ip_checksum_errs = port_stats->rx_ip_checksum_errs;
 461         drvs->rx_tcp_checksum_errs = port_stats->rx_tcp_checksum_errs;
 462         drvs->rx_udp_checksum_errs = port_stats->rx_udp_checksum_errs;
 463         drvs->rx_dropped_tcp_length = port_stats->rx_dropped_tcp_length;
 464         drvs->rx_dropped_too_small = port_stats->rx_dropped_too_small;
 465         drvs->rx_dropped_too_short = port_stats->rx_dropped_too_short;
 466         drvs->rx_out_range_errors = port_stats->rx_out_range_errors;
 467         drvs->rx_dropped_header_too_small =
 468                 port_stats->rx_dropped_header_too_small;
 469         drvs->rx_input_fifo_overflow_drop =
 470                 port_stats->rx_input_fifo_overflow_drop;
 471         drvs->rx_address_filtered = port_stats->rx_address_filtered;
 472         drvs->rx_alignment_symbol_errors =
 473                 port_stats->rx_alignment_symbol_errors;
 474         drvs->rxpp_fifo_overflow_drop = port_stats->rxpp_fifo_overflow_drop;
 475         drvs->tx_pauseframes = port_stats->tx_pauseframes;
 476         drvs->tx_controlframes = port_stats->tx_controlframes;
 477         drvs->tx_priority_pauseframes = port_stats->tx_priority_pauseframes;
 478         drvs->jabber_events = port_stats->jabber_events;
 479         drvs->rx_drops_no_pbuf = rxf_stats->rx_drops_no_pbuf;
 480         drvs->rx_drops_no_erx_descr = rxf_stats->rx_drops_no_erx_descr;
 481         drvs->forwarded_packets = rxf_stats->forwarded_packets;
 482         drvs->rx_drops_mtu = rxf_stats->rx_drops_mtu;
 483         drvs->rx_drops_no_tpre_descr = rxf_stats->rx_drops_no_tpre_descr;
 484         drvs->rx_drops_too_many_frags = rxf_stats->rx_drops_too_many_frags;
 485         adapter->drv_stats.eth_red_drops = pmem_sts->eth_red_drops;
 486         if (be_roce_supported(adapter)) {
 487                 drvs->rx_roce_bytes_lsd = port_stats->roce_bytes_received_lsd;
 488                 drvs->rx_roce_bytes_msd = port_stats->roce_bytes_received_msd;
 489                 drvs->rx_roce_frames = port_stats->roce_frames_received;
 490                 drvs->roce_drops_crc = port_stats->roce_drops_crc;
 491                 drvs->roce_drops_payload_len =
 492                         port_stats->roce_drops_payload_len;
 493         }
 494 }
 495
 496 static void populate_lancer_stats(struct be_adapter *adapter)
 497 {
 498         struct be_drv_stats *drvs = &adapter->drv_stats;
 499         struct lancer_pport_stats *pport_stats = pport_stats_from_cmd(adapter);
 500
 501         be_dws_le_to_cpu(pport_stats, sizeof(*pport_stats));
 502         drvs->rx_pause_frames = pport_stats->rx_pause_frames_lo;
 503         drvs->rx_crc_errors = pport_stats->rx_crc_errors_lo;
 504         drvs->rx_control_frames = pport_stats->rx_control_frames_lo;
 505         drvs->rx_in_range_errors = pport_stats->rx_in_range_errors;
 506         drvs->rx_frame_too_long = pport_stats->rx_frames_too_long_lo;
 507         drvs->rx_dropped_runt = pport_stats->rx_dropped_runt;
 508         drvs->rx_ip_checksum_errs = pport_stats->rx_ip_checksum_errors;
 509         drvs->rx_tcp_checksum_errs = pport_stats->rx_tcp_checksum_errors;
 510         drvs->rx_udp_checksum_errs = pport_stats->rx_udp_checksum_errors;
 511         drvs->rx_dropped_tcp_length =
 512                                 pport_stats->rx_dropped_invalid_tcp_length;
 513         drvs->rx_dropped_too_small = pport_stats->rx_dropped_too_small;
 514         drvs->rx_dropped_too_short = pport_stats->rx_dropped_too_short;
 515         drvs->rx_out_range_errors = pport_stats->rx_out_of_range_errors;
 516         drvs->rx_dropped_header_too_small =
 517                                 pport_stats->rx_dropped_header_too_small;
 518         drvs->rx_input_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
 519         drvs->rx_address_filtered =
 520                                         pport_stats->rx_address_filtered +
 521                                         pport_stats->rx_vlan_filtered;
 522         drvs->rx_alignment_symbol_errors = pport_stats->rx_symbol_errors_lo;
 523         drvs->rxpp_fifo_overflow_drop = pport_stats->rx_fifo_overflow;
 524         drvs->tx_pauseframes = pport_stats->tx_pause_frames_lo;
 525         drvs->tx_controlframes = pport_stats->tx_control_frames_lo;
 526         drvs->jabber_events = pport_stats->rx_jabbers;
 527         drvs->forwarded_packets = pport_stats->num_forwards_lo;
 528         drvs->rx_drops_mtu = pport_stats->rx_drops_mtu_lo;
 529         drvs->rx_drops_too_many_frags =
 530                                 pport_stats->rx_drops_too_many_frags_lo;
 531 }
 532
 533 static void accumulate_16bit_val(u32 *acc, u16 val)
 534 {
 535 #define lo(x)                   (x & 0xFFFF)
 536 #define hi(x)                   (x & 0xFFFF0000)
 537         bool wrapped = val < lo(*acc);
 538         u32 newacc = hi(*acc) + val;
 539
 540         if (wrapped)
 541                 newacc += 65536;
 542         ACCESS_ONCE(*acc) = newacc;
 543 }
 544
 545 static void populate_erx_stats(struct be_adapter *adapter,
 546                                struct be_rx_obj *rxo, u32 erx_stat)
 547 {
 548         if (!BEx_chip(adapter))
 549                 rx_stats(rxo)->rx_drops_no_frags = erx_stat;
 550         else
 551                 /* below erx HW counter can actually wrap around after
 552                  * 65535. Driver accumulates a 32-bit value
 553                  */
 554                 accumulate_16bit_val(&rx_stats(rxo)->rx_drops_no_frags,
 555                                      (u16)erx_stat);
 556 }
 557
 558 void be_parse_stats(struct be_adapter *adapter)
 559 {
 560         struct be_erx_stats_v2 *erx = be_erx_stats_from_cmd(adapter);
 561         struct be_rx_obj *rxo;
 562         int i;
 563         u32 erx_stat;
 564
 565         if (lancer_chip(adapter)) {
 566                 populate_lancer_stats(adapter);
 567         } else {
 568                 if (BE2_chip(adapter))
 569                         populate_be_v0_stats(adapter);
 570                 else if (BE3_chip(adapter))
 571                         /* for BE3 */
 572                         populate_be_v1_stats(adapter);
 573                 else
 574                         populate_be_v2_stats(adapter);
 575
 576                 /* erx_v2 is longer than v0, v1. use v2 for v0, v1 access */
 577                 for_all_rx_queues(adapter, rxo, i) {
 578                         erx_stat = erx->rx_drops_no_fragments[rxo->q.id];
 579                         populate_erx_stats(adapter, rxo, erx_stat);
 580                 }
 581         }
 582 }
 583
 584 static struct rtnl_link_stats64 *be_get_stats64(struct net_device *netdev,
 585                                                 struct rtnl_link_stats64 *stats)
 586 {
 587         struct be_adapter *adapter = netdev_priv(netdev);
 588         struct be_drv_stats *drvs = &adapter->drv_stats;
 589         struct be_rx_obj *rxo;
 590         struct be_tx_obj *txo;
 591         u64 pkts, bytes;
 592         unsigned int start;
 593         int i;
 594
 595         for_all_rx_queues(adapter, rxo, i) {
 596                 const struct be_rx_stats *rx_stats = rx_stats(rxo);
 597
 598                 do {
 599                         start = u64_stats_fetch_begin_irq(&rx_stats->sync);
 600                         pkts = rx_stats(rxo)->rx_pkts;
 601                         bytes = rx_stats(rxo)->rx_bytes;
 602                 } while (u64_stats_fetch_retry_irq(&rx_stats->sync, start));
 603                 stats->rx_packets += pkts;
 604                 stats->rx_bytes += bytes;
 605                 stats->multicast += rx_stats(rxo)->rx_mcast_pkts;
 606                 stats->rx_dropped += rx_stats(rxo)->rx_drops_no_skbs +
 607                                         rx_stats(rxo)->rx_drops_no_frags;
 608         }
 609
 610         for_all_tx_queues(adapter, txo, i) {
 611                 const struct be_tx_stats *tx_stats = tx_stats(txo);
 612
 613                 do {
 614                         start = u64_stats_fetch_begin_irq(&tx_stats->sync);
 615                         pkts = tx_stats(txo)->tx_pkts;
 616                         bytes = tx_stats(txo)->tx_bytes;
 617                 } while (u64_stats_fetch_retry_irq(&tx_stats->sync, start));
 618                 stats->tx_packets += pkts;
 619                 stats->tx_bytes += bytes;
 620         }
 621
 622         /* bad pkts received */
 623         stats->rx_errors = drvs->rx_crc_errors +
 624                 drvs->rx_alignment_symbol_errors +
 625                 drvs->rx_in_range_errors +
 626                 drvs->rx_out_range_errors +
 627                 drvs->rx_frame_too_long +
 628                 drvs->rx_dropped_too_small +
 629                 drvs->rx_dropped_too_short +
 630                 drvs->rx_dropped_header_too_small +
 631                 drvs->rx_dropped_tcp_length +
 632                 drvs->rx_dropped_runt;
 633
 634         /* detailed rx errors */
 635         stats->rx_length_errors = drvs->rx_in_range_errors +
 636                 drvs->rx_out_range_errors +
 637                 drvs->rx_frame_too_long;
 638
 639         stats->rx_crc_errors = drvs->rx_crc_errors;
 640
 641         /* frame alignment errors */
 642         stats->rx_frame_errors = drvs->rx_alignment_symbol_errors;
 643
 644         /* receiver fifo overrun */
 645         /* drops_no_pbuf is no per i/f, it's per BE card */
 646         stats->rx_fifo_errors = drvs->rxpp_fifo_overflow_drop +
 647                                 drvs->rx_input_fifo_overflow_drop +
 648                                 drvs->rx_drops_no_pbuf;
 649         return stats;
 650 }
 651
 652 void be_link_status_update(struct be_adapter *adapter, u8 link_status)
 653 {
 654         struct net_device *netdev = adapter->netdev;
 655
 656         if (!(adapter->flags & BE_FLAGS_LINK_STATUS_INIT)) {
 657                 netif_carrier_off(netdev);
 658                 adapter->flags |= BE_FLAGS_LINK_STATUS_INIT;
 659         }
 660
 661         if (link_status)
 662                 netif_carrier_on(netdev);
 663         else
 664                 netif_carrier_off(netdev);
 665 }
 666
 667 static void be_tx_stats_update(struct be_tx_obj *txo, struct sk_buff *skb)
 668 {
 669         struct be_tx_stats *stats = tx_stats(txo);
 670
 671         u64_stats_update_begin(&stats->sync);
 672         stats->tx_reqs++;
 673         stats->tx_bytes += skb->len;
 674         stats->tx_pkts += (skb_shinfo(skb)->gso_segs ? : 1);
 675         u64_stats_update_end(&stats->sync);
 676 }
 677
 678 /* Returns number of WRBs needed for the skb */
 679 static u32 skb_wrb_cnt(struct sk_buff *skb)
 680 {
 681         /* +1 for the header wrb */
 682         return 1 + (skb_headlen(skb) ? 1 : 0) + skb_shinfo(skb)->nr_frags;
 683 }
 684
 685 static inline void wrb_fill(struct be_eth_wrb *wrb, u64 addr, int len)
 686 {
 687         wrb->frag_pa_hi = cpu_to_le32(upper_32_bits(addr));
 688         wrb->frag_pa_lo = cpu_to_le32(lower_32_bits(addr));
 689         wrb->frag_len = cpu_to_le32(len & ETH_WRB_FRAG_LEN_MASK);
 690         wrb->rsvd0 = 0;
 691 }
 692
 693 /* A dummy wrb is just all zeros. Using a separate routine for dummy-wrb
 694  * to avoid the swap and shift/mask operations in wrb_fill().
 695  */
 696 static inline void wrb_fill_dummy(struct be_eth_wrb *wrb)
 697 {
 698         wrb->frag_pa_hi = 0;
 699         wrb->frag_pa_lo = 0;
 700         wrb->frag_len = 0;
 701         wrb->rsvd0 = 0;
 702 }
 703
 704 static inline u16 be_get_tx_vlan_tag(struct be_adapter *adapter,
 705                                      struct sk_buff *skb)
 706 {
 707         u8 vlan_prio;
 708         u16 vlan_tag;
 709
 710         vlan_tag = skb_vlan_tag_get(skb);
 711         vlan_prio = (vlan_tag & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
 712         /* If vlan priority provided by OS is NOT in available bmap */
 713         if (!(adapter->vlan_prio_bmap & (1 << vlan_prio)))
 714                 vlan_tag = (vlan_tag & ~VLAN_PRIO_MASK) |
 715                                 adapter->recommended_prio;
 716
 717         return vlan_tag;
 718 }
 719
 720 /* Used only for IP tunnel packets */
 721 static u16 skb_inner_ip_proto(struct sk_buff *skb)
 722 {
 723         return (inner_ip_hdr(skb)->version == 4) ?
 724                 inner_ip_hdr(skb)->protocol : inner_ipv6_hdr(skb)->nexthdr;
 725 }
 726
 727 static u16 skb_ip_proto(struct sk_buff *skb)
 728 {
 729         return (ip_hdr(skb)->version == 4) ?
 730                 ip_hdr(skb)->protocol : ipv6_hdr(skb)->nexthdr;
 731 }
 732
 733 static inline bool be_is_txq_full(struct be_tx_obj *txo)
 734 {
 735         return atomic_read(&txo->q.used) + BE_MAX_TX_FRAG_COUNT >= txo->q.len;
 736 }
 737
 738 static inline bool be_can_txq_wake(struct be_tx_obj *txo)
 739 {
 740         return atomic_read(&txo->q.used) < txo->q.len / 2;
 741 }
 742
 743 static inline bool be_is_tx_compl_pending(struct be_tx_obj *txo)
 744 {
 745         return atomic_read(&txo->q.used) > txo->pend_wrb_cnt;
 746 }
 747
 748 static void be_get_wrb_params_from_skb(struct be_adapter *adapter,
 749                                        struct sk_buff *skb,
 750                                        struct be_wrb_params *wrb_params)
 751 {
 752         u16 proto;
 753
 754         if (skb_is_gso(skb)) {
 755                 BE_WRB_F_SET(wrb_params->features, LSO, 1);
 756                 wrb_params->lso_mss = skb_shinfo(skb)->gso_size;
 757                 if (skb_is_gso_v6(skb) && !lancer_chip(adapter))
 758                         BE_WRB_F_SET(wrb_params->features, LSO6, 1);
 759         } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
 760                 if (skb->encapsulation) {
 761                         BE_WRB_F_SET(wrb_params->features, IPCS, 1);
 762                         proto = skb_inner_ip_proto(skb);
 763                 } else {
 764                         proto = skb_ip_proto(skb);
 765                 }
 766                 if (proto == IPPROTO_TCP)
 767                         BE_WRB_F_SET(wrb_params->features, TCPCS, 1);
 768                 else if (proto == IPPROTO_UDP)
 769                         BE_WRB_F_SET(wrb_params->features, UDPCS, 1);
 770         }
 771
 772         if (skb_vlan_tag_present(skb)) {
 773                 BE_WRB_F_SET(wrb_params->features, VLAN, 1);
 774                 wrb_params->vlan_tag = be_get_tx_vlan_tag(adapter, skb);
 775         }
 776
 777         BE_WRB_F_SET(wrb_params->features, CRC, 1);
 778 }
 779
 780 static void wrb_fill_hdr(struct be_adapter *adapter,
 781                          struct be_eth_hdr_wrb *hdr,
 782                          struct be_wrb_params *wrb_params,
 783                          struct sk_buff *skb)
 784 {
 785         memset(hdr, 0, sizeof(*hdr));
 786
 787         SET_TX_WRB_HDR_BITS(crc, hdr,
 788                             BE_WRB_F_GET(wrb_params->features, CRC));
 789         SET_TX_WRB_HDR_BITS(ipcs, hdr,
 790                             BE_WRB_F_GET(wrb_params->features, IPCS));
 791         SET_TX_WRB_HDR_BITS(tcpcs, hdr,
 792                             BE_WRB_F_GET(wrb_params->features, TCPCS));
 793         SET_TX_WRB_HDR_BITS(udpcs, hdr,
 794                             BE_WRB_F_GET(wrb_params->features, UDPCS));
 795
 796         SET_TX_WRB_HDR_BITS(lso, hdr,
 797                             BE_WRB_F_GET(wrb_params->features, LSO));
 798         SET_TX_WRB_HDR_BITS(lso6, hdr,
 799                             BE_WRB_F_GET(wrb_params->features, LSO6));
 800         SET_TX_WRB_HDR_BITS(lso_mss, hdr, wrb_params->lso_mss);
 801
 802         /* Hack to skip HW VLAN tagging needs evt = 1, compl = 0. When this
 803          * hack is not needed, the evt bit is set while ringing DB.
 804          */
 805         SET_TX_WRB_HDR_BITS(event, hdr,
 806                             BE_WRB_F_GET(wrb_params->features, VLAN_SKIP_HW));
 807         SET_TX_WRB_HDR_BITS(vlan, hdr,
 808                             BE_WRB_F_GET(wrb_params->features, VLAN));
 809         SET_TX_WRB_HDR_BITS(vlan_tag, hdr, wrb_params->vlan_tag);
 810
 811         SET_TX_WRB_HDR_BITS(num_wrb, hdr, skb_wrb_cnt(skb));
 812         SET_TX_WRB_HDR_BITS(len, hdr, skb->len);
 813 }
 814
 815 static void unmap_tx_frag(struct device *dev, struct be_eth_wrb *wrb,
 816                           bool unmap_single)
 817 {
 818         dma_addr_t dma;
 819         u32 frag_len = le32_to_cpu(wrb->frag_len);
 820
 821
 822         dma = (u64)le32_to_cpu(wrb->frag_pa_hi) << 32 |
 823                 (u64)le32_to_cpu(wrb->frag_pa_lo);
 824         if (frag_len) {
 825                 if (unmap_single)
 826                         dma_unmap_single(dev, dma, frag_len, DMA_TO_DEVICE);
 827                 else
 828                         dma_unmap_page(dev, dma, frag_len, DMA_TO_DEVICE);
 829         }
 830 }
 831
 832 /* Grab a WRB header for xmit */
 833 static u16 be_tx_get_wrb_hdr(struct be_tx_obj *txo)
 834 {
 835         u16 head = txo->q.head;
 836
 837         queue_head_inc(&txo->q);
 838         return head;
 839 }
 840
 841 /* Set up the WRB header for xmit */
 842 static void be_tx_setup_wrb_hdr(struct be_adapter *adapter,
 843                                 struct be_tx_obj *txo,
 844                                 struct be_wrb_params *wrb_params,
 845                                 struct sk_buff *skb, u16 head)
 846 {
 847         u32 num_frags = skb_wrb_cnt(skb);
 848         struct be_queue_info *txq = &txo->q;
 849         struct be_eth_hdr_wrb *hdr = queue_index_node(txq, head);
 850
 851         wrb_fill_hdr(adapter, hdr, wrb_params, skb);
 852         be_dws_cpu_to_le(hdr, sizeof(*hdr));
 853
 854         BUG_ON(txo->sent_skb_list[head]);
 855         txo->sent_skb_list[head] = skb;
 856         txo->last_req_hdr = head;
 857         atomic_add(num_frags, &txq->used);
 858         txo->last_req_wrb_cnt = num_frags;
 859         txo->pend_wrb_cnt += num_frags;
 860 }
 861
 862 /* Setup a WRB fragment (buffer descriptor) for xmit */
 863 static void be_tx_setup_wrb_frag(struct be_tx_obj *txo, dma_addr_t busaddr,
 864                                  int len)
 865 {
 866         struct be_eth_wrb *wrb;
 867         struct be_queue_info *txq = &txo->q;
 868
 869         wrb = queue_head_node(txq);
 870         wrb_fill(wrb, busaddr, len);
 871         queue_head_inc(txq);
 872 }
 873
 874 /* Bring the queue back to the state it was in before be_xmit_enqueue() routine
 875  * was invoked. The producer index is restored to the previous packet and the
 876  * WRBs of the current packet are unmapped. Invoked to handle tx setup errors.
 877  */
 878 static void be_xmit_restore(struct be_adapter *adapter,
 879                             struct be_tx_obj *txo, u16 head, bool map_single,
 880                             u32 copied)
 881 {
 882         struct device *dev;
 883         struct be_eth_wrb *wrb;
 884         struct be_queue_info *txq = &txo->q;
 885
 886         dev = &adapter->pdev->dev;
 887         txq->head = head;
 888
 889         /* skip the first wrb (hdr); it's not mapped */
 890         queue_head_inc(txq);
 891         while (copied) {
 892                 wrb = queue_head_node(txq);
 893                 unmap_tx_frag(dev, wrb, map_single);
 894                 map_single = false;
 895                 copied -= le32_to_cpu(wrb->frag_len);
 896                 queue_head_inc(txq);
 897         }
 898
 899         txq->head = head;
 900 }
 901
 902 /* Enqueue the given packet for transmit. This routine allocates WRBs for the
 903  * packet, dma maps the packet buffers and sets up the WRBs. Returns the number
 904  * of WRBs used up by the packet.
 905  */
 906 static u32 be_xmit_enqueue(struct be_adapter *adapter, struct be_tx_obj *txo,
 907                            struct sk_buff *skb,
 908                            struct be_wrb_params *wrb_params)
 909 {
 910         u32 i, copied = 0, wrb_cnt = skb_wrb_cnt(skb);
 911         struct device *dev = &adapter->pdev->dev;
 912         struct be_queue_info *txq = &txo->q;
 913         bool map_single = false;
 914         u16 head = txq->head;
 915         dma_addr_t busaddr;
 916         int len;
 917
 918         head = be_tx_get_wrb_hdr(txo);
 919
 920         if (skb->len > skb->data_len) {
 921                 len = skb_headlen(skb);
 922
 923                 busaddr = dma_map_single(dev, skb->data, len, DMA_TO_DEVICE);
 924                 if (dma_mapping_error(dev, busaddr))
 925                         goto dma_err;
 926                 map_single = true;
 927                 be_tx_setup_wrb_frag(txo, busaddr, len);
 928                 copied += len;
 929         }
 930
 931         for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
 932                 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
 933                 len = skb_frag_size(frag);
 934
 935                 busaddr = skb_frag_dma_map(dev, frag, 0, len, DMA_TO_DEVICE);
 936                 if (dma_mapping_error(dev, busaddr))
 937                         goto dma_err;
 938                 be_tx_setup_wrb_frag(txo, busaddr, len);
 939                 copied += len;
 940         }
 941
 942         be_tx_setup_wrb_hdr(adapter, txo, wrb_params, skb, head);
 943
 944         be_tx_stats_update(txo, skb);
 945         return wrb_cnt;
 946
 947 dma_err:
 948         adapter->drv_stats.dma_map_errors++;
 949         be_xmit_restore(adapter, txo, head, map_single, copied);
 950         return 0;
 951 }
 952
 953 static inline int qnq_async_evt_rcvd(struct be_adapter *adapter)
 954 {
 955         return adapter->flags & BE_FLAGS_QNQ_ASYNC_EVT_RCVD;
 956 }
 957
 958 static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
 959                                              struct sk_buff *skb,
 960                                              struct be_wrb_params
 961                                              *wrb_params)
 962 {
 963         u16 vlan_tag = 0;
 964
 965         skb = skb_share_check(skb, GFP_ATOMIC);
 966         if (unlikely(!skb))
 967                 return skb;
 968
 969         if (skb_vlan_tag_present(skb))
 970                 vlan_tag = be_get_tx_vlan_tag(adapter, skb);
 971
 972         if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
 973                 if (!vlan_tag)
 974                         vlan_tag = adapter->pvid;
 975                 /* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
 976                  * skip VLAN insertion
 977                  */
 978                 BE_WRB_F_SET(wrb_params->features, VLAN_SKIP_HW, 1);
 979         }
 980
 981         if (vlan_tag) {
 982                 skb = vlan_insert_tag_set_proto(skb, htons(ETH_P_8021Q),
 983                                                 vlan_tag);
 984                 if (unlikely(!skb))
 985                         return skb;
 986                 skb->vlan_tci = 0;
 987         }
 988
 989         /* Insert the outer VLAN, if any */
 990         if (adapter->qnq_vid) {
 991                 vlan_tag = adapter->qnq_vid;
 992                 skb = vlan_insert_tag_set_proto(skb, htons(ETH_P_8021Q),
 993                                                 vlan_tag);
 994                 if (unlikely(!skb))
 995                         return skb;
 996                 BE_WRB_F_SET(wrb_params->features, VLAN_SKIP_HW, 1);
 997         }
 998
 999         return skb;
1000 }
1001
1002 static bool be_ipv6_exthdr_check(struct sk_buff *skb)
1003 {
1004         struct ethhdr *eh = (struct ethhdr *)skb->data;
1005         u16 offset = ETH_HLEN;
1006
1007         if (eh->h_proto == htons(ETH_P_IPV6)) {
1008                 struct ipv6hdr *ip6h = (struct ipv6hdr *)(skb->data + offset);
1009
1010                 offset += sizeof(struct ipv6hdr);
1011                 if (ip6h->nexthdr != NEXTHDR_TCP &&
1012                     ip6h->nexthdr != NEXTHDR_UDP) {
1013                         struct ipv6_opt_hdr *ehdr =
1014                                 (struct ipv6_opt_hdr *)(skb->data + offset);
1015
1016                         /* offending pkt: 2nd byte following IPv6 hdr is 0xff */
1017                         if (ehdr->hdrlen == 0xff)
1018                                 return true;
1019                 }
1020         }
1021         return false;
1022 }
1023
1024 static int be_vlan_tag_tx_chk(struct be_adapter *adapter, struct sk_buff *skb)
1025 {
1026         return skb_vlan_tag_present(skb) || adapter->pvid || adapter->qnq_vid;
1027 }
1028
1029 static int be_ipv6_tx_stall_chk(struct be_adapter *adapter, struct sk_buff *skb)
1030 {
1031         return BE3_chip(adapter) && be_ipv6_exthdr_check(skb);
1032 }
1033
1034 static struct sk_buff *be_lancer_xmit_workarounds(struct be_adapter *adapter,
1035                                                   struct sk_buff *skb,
1036                                                   struct be_wrb_params
1037                                                   *wrb_params)
1038 {
1039         struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
1040         unsigned int eth_hdr_len;
1041         struct iphdr *ip;
1042
1043         /* For padded packets, BE HW modifies tot_len field in IP header
1044          * incorrecly when VLAN tag is inserted by HW.
1045          * For padded packets, Lancer computes incorrect checksum.
1046          */
1047         eth_hdr_len = ntohs(skb->protocol) == ETH_P_8021Q ?
1048                                                 VLAN_ETH_HLEN : ETH_HLEN;
1049         if (skb->len <= 60 &&
1050             (lancer_chip(adapter) || skb_vlan_tag_present(skb)) &&
1051             is_ipv4_pkt(skb)) {
1052                 ip = (struct iphdr *)ip_hdr(skb);
1053                 pskb_trim(skb, eth_hdr_len + ntohs(ip->tot_len));
1054         }
1055
1056         /* If vlan tag is already inlined in the packet, skip HW VLAN
1057          * tagging in pvid-tagging mode
1058          */
1059         if (be_pvid_tagging_enabled(adapter) &&
1060             veh->h_vlan_proto == htons(ETH_P_8021Q))
1061                 BE_WRB_F_SET(wrb_params->features, VLAN_SKIP_HW, 1);
1062
1063         /* HW has a bug wherein it will calculate CSUM for VLAN
1064          * pkts even though it is disabled.
1065          * Manually insert VLAN in pkt.
1066          */
1067         if (skb->ip_summed != CHECKSUM_PARTIAL &&
1068             skb_vlan_tag_present(skb)) {
1069                 skb = be_insert_vlan_in_pkt(adapter, skb, wrb_params);
1070                 if (unlikely(!skb))
1071                         goto err;
1072         }
1073
1074         /* HW may lockup when VLAN HW tagging is requested on
1075          * certain ipv6 packets. Drop such pkts if the HW workaround to
1076          * skip HW tagging is not enabled by FW.
1077          */
1078         if (unlikely(be_ipv6_tx_stall_chk(adapter, skb) &&
1079                      (adapter->pvid || adapter->qnq_vid) &&
1080                      !qnq_async_evt_rcvd(adapter)))
1081                 goto tx_drop;
1082
1083         /* Manual VLAN tag insertion to prevent:
1084          * ASIC lockup when the ASIC inserts VLAN tag into
1085          * certain ipv6 packets. Insert VLAN tags in driver,
1086          * and set event, completion, vlan bits accordingly
1087          * in the Tx WRB.
1088          */
1089         if (be_ipv6_tx_stall_chk(adapter, skb) &&
1090             be_vlan_tag_tx_chk(adapter, skb)) {
1091                 skb = be_insert_vlan_in_pkt(adapter, skb, wrb_params);
1092                 if (unlikely(!skb))
1093                         goto err;
1094         }
1095
1096         return skb;
1097 tx_drop:
1098         dev_kfree_skb_any(skb);
1099 err:
1100         return NULL;
1101 }
1102
1103 static struct sk_buff *be_xmit_workarounds(struct be_adapter *adapter,
1104                                            struct sk_buff *skb,
1105                                            struct be_wrb_params *wrb_params)
1106 {
1107         /* Lancer, SH-R ASICs have a bug wherein Packets that are 32 bytes or
1108          * less may cause a transmit stall on that port. So the work-around is
1109          * to pad short packets (<= 32 bytes) to a 36-byte length.
1110          */
1111         if (unlikely(!BEx_chip(adapter) && skb->len <= 32)) {
1112                 if (skb_put_padto(skb, 36))
1113                         return NULL;
1114         }
1115
1116         if (BEx_chip(adapter) || lancer_chip(adapter)) {
1117                 skb = be_lancer_xmit_workarounds(adapter, skb, wrb_params);
1118                 if (!skb)
1119                         return NULL;
1120         }
1121
1122         return skb;
1123 }
1124
1125 static void be_xmit_flush(struct be_adapter *adapter, struct be_tx_obj *txo)
1126 {
1127         struct be_queue_info *txq = &txo->q;
1128         struct be_eth_hdr_wrb *hdr = queue_index_node(txq, txo->last_req_hdr);
1129
1130         /* Mark the last request eventable if it hasn't been marked already */
1131         if (!(hdr->dw[2] & cpu_to_le32(TX_HDR_WRB_EVT)))
1132                 hdr->dw[2] |= cpu_to_le32(TX_HDR_WRB_EVT | TX_HDR_WRB_COMPL);
1133
1134         /* compose a dummy wrb if there are odd set of wrbs to notify */
1135         if (!lancer_chip(adapter) && (txo->pend_wrb_cnt & 1)) {
1136                 wrb_fill_dummy(queue_head_node(txq));
1137                 queue_head_inc(txq);
1138                 atomic_inc(&txq->used);
1139                 txo->pend_wrb_cnt++;
1140                 hdr->dw[2] &= ~cpu_to_le32(TX_HDR_WRB_NUM_MASK <<
1141                                            TX_HDR_WRB_NUM_SHIFT);
1142                 hdr->dw[2] |= cpu_to_le32((txo->last_req_wrb_cnt + 1) <<
1143                                           TX_HDR_WRB_NUM_SHIFT);
1144         }
1145         be_txq_notify(adapter, txo, txo->pend_wrb_cnt);
1146         txo->pend_wrb_cnt = 0;
1147 }
1148
1149 static netdev_tx_t be_xmit(struct sk_buff *skb, struct net_device *netdev)
1150 {
1151         struct be_adapter *adapter = netdev_priv(netdev);
1152         u16 q_idx = skb_get_queue_mapping(skb);
1153         struct be_tx_obj *txo = &adapter->tx_obj[q_idx];
1154         struct be_wrb_params wrb_params = { 0 };
1155         bool flush = !skb->xmit_more;
1156         u16 wrb_cnt;
1157
1158         skb = be_xmit_workarounds(adapter, skb, &wrb_params);
1159         if (unlikely(!skb))
1160                 goto drop;
1161
1162         be_get_wrb_params_from_skb(adapter, skb, &wrb_params);
1163
1164         wrb_cnt = be_xmit_enqueue(adapter, txo, skb, &wrb_params);
1165         if (unlikely(!wrb_cnt)) {
1166                 dev_kfree_skb_any(skb);
1167                 goto drop;
1168         }
1169
1170         if (be_is_txq_full(txo)) {
1171                 netif_stop_subqueue(netdev, q_idx);
1172                 tx_stats(txo)->tx_stops++;
1173         }
1174
1175         if (flush || __netif_subqueue_stopped(netdev, q_idx))
1176                 be_xmit_flush(adapter, txo);
1177
1178         return NETDEV_TX_OK;
1179 drop:
1180         tx_stats(txo)->tx_drv_drops++;
1181         /* Flush the already enqueued tx requests */
1182         if (flush && txo->pend_wrb_cnt)
1183                 be_xmit_flush(adapter, txo);
1184
1185         return NETDEV_TX_OK;
1186 }
1187
1188 static int be_change_mtu(struct net_device *netdev, int new_mtu)
1189 {
1190         struct be_adapter *adapter = netdev_priv(netdev);
1191         struct device *dev = &adapter->pdev->dev;
1192
1193         if (new_mtu < BE_MIN_MTU || new_mtu > BE_MAX_MTU) {
1194                 dev_info(dev, "MTU must be between %d and %d bytes\n",
1195                          BE_MIN_MTU, BE_MAX_MTU);
1196                 return -EINVAL;
1197         }
1198
1199         dev_info(dev, "MTU changed from %d to %d bytes\n",
1200                  netdev->mtu, new_mtu);
1201         netdev->mtu = new_mtu;
1202         return 0;
1203 }
1204
1205 static inline bool be_in_all_promisc(struct be_adapter *adapter)
1206 {
1207         return (adapter->if_flags & BE_IF_FLAGS_ALL_PROMISCUOUS) ==
1208                         BE_IF_FLAGS_ALL_PROMISCUOUS;
1209 }
1210
1211 static int be_set_vlan_promisc(struct be_adapter *adapter)
1212 {
1213         struct device *dev = &adapter->pdev->dev;
1214         int status;
1215
1216         if (adapter->if_flags & BE_IF_FLAGS_VLAN_PROMISCUOUS)
1217                 return 0;
1218
1219         status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_VLAN_PROMISCUOUS, ON);
1220         if (!status) {
1221                 dev_info(dev, "Enabled VLAN promiscuous mode\n");
1222                 adapter->if_flags |= BE_IF_FLAGS_VLAN_PROMISCUOUS;
1223         } else {
1224                 dev_err(dev, "Failed to enable VLAN promiscuous mode\n");
1225         }
1226         return status;
1227 }
1228
1229 static int be_clear_vlan_promisc(struct be_adapter *adapter)
1230 {
1231         struct device *dev = &adapter->pdev->dev;
1232         int status;
1233
1234         status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_VLAN_PROMISCUOUS, OFF);
1235         if (!status) {
1236                 dev_info(dev, "Disabling VLAN promiscuous mode\n");
1237                 adapter->if_flags &= ~BE_IF_FLAGS_VLAN_PROMISCUOUS;
1238         }
1239         return status;
1240 }
1241
1242 /*
1243  * A max of 64 (BE_NUM_VLANS_SUPPORTED) vlans can be configured in BE.
1244  * If the user configures more, place BE in vlan promiscuous mode.
1245  */
1246 static int be_vid_config(struct be_adapter *adapter)
1247 {
1248         struct device *dev = &adapter->pdev->dev;
1249         u16 vids[BE_NUM_VLANS_SUPPORTED];
1250         u16 num = 0, i = 0;
1251         int status = 0;
1252
1253         /* No need to further configure vids if in promiscuous mode */
1254         if (be_in_all_promisc(adapter))
1255                 return 0;
1256
1257         if (adapter->vlans_added > be_max_vlans(adapter))
1258                 return be_set_vlan_promisc(adapter);
1259
1260         /* Construct VLAN Table to give to HW */
1261         for_each_set_bit(i, adapter->vids, VLAN_N_VID)
1262                 vids[num++] = cpu_to_le16(i);
1263
1264         status = be_cmd_vlan_config(adapter, adapter->if_handle, vids, num, 0);
1265         if (status) {
1266                 dev_err(dev, "Setting HW VLAN filtering failed\n");
1267                 /* Set to VLAN promisc mode as setting VLAN filter failed */
1268                 if (addl_status(status) ==
1269                                 MCC_ADDL_STATUS_INSUFFICIENT_RESOURCES)
1270                         return be_set_vlan_promisc(adapter);
1271         } else if (adapter->if_flags & BE_IF_FLAGS_VLAN_PROMISCUOUS) {
1272                 status = be_clear_vlan_promisc(adapter);
1273         }
1274         return status;
1275 }
1276
1277 static int be_vlan_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
1278 {
1279         struct be_adapter *adapter = netdev_priv(netdev);
1280         int status = 0;
1281
1282         /* Packets with VID 0 are always received by Lancer by default */
1283         if (lancer_chip(adapter) && vid == 0)
1284                 return status;
1285
1286         if (test_bit(vid, adapter->vids))
1287                 return status;
1288
1289         set_bit(vid, adapter->vids);
1290         adapter->vlans_added++;
1291
1292         status = be_vid_config(adapter);
1293         if (status) {
1294                 adapter->vlans_added--;
1295                 clear_bit(vid, adapter->vids);
1296         }
1297
1298         return status;
1299 }
1300
1301 static int be_vlan_rem_vid(struct net_device *netdev, __be16 proto, u16 vid)
1302 {
1303         struct be_adapter *adapter = netdev_priv(netdev);
1304
1305         /* Packets with VID 0 are always received by Lancer by default */
1306         if (lancer_chip(adapter) && vid == 0)
1307                 return 0;
1308
1309         clear_bit(vid, adapter->vids);
1310         adapter->vlans_added--;
1311
1312         return be_vid_config(adapter);
1313 }
1314
1315 static void be_clear_all_promisc(struct be_adapter *adapter)
1316 {
1317         be_cmd_rx_filter(adapter, BE_IF_FLAGS_ALL_PROMISCUOUS, OFF);
1318         adapter->if_flags &= ~BE_IF_FLAGS_ALL_PROMISCUOUS;
1319 }
1320
1321 static void be_set_all_promisc(struct be_adapter *adapter)
1322 {
1323         be_cmd_rx_filter(adapter, BE_IF_FLAGS_ALL_PROMISCUOUS, ON);
1324         adapter->if_flags |= BE_IF_FLAGS_ALL_PROMISCUOUS;
1325 }
1326
1327 static void be_set_mc_promisc(struct be_adapter *adapter)
1328 {
1329         int status;
1330
1331         if (adapter->if_flags & BE_IF_FLAGS_MCAST_PROMISCUOUS)
1332                 return;
1333
1334         status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_MCAST_PROMISCUOUS, ON);
1335         if (!status)
1336                 adapter->if_flags |= BE_IF_FLAGS_MCAST_PROMISCUOUS;
1337 }
1338
1339 static void be_set_mc_list(struct be_adapter *adapter)
1340 {
1341         int status;
1342
1343         status = be_cmd_rx_filter(adapter, BE_IF_FLAGS_MULTICAST, ON);
1344         if (!status)
1345                 adapter->if_flags &= ~BE_IF_FLAGS_MCAST_PROMISCUOUS;
1346         else
1347                 be_set_mc_promisc(adapter);
1348 }
1349
1350 static void be_set_uc_list(struct be_adapter *adapter)
1351 {
1352         struct netdev_hw_addr *ha;
1353         int i = 1; /* First slot is claimed by the Primary MAC */
1354
1355         for (; adapter->uc_macs > 0; adapter->uc_macs--, i++)
1356                 be_cmd_pmac_del(adapter, adapter->if_handle,
1357                                 adapter->pmac_id[i], 0);
1358
1359         if (netdev_uc_count(adapter->netdev) > be_max_uc(adapter)) {
1360                 be_set_all_promisc(adapter);
1361                 return;
1362         }
1363
1364         netdev_for_each_uc_addr(ha, adapter->netdev) {
1365                 adapter->uc_macs++; /* First slot is for Primary MAC */
1366                 be_cmd_pmac_add(adapter, (u8 *)ha->addr, adapter->if_handle,
1367                                 &adapter->pmac_id[adapter->uc_macs], 0);
1368         }
1369 }
1370
1371 static void be_clear_uc_list(struct be_adapter *adapter)
1372 {
1373         int i;
1374
1375         for (i = 1; i < (adapter->uc_macs + 1); i++)
1376                 be_cmd_pmac_del(adapter, adapter->if_handle,
1377                                 adapter->pmac_id[i], 0);
1378         adapter->uc_macs = 0;
1379 }
1380
1381 static void be_set_rx_mode(struct net_device *netdev)
1382 {
1383         struct be_adapter *adapter = netdev_priv(netdev);
1384
1385         if (netdev->flags & IFF_PROMISC) {
1386                 be_set_all_promisc(adapter);
1387                 return;
1388         }
1389
1390         /* Interface was previously in promiscuous mode; disable it */
1391         if (be_in_all_promisc(adapter)) {
1392                 be_clear_all_promisc(adapter);
1393                 if (adapter->vlans_added)
1394                         be_vid_config(adapter);
1395         }
1396
1397         /* Enable multicast promisc if num configured exceeds what we support */
1398         if (netdev->flags & IFF_ALLMULTI ||
1399             netdev_mc_count(netdev) > be_max_mc(adapter)) {
1400                 be_set_mc_promisc(adapter);
1401                 return;
1402         }
1403
1404         if (netdev_uc_count(netdev) != adapter->uc_macs)
1405                 be_set_uc_list(adapter);
1406
1407         be_set_mc_list(adapter);
1408 }
1409
1410 static int be_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1411 {
1412         struct be_adapter *adapter = netdev_priv(netdev);
1413         struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1414         int status;
1415
1416         if (!sriov_enabled(adapter))
1417                 return -EPERM;
1418
1419         if (!is_valid_ether_addr(mac) || vf >= adapter->num_vfs)
1420                 return -EINVAL;
1421
1422         /* Proceed further only if user provided MAC is different
1423          * from active MAC
1424          */
1425         if (ether_addr_equal(mac, vf_cfg->mac_addr))
1426                 return 0;
1427
1428         if (BEx_chip(adapter)) {
1429                 be_cmd_pmac_del(adapter, vf_cfg->if_handle, vf_cfg->pmac_id,
1430                                 vf + 1);
1431
1432                 status = be_cmd_pmac_add(adapter, mac, vf_cfg->if_handle,
1433                                          &vf_cfg->pmac_id, vf + 1);
1434         } else {
1435                 status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
1436                                         vf + 1);
1437         }
1438
1439         if (status) {
1440                 dev_err(&adapter->pdev->dev, "MAC %pM set on VF %d Failed: %#x",
1441                         mac, vf, status);
1442                 return be_cmd_status(status);
1443         }
1444
1445         ether_addr_copy(vf_cfg->mac_addr, mac);
1446
1447         return 0;
1448 }
1449
1450 static int be_get_vf_config(struct net_device *netdev, int vf,
1451                             struct ifla_vf_info *vi)
1452 {
1453         struct be_adapter *adapter = netdev_priv(netdev);
1454         struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1455
1456         if (!sriov_enabled(adapter))
1457                 return -EPERM;
1458
1459         if (vf >= adapter->num_vfs)
1460                 return -EINVAL;
1461
1462         vi->vf = vf;
1463         vi->max_tx_rate = vf_cfg->tx_rate;
1464         vi->min_tx_rate = 0;
1465         vi->vlan = vf_cfg->vlan_tag & VLAN_VID_MASK;
1466         vi->qos = vf_cfg->vlan_tag >> VLAN_PRIO_SHIFT;
1467         memcpy(&vi->mac, vf_cfg->mac_addr, ETH_ALEN);
1468         vi->linkstate = adapter->vf_cfg[vf].plink_tracking;
1469
1470         return 0;
1471 }
1472
1473 static int be_set_vf_tvt(struct be_adapter *adapter, int vf, u16 vlan)
1474 {
1475         struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1476         u16 vids[BE_NUM_VLANS_SUPPORTED];
1477         int vf_if_id = vf_cfg->if_handle;
1478         int status;
1479
1480         /* Enable Transparent VLAN Tagging */
1481         status = be_cmd_set_hsw_config(adapter, vlan, vf + 1, vf_if_id, 0);
1482         if (status)
1483                 return status;
1484
1485         /* Clear pre-programmed VLAN filters on VF if any, if TVT is enabled */
1486         vids[0] = 0;
1487         status = be_cmd_vlan_config(adapter, vf_if_id, vids, 1, vf + 1);
1488         if (!status)
1489                 dev_info(&adapter->pdev->dev,
1490                          "Cleared guest VLANs on VF%d", vf);
1491
1492         /* After TVT is enabled, disallow VFs to program VLAN filters */
1493         if (vf_cfg->privileges & BE_PRIV_FILTMGMT) {
1494                 status = be_cmd_set_fn_privileges(adapter, vf_cfg->privileges &
1495                                                   ~BE_PRIV_FILTMGMT, vf + 1);
1496                 if (!status)
1497                         vf_cfg->privileges &= ~BE_PRIV_FILTMGMT;
1498         }
1499         return 0;
1500 }
1501
1502 static int be_clear_vf_tvt(struct be_adapter *adapter, int vf)
1503 {
1504         struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1505         struct device *dev = &adapter->pdev->dev;
1506         int status;
1507
1508         /* Reset Transparent VLAN Tagging. */
1509         status = be_cmd_set_hsw_config(adapter, BE_RESET_VLAN_TAG_ID, vf + 1,
1510                                        vf_cfg->if_handle, 0);
1511         if (status)
1512                 return status;
1513
1514         /* Allow VFs to program VLAN filtering */
1515         if (!(vf_cfg->privileges & BE_PRIV_FILTMGMT)) {
1516                 status = be_cmd_set_fn_privileges(adapter, vf_cfg->privileges |
1517                                                   BE_PRIV_FILTMGMT, vf + 1);
1518                 if (!status) {
1519                         vf_cfg->privileges |= BE_PRIV_FILTMGMT;
1520                         dev_info(dev, "VF%d: FILTMGMT priv enabled", vf);
1521                 }
1522         }
1523
1524         dev_info(dev,
1525                  "Disable/re-enable i/f in VM to clear Transparent VLAN tag");
1526         return 0;
1527 }
1528
1529 static int be_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1530 {
1531         struct be_adapter *adapter = netdev_priv(netdev);
1532         struct be_vf_cfg *vf_cfg = &adapter->vf_cfg[vf];
1533         int status;
1534
1535         if (!sriov_enabled(adapter))
1536                 return -EPERM;
1537
1538         if (vf >= adapter->num_vfs || vlan > 4095 || qos > 7)
1539                 return -EINVAL;
1540
1541         if (vlan || qos) {
1542                 vlan |= qos << VLAN_PRIO_SHIFT;
1543                 status = be_set_vf_tvt(adapter, vf, vlan);
1544         } else {
1545                 status = be_clear_vf_tvt(adapter, vf);
1546         }
1547
1548         if (status) {
1549                 dev_err(&adapter->pdev->dev,
1550                         "VLAN %d config on VF %d failed : %#x\n", vlan, vf,
1551                         status);
1552                 return be_cmd_status(status);
1553         }
1554
1555         vf_cfg->vlan_tag = vlan;
1556         return 0;
1557 }
1558
1559 static int be_set_vf_tx_rate(struct net_device *netdev, int vf,
1560                              int min_tx_rate, int max_tx_rate)
1561 {
1562         struct be_adapter *adapter = netdev_priv(netdev);
1563         struct device *dev = &adapter->pdev->dev;
1564         int percent_rate, status = 0;
1565         u16 link_speed = 0;
1566         u8 link_status;
1567
1568         if (!sriov_enabled(adapter))
1569                 return -EPERM;
1570
1571         if (vf >= adapter->num_vfs)
1572                 return -EINVAL;
1573
1574         if (min_tx_rate)
1575                 return -EINVAL;
1576
1577         if (!max_tx_rate)
1578                 goto config_qos;
1579
1580         status = be_cmd_link_status_query(adapter, &link_speed,
1581                                           &link_status, 0);
1582         if (status)
1583                 goto err;
1584
1585         if (!link_status) {
1586                 dev_err(dev, "TX-rate setting not allowed when link is down\n");
1587                 status = -ENETDOWN;
1588                 goto err;
1589         }
1590
1591         if (max_tx_rate < 100 || max_tx_rate > link_speed) {
1592                 dev_err(dev, "TX-rate must be between 100 and %d Mbps\n",
1593                         link_speed);
1594                 status = -EINVAL;
1595                 goto err;
1596         }
1597
1598         /* On Skyhawk the QOS setting must be done only as a % value */
1599         percent_rate = link_speed / 100;
1600         if (skyhawk_chip(adapter) && (max_tx_rate % percent_rate)) {
1601                 dev_err(dev, "TX-rate must be a multiple of %d Mbps\n",
1602                         percent_rate);
1603                 status = -EINVAL;
1604                 goto err;
1605         }
1606
1607 config_qos:
1608         status = be_cmd_config_qos(adapter, max_tx_rate, link_speed, vf + 1);
1609         if (status)
1610                 goto err;
1611
1612         adapter->vf_cfg[vf].tx_rate = max_tx_rate;
1613         return 0;
1614
1615 err:
1616         dev_err(dev, "TX-rate setting of %dMbps on VF%d failed\n",
1617                 max_tx_rate, vf);
1618         return be_cmd_status(status);
1619 }
1620
1621 static int be_set_vf_link_state(struct net_device *netdev, int vf,
1622                                 int link_state)
1623 {
1624         struct be_adapter *adapter = netdev_priv(netdev);
1625         int status;
1626
1627         if (!sriov_enabled(adapter))
1628                 return -EPERM;
1629
1630         if (vf >= adapter->num_vfs)
1631                 return -EINVAL;
1632
1633         status = be_cmd_set_logical_link_config(adapter, link_state, vf+1);
1634         if (status) {
1635                 dev_err(&adapter->pdev->dev,
1636                         "Link state change on VF %d failed: %#x\n", vf, status);
1637                 return be_cmd_status(status);
1638         }
1639
1640         adapter->vf_cfg[vf].plink_tracking = link_state;
1641
1642         return 0;
1643 }
1644
1645 static void be_aic_update(struct be_aic_obj *aic, u64 rx_pkts, u64 tx_pkts,
1646                           ulong now)
1647 {
1648         aic->rx_pkts_prev = rx_pkts;
1649         aic->tx_reqs_prev = tx_pkts;
1650         aic->jiffies = now;
1651 }
1652
1653 static void be_eqd_update(struct be_adapter *adapter)
1654 {
1655         struct be_set_eqd set_eqd[MAX_EVT_QS];
1656         int eqd, i, num = 0, start;
1657         struct be_aic_obj *aic;
1658         struct be_eq_obj *eqo;
1659         struct be_rx_obj *rxo;
1660         struct be_tx_obj *txo;
1661         u64 rx_pkts, tx_pkts;
1662         ulong now;
1663         u32 pps, delta;
1664
1665         for_all_evt_queues(adapter, eqo, i) {
1666                 aic = &adapter->aic_obj[eqo->idx];
1667                 if (!aic->enable) {
1668                         if (aic->jiffies)
1669                                 aic->jiffies = 0;
1670                         eqd = aic->et_eqd;
1671                         goto modify_eqd;
1672                 }
1673
1674                 rxo = &adapter->rx_obj[eqo->idx];
1675                 do {
1676                         start = u64_stats_fetch_begin_irq(&rxo->stats.sync);
1677                         rx_pkts = rxo->stats.rx_pkts;
1678                 } while (u64_stats_fetch_retry_irq(&rxo->stats.sync, start));
1679
1680                 txo = &adapter->tx_obj[eqo->idx];
1681                 do {
1682                         start = u64_stats_fetch_begin_irq(&txo->stats.sync);
1683                         tx_pkts = txo->stats.tx_reqs;
1684                 } while (u64_stats_fetch_retry_irq(&txo->stats.sync, start));
1685
1686                 /* Skip, if wrapped around or first calculation */
1687                 now = jiffies;
1688                 if (!aic->jiffies || time_before(now, aic->jiffies) ||
1689                     rx_pkts < aic->rx_pkts_prev ||
1690                     tx_pkts < aic->tx_reqs_prev) {
1691                         be_aic_update(aic, rx_pkts, tx_pkts, now);
1692                         continue;
1693                 }
1694
1695                 delta = jiffies_to_msecs(now - aic->jiffies);
1696                 pps = (((u32)(rx_pkts - aic->rx_pkts_prev) * 1000) / delta) +
1697                         (((u32)(tx_pkts - aic->tx_reqs_prev) * 1000) / delta);
1698                 eqd = (pps / 15000) << 2;
1699
1700                 if (eqd < 8)
1701                         eqd = 0;
1702                 eqd = min_t(u32, eqd, aic->max_eqd);
1703                 eqd = max_t(u32, eqd, aic->min_eqd);
1704
1705                 be_aic_update(aic, rx_pkts, tx_pkts, now);
1706 modify_eqd:
1707                 if (eqd != aic->prev_eqd) {
1708                         set_eqd[num].delay_multiplier = (eqd * 65)/100;
1709                         set_eqd[num].eq_id = eqo->q.id;
1710                         aic->prev_eqd = eqd;
1711                         num++;
1712                 }
1713         }
1714
1715         if (num)
1716                 be_cmd_modify_eqd(adapter, set_eqd, num);
1717 }
1718
1719 static void be_rx_stats_update(struct be_rx_obj *rxo,
1720                                struct be_rx_compl_info *rxcp)
1721 {
1722         struct be_rx_stats *stats = rx_stats(rxo);
1723
1724         u64_stats_update_begin(&stats->sync);
1725         stats->rx_compl++;
1726         stats->rx_bytes += rxcp->pkt_size;
1727         stats->rx_pkts++;
1728         if (rxcp->pkt_type == BE_MULTICAST_PACKET)
1729                 stats->rx_mcast_pkts++;
1730         if (rxcp->err)
1731                 stats->rx_compl_err++;
1732         u64_stats_update_end(&stats->sync);
1733 }
1734
1735 static inline bool csum_passed(struct be_rx_compl_info *rxcp)
1736 {
1737         /* L4 checksum is not reliable for non TCP/UDP packets.
1738          * Also ignore ipcksm for ipv6 pkts
1739          */
1740         return (rxcp->tcpf || rxcp->udpf) && rxcp->l4_csum &&
1741                 (rxcp->ip_csum || rxcp->ipv6) && !rxcp->err;
1742 }
1743
1744 static struct be_rx_page_info *get_rx_page_info(struct be_rx_obj *rxo)
1745 {
1746         struct be_adapter *adapter = rxo->adapter;
1747         struct be_rx_page_info *rx_page_info;
1748         struct be_queue_info *rxq = &rxo->q;
1749         u16 frag_idx = rxq->tail;
1750
1751         rx_page_info = &rxo->page_info_tbl[frag_idx];
1752         BUG_ON(!rx_page_info->page);
1753
1754         if (rx_page_info->last_frag) {
1755                 dma_unmap_page(&adapter->pdev->dev,
1756                                dma_unmap_addr(rx_page_info, bus),
1757                                adapter->big_page_size, DMA_FROM_DEVICE);
1758                 rx_page_info->last_frag = false;
1759         } else {
1760                 dma_sync_single_for_cpu(&adapter->pdev->dev,
1761                                         dma_unmap_addr(rx_page_info, bus),
1762                                         rx_frag_size, DMA_FROM_DEVICE);
1763         }
1764
1765         queue_tail_inc(rxq);
1766         atomic_dec(&rxq->used);
1767         return rx_page_info;
1768 }
1769
1770 /* Throwaway the data in the Rx completion */
1771 static void be_rx_compl_discard(struct be_rx_obj *rxo,
1772                                 struct be_rx_compl_info *rxcp)
1773 {
1774         struct be_rx_page_info *page_info;
1775         u16 i, num_rcvd = rxcp->num_rcvd;
1776
1777         for (i = 0; i < num_rcvd; i++) {
1778                 page_info = get_rx_page_info(rxo);
1779                 put_page(page_info->page);
1780                 memset(page_info, 0, sizeof(*page_info));
1781         }
1782 }
1783
1784 /*
1785  * skb_fill_rx_data forms a complete skb for an ether frame
1786  * indicated by rxcp.
1787  */
1788 static void skb_fill_rx_data(struct be_rx_obj *rxo, struct sk_buff *skb,
1789                              struct be_rx_compl_info *rxcp)
1790 {
1791         struct be_rx_page_info *page_info;
1792         u16 i, j;
1793         u16 hdr_len, curr_frag_len, remaining;
1794         u8 *start;
1795
1796         page_info = get_rx_page_info(rxo);
1797         start = page_address(page_info->page) + page_info->page_offset;
1798         prefetch(start);
1799
1800         /* Copy data in the first descriptor of this completion */
1801         curr_frag_len = min(rxcp->pkt_size, rx_frag_size);
1802
1803         skb->len = curr_frag_len;
1804         if (curr_frag_len <= BE_HDR_LEN) { /* tiny packet */
1805                 memcpy(skb->data, start, curr_frag_len);
1806                 /* Complete packet has now been moved to data */
1807                 put_page(page_info->page);
1808                 skb->data_len = 0;
1809                 skb->tail += curr_frag_len;
1810         } else {
1811                 hdr_len = ETH_HLEN;
1812                 memcpy(skb->data, start, hdr_len);
1813                 skb_shinfo(skb)->nr_frags = 1;
1814                 skb_frag_set_page(skb, 0, page_info->page);
1815                 skb_shinfo(skb)->frags[0].page_offset =
1816                                         page_info->page_offset + hdr_len;
1817                 skb_frag_size_set(&skb_shinfo(skb)->frags[0],
1818                                   curr_frag_len - hdr_len);
1819                 skb->data_len = curr_frag_len - hdr_len;
1820                 skb->truesize += rx_frag_size;
1821                 skb->tail += hdr_len;
1822         }
1823         page_info->page = NULL;
1824
1825         if (rxcp->pkt_size <= rx_frag_size) {
1826                 BUG_ON(rxcp->num_rcvd != 1);
1827                 return;
1828         }
1829
1830         /* More frags present for this completion */
1831         remaining = rxcp->pkt_size - curr_frag_len;
1832         for (i = 1, j = 0; i < rxcp->num_rcvd; i++) {
1833                 page_info = get_rx_page_info(rxo);
1834                 curr_frag_len = min(remaining, rx_frag_size);
1835
1836                 /* Coalesce all frags from the same physical page in one slot */
1837                 if (page_info->page_offset == 0) {
1838                         /* Fresh page */
1839                         j++;
1840                         skb_frag_set_page(skb, j, page_info->page);
1841                         skb_shinfo(skb)->frags[j].page_offset =
1842                                                         page_info->page_offset;
1843                         skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1844                         skb_shinfo(skb)->nr_frags++;
1845                 } else {
1846                         put_page(page_info->page);
1847                 }
1848
1849                 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1850                 skb->len += curr_frag_len;
1851                 skb->data_len += curr_frag_len;
1852                 skb->truesize += rx_frag_size;
1853                 remaining -= curr_frag_len;
1854                 page_info->page = NULL;
1855         }
1856         BUG_ON(j > MAX_SKB_FRAGS);
1857 }
1858
1859 /* Process the RX completion indicated by rxcp when GRO is disabled */
1860 static void be_rx_compl_process(struct be_rx_obj *rxo, struct napi_struct *napi,
1861                                 struct be_rx_compl_info *rxcp)
1862 {
1863         struct be_adapter *adapter = rxo->adapter;
1864         struct net_device *netdev = adapter->netdev;
1865         struct sk_buff *skb;
1866
1867         skb = netdev_alloc_skb_ip_align(netdev, BE_RX_SKB_ALLOC_SIZE);
1868         if (unlikely(!skb)) {
1869                 rx_stats(rxo)->rx_drops_no_skbs++;
1870                 be_rx_compl_discard(rxo, rxcp);
1871                 return;
1872         }
1873
1874         skb_fill_rx_data(rxo, skb, rxcp);
1875
1876         if (likely((netdev->features & NETIF_F_RXCSUM) && csum_passed(rxcp)))
1877                 skb->ip_summed = CHECKSUM_UNNECESSARY;
1878         else
1879                 skb_checksum_none_assert(skb);
1880
1881         skb->protocol = eth_type_trans(skb, netdev);
1882         skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1883         if (netdev->features & NETIF_F_RXHASH)
1884                 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1885
1886         skb->csum_level = rxcp->tunneled;
1887         skb_mark_napi_id(skb, napi);
1888
1889         if (rxcp->vlanf)
1890                 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1891
1892         netif_receive_skb(skb);
1893 }
1894
1895 /* Process the RX completion indicated by rxcp when GRO is enabled */
1896 static void be_rx_compl_process_gro(struct be_rx_obj *rxo,
1897                                     struct napi_struct *napi,
1898                                     struct be_rx_compl_info *rxcp)
1899 {
1900         struct be_adapter *adapter = rxo->adapter;
1901         struct be_rx_page_info *page_info;
1902         struct sk_buff *skb = NULL;
1903         u16 remaining, curr_frag_len;
1904         u16 i, j;
1905
1906         skb = napi_get_frags(napi);
1907         if (!skb) {
1908                 be_rx_compl_discard(rxo, rxcp);
1909                 return;
1910         }
1911
1912         remaining = rxcp->pkt_size;
1913         for (i = 0, j = -1; i < rxcp->num_rcvd; i++) {
1914                 page_info = get_rx_page_info(rxo);
1915
1916                 curr_frag_len = min(remaining, rx_frag_size);
1917
1918                 /* Coalesce all frags from the same physical page in one slot */
1919                 if (i == 0 || page_info->page_offset == 0) {
1920                         /* First frag or Fresh page */
1921                         j++;
1922                         skb_frag_set_page(skb, j, page_info->page);
1923                         skb_shinfo(skb)->frags[j].page_offset =
1924                                                         page_info->page_offset;
1925                         skb_frag_size_set(&skb_shinfo(skb)->frags[j], 0);
1926                 } else {
1927                         put_page(page_info->page);
1928                 }
1929                 skb_frag_size_add(&skb_shinfo(skb)->frags[j], curr_frag_len);
1930                 skb->truesize += rx_frag_size;
1931                 remaining -= curr_frag_len;
1932                 memset(page_info, 0, sizeof(*page_info));
1933         }
1934         BUG_ON(j > MAX_SKB_FRAGS);
1935
1936         skb_shinfo(skb)->nr_frags = j + 1;
1937         skb->len = rxcp->pkt_size;
1938         skb->data_len = rxcp->pkt_size;
1939         skb->ip_summed = CHECKSUM_UNNECESSARY;
1940         skb_record_rx_queue(skb, rxo - &adapter->rx_obj[0]);
1941         if (adapter->netdev->features & NETIF_F_RXHASH)
1942                 skb_set_hash(skb, rxcp->rss_hash, PKT_HASH_TYPE_L3);
1943
1944         skb->csum_level = rxcp->tunneled;
1945         skb_mark_napi_id(skb, napi);
1946
1947         if (rxcp->vlanf)
1948                 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rxcp->vlan_tag);
1949
1950         napi_gro_frags(napi);
1951 }
1952
1953 static void be_parse_rx_compl_v1(struct be_eth_rx_compl *compl,
1954                                  struct be_rx_compl_info *rxcp)
1955 {
1956         rxcp->pkt_size = GET_RX_COMPL_V1_BITS(pktsize, compl);
1957         rxcp->vlanf = GET_RX_COMPL_V1_BITS(vtp, compl);
1958         rxcp->err = GET_RX_COMPL_V1_BITS(err, compl);
1959         rxcp->tcpf = GET_RX_COMPL_V1_BITS(tcpf, compl);
1960         rxcp->udpf = GET_RX_COMPL_V1_BITS(udpf, compl);
1961         rxcp->ip_csum = GET_RX_COMPL_V1_BITS(ipcksm, compl);
1962         rxcp->l4_csum = GET_RX_COMPL_V1_BITS(l4_cksm, compl);
1963         rxcp->ipv6 = GET_RX_COMPL_V1_BITS(ip_version, compl);
1964         rxcp->num_rcvd = GET_RX_COMPL_V1_BITS(numfrags, compl);
1965         rxcp->pkt_type = GET_RX_COMPL_V1_BITS(cast_enc, compl);
1966         rxcp->rss_hash = GET_RX_COMPL_V1_BITS(rsshash, compl);
1967         if (rxcp->vlanf) {
1968                 rxcp->qnq = GET_RX_COMPL_V1_BITS(qnq, compl);
1969                 rxcp->vlan_tag = GET_RX_COMPL_V1_BITS(vlan_tag, compl);
1970         }
1971         rxcp->port = GET_RX_COMPL_V1_BITS(port, compl);
1972         rxcp->tunneled =
1973                 GET_RX_COMPL_V1_BITS(tunneled, compl);
1974 }
1975
1976 static void be_parse_rx_compl_v0(struct be_eth_rx_compl *compl,
1977                                  struct be_rx_compl_info *rxcp)
1978 {
1979         rxcp->pkt_size = GET_RX_COMPL_V0_BITS(pktsize, compl);
1980         rxcp->vlanf = GET_RX_COMPL_V0_BITS(vtp, compl);
1981         rxcp->err = GET_RX_COMPL_V0_BITS(err, compl);
1982         rxcp->tcpf = GET_RX_COMPL_V0_BITS(tcpf, compl);
1983         rxcp->udpf = GET_RX_COMPL_V0_BITS(udpf, compl);
1984         rxcp->ip_csum = GET_RX_COMPL_V0_BITS(ipcksm, compl);
1985         rxcp->l4_csum = GET_RX_COMPL_V0_BITS(l4_cksm, compl);
1986         rxcp->ipv6 = GET_RX_COMPL_V0_BITS(ip_version, compl);
1987         rxcp->num_rcvd = GET_RX_COMPL_V0_BITS(numfrags, compl);
1988         rxcp->pkt_type = GET_RX_COMPL_V0_BITS(cast_enc, compl);
1989         rxcp->rss_hash = GET_RX_COMPL_V0_BITS(rsshash, compl);
1990         if (rxcp->vlanf) {
1991                 rxcp->qnq = GET_RX_COMPL_V0_BITS(qnq, compl);
1992                 rxcp->vlan_tag = GET_RX_COMPL_V0_BITS(vlan_tag, compl);
1993         }
1994         rxcp->port = GET_RX_COMPL_V0_BITS(port, compl);
1995         rxcp->ip_frag = GET_RX_COMPL_V0_BITS(ip_frag, compl);
1996 }
1997
1998 static struct be_rx_compl_info *be_rx_compl_get(struct be_rx_obj *rxo)
1999 {
2000         struct be_eth_rx_compl *compl = queue_tail_node(&rxo->cq);
2001         struct be_rx_compl_info *rxcp = &rxo->rxcp;
2002         struct be_adapter *adapter = rxo->adapter;
2003
2004         /* For checking the valid bit it is Ok to use either definition as the
2005          * valid bit is at the same position in both v0 and v1 Rx compl */
2006         if (compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] == 0)
2007                 return NULL;
2008
2009         rmb();
2010         be_dws_le_to_cpu(compl, sizeof(*compl));
2011
2012         if (adapter->be3_native)
2013                 be_parse_rx_compl_v1(compl, rxcp);
2014         else
2015                 be_parse_rx_compl_v0(compl, rxcp);
2016
2017         if (rxcp->ip_frag)
2018                 rxcp->l4_csum = 0;
2019
2020         if (rxcp->vlanf) {
2021                 /* In QNQ modes, if qnq bit is not set, then the packet was
2022                  * tagged only with the transparent outer vlan-tag and must
2023                  * not be treated as a vlan packet by host
2024                  */
2025                 if (be_is_qnq_mode(adapter) && !rxcp->qnq)
2026                         rxcp->vlanf = 0;
2027
2028                 if (!lancer_chip(adapter))
2029                         rxcp->vlan_tag = swab16(rxcp->vlan_tag);
2030
2031                 if (adapter->pvid == (rxcp->vlan_tag & VLAN_VID_MASK) &&
2032                     !test_bit(rxcp->vlan_tag, adapter->vids))
2033                         rxcp->vlanf = 0;
2034         }
2035
2036         /* As the compl has been parsed, reset it; we wont touch it again */
2037         compl->dw[offsetof(struct amap_eth_rx_compl_v1, valid) / 32] = 0;
2038
2039         queue_tail_inc(&rxo->cq);
2040         return rxcp;
2041 }
2042
2043 static inline struct page *be_alloc_pages(u32 size, gfp_t gfp)
2044 {
2045         u32 order = get_order(size);
2046
2047         if (order > 0)
2048                 gfp |= __GFP_COMP;
2049         return  alloc_pages(gfp, order);
2050 }
2051
2052 /*
2053  * Allocate a page, split it to fragments of size rx_frag_size and post as
2054  * receive buffers to BE
2055  */
2056 static void be_post_rx_frags(struct be_rx_obj *rxo, gfp_t gfp, u32 frags_needed)
2057 {
2058         struct be_adapter *adapter = rxo->adapter;
2059         struct be_rx_page_info *page_info = NULL, *prev_page_info = NULL;
2060         struct be_queue_info *rxq = &rxo->q;
2061         struct page *pagep = NULL;
2062         struct device *dev = &adapter->pdev->dev;
2063         struct be_eth_rx_d *rxd;
2064         u64 page_dmaaddr = 0, frag_dmaaddr;
2065         u32 posted, page_offset = 0, notify = 0;
2066
2067         page_info = &rxo->page_info_tbl[rxq->head];
2068         for (posted = 0; posted < frags_needed && !page_info->page; posted++) {
2069                 if (!pagep) {
2070                         pagep = be_alloc_pages(adapter->big_page_size, gfp);
2071                         if (unlikely(!pagep)) {
2072                                 rx_stats(rxo)->rx_post_fail++;
2073                                 break;
2074                         }
2075                         page_dmaaddr = dma_map_page(dev, pagep, 0,
2076                                                     adapter->big_page_size,
2077                                                     DMA_FROM_DEVICE);
2078                         if (dma_mapping_error(dev, page_dmaaddr)) {
2079                                 put_page(pagep);
2080                                 pagep = NULL;
2081                                 adapter->drv_stats.dma_map_errors++;
2082                                 break;
2083                         }
2084                         page_offset = 0;
2085                 } else {
2086                         get_page(pagep);
2087                         page_offset += rx_frag_size;
2088                 }
2089                 page_info->page_offset = page_offset;
2090                 page_info->page = pagep;
2091
2092                 rxd = queue_head_node(rxq);
2093                 frag_dmaaddr = page_dmaaddr + page_info->page_offset;
2094                 rxd->fragpa_lo = cpu_to_le32(frag_dmaaddr & 0xFFFFFFFF);
2095                 rxd->fragpa_hi = cpu_to_le32(upper_32_bits(frag_dmaaddr));
2096
2097                 /* Any space left in the current big page for another frag? */
2098                 if ((page_offset + rx_frag_size + rx_frag_size) >
2099                                         adapter->big_page_size) {
2100                         pagep = NULL;
2101                         page_info->last_frag = true;
2102                         dma_unmap_addr_set(page_info, bus, page_dmaaddr);
2103                 } else {
2104                         dma_unmap_addr_set(page_info, bus, frag_dmaaddr);
2105                 }
2106
2107                 prev_page_info = page_info;
2108                 queue_head_inc(rxq);
2109                 page_info = &rxo->page_info_tbl[rxq->head];
2110         }
2111
2112         /* Mark the last frag of a page when we break out of the above loop
2113          * with no more slots available in the RXQ
2114          */
2115         if (pagep) {
2116                 prev_page_info->last_frag = true;
2117                 dma_unmap_addr_set(prev_page_info, bus, page_dmaaddr);
2118         }
2119
2120         if (posted) {
2121                 atomic_add(posted, &rxq->used);
2122                 if (rxo->rx_post_starved)
2123                         rxo->rx_post_starved = false;
2124                 do {
2125                         notify = min(256u, posted);
2126                         be_rxq_notify(adapter, rxq->id, notify);
2127                         posted -= notify;
2128                 } while (posted);
2129         } else if (atomic_read(&rxq->used) == 0) {
2130                 /* Let be_worker replenish when memory is available */
2131                 rxo->rx_post_starved = true;
2132         }
2133 }
2134
2135 static struct be_tx_compl_info *be_tx_compl_get(struct be_tx_obj *txo)
2136 {
2137         struct be_queue_info *tx_cq = &txo->cq;
2138         struct be_tx_compl_info *txcp = &txo->txcp;
2139         struct be_eth_tx_compl *compl = queue_tail_node(tx_cq);
2140
2141         if (compl->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] == 0)
2142                 return NULL;
2143
2144         /* Ensure load ordering of valid bit dword and other dwords below */
2145         rmb();
2146         be_dws_le_to_cpu(compl, sizeof(*compl));
2147
2148         txcp->status = GET_TX_COMPL_BITS(status, compl);
2149         txcp->end_index = GET_TX_COMPL_BITS(wrb_index, compl);
2150
2151         compl->dw[offsetof(struct amap_eth_tx_compl, valid) / 32] = 0;
2152         queue_tail_inc(tx_cq);
2153         return txcp;
2154 }
2155
2156 static u16 be_tx_compl_process(struct be_adapter *adapter,
2157                                struct be_tx_obj *txo, u16 last_index)
2158 {
2159         struct sk_buff **sent_skbs = txo->sent_skb_list;
2160         struct be_queue_info *txq = &txo->q;
2161         u16 frag_index, num_wrbs = 0;
2162         struct sk_buff *skb = NULL;
2163         bool unmap_skb_hdr = false;
2164         struct be_eth_wrb *wrb;
2165
2166         do {
2167                 if (sent_skbs[txq->tail]) {
2168                         /* Free skb from prev req */
2169                         if (skb)
2170                                 dev_consume_skb_any(skb);
2171                         skb = sent_skbs[txq->tail];
2172                         sent_skbs[txq->tail] = NULL;
2173                         queue_tail_inc(txq);  /* skip hdr wrb */
2174                         num_wrbs++;
2175                         unmap_skb_hdr = true;
2176                 }
2177                 wrb = queue_tail_node(txq);
2178                 frag_index = txq->tail;
2179                 unmap_tx_frag(&adapter->pdev->dev, wrb,
2180                               (unmap_skb_hdr && skb_headlen(skb)));
2181                 unmap_skb_hdr = false;
2182                 queue_tail_inc(txq);
2183                 num_wrbs++;
2184         } while (frag_index != last_index);
2185         dev_consume_skb_any(skb);
2186
2187         return num_wrbs;
2188 }
2189
2190 /* Return the number of events in the event queue */
2191 static inline int events_get(struct be_eq_obj *eqo)
2192 {
2193         struct be_eq_entry *eqe;
2194         int num = 0;
2195
2196         do {
2197                 eqe = queue_tail_node(&eqo->q);
2198                 if (eqe->evt == 0)
2199                         break;
2200
2201                 rmb();
2202                 eqe->evt = 0;
2203                 num++;
2204                 queue_tail_inc(&eqo->q);
2205         } while (true);
2206
2207         return num;
2208 }
2209
2210 /* Leaves the EQ is disarmed state */
2211 static void be_eq_clean(struct be_eq_obj *eqo)
2212 {
2213         int num = events_get(eqo);
2214
2215         be_eq_notify(eqo->adapter, eqo->q.id, false, true, num);
2216 }
2217
2218 static void be_rx_cq_clean(struct be_rx_obj *rxo)
2219 {
2220         struct be_rx_page_info *page_info;
2221         struct be_queue_info *rxq = &rxo->q;
2222         struct be_queue_info *rx_cq = &rxo->cq;
2223         struct be_rx_compl_info *rxcp;
2224         struct be_adapter *adapter = rxo->adapter;
2225         int flush_wait = 0;
2226
2227         /* Consume pending rx completions.
2228          * Wait for the flush completion (identified by zero num_rcvd)
2229          * to arrive. Notify CQ even when there are no more CQ entries
2230          * for HW to flush partially coalesced CQ entries.
2231          * In Lancer, there is no need to wait for flush compl.
2232          */
2233         for (;;) {
2234                 rxcp = be_rx_compl_get(rxo);
2235                 if (!rxcp) {
2236                         if (lancer_chip(adapter))
2237                                 break;
2238
2239                         if (flush_wait++ > 10 || be_hw_error(adapter)) {
2240                                 dev_warn(&adapter->pdev->dev,
2241                                          "did not receive flush compl\n");
2242                                 break;
2243                         }
2244                         be_cq_notify(adapter, rx_cq->id, true, 0);
2245                         mdelay(1);
2246                 } else {
2247                         be_rx_compl_discard(rxo, rxcp);
2248                         be_cq_notify(adapter, rx_cq->id, false, 1);
2249                         if (rxcp->num_rcvd == 0)
2250                                 break;
2251                 }
2252         }
2253
2254         /* After cleanup, leave the CQ in unarmed state */
2255         be_cq_notify(adapter, rx_cq->id, false, 0);
2256
2257         /* Then free posted rx buffers that were not used */
2258         while (atomic_read(&rxq->used) > 0) {
2259                 page_info = get_rx_page_info(rxo);
2260                 put_page(page_info->page);
2261                 memset(page_info, 0, sizeof(*page_info));
2262         }
2263         BUG_ON(atomic_read(&rxq->used));
2264         rxq->tail = 0;
2265         rxq->head = 0;
2266 }
2267
2268 static void be_tx_compl_clean(struct be_adapter *adapter)
2269 {
2270         u16 end_idx, notified_idx, cmpl = 0, timeo = 0, num_wrbs = 0;
2271         struct device *dev = &adapter->pdev->dev;
2272         struct be_tx_compl_info *txcp;
2273         struct be_queue_info *txq;
2274         struct be_tx_obj *txo;
2275         int i, pending_txqs;
2276
2277         /* Stop polling for compls when HW has been silent for 10ms */
2278         do {
2279                 pending_txqs = adapter->num_tx_qs;
2280
2281                 for_all_tx_queues(adapter, txo, i) {
2282                         cmpl = 0;
2283                         num_wrbs = 0;
2284                         txq = &txo->q;
2285                         while ((txcp = be_tx_compl_get(txo))) {
2286                                 num_wrbs +=
2287                                         be_tx_compl_process(adapter, txo,
2288                                                             txcp->end_index);
2289                                 cmpl++;
2290                         }
2291                         if (cmpl) {
2292                                 be_cq_notify(adapter, txo->cq.id, false, cmpl);
2293                                 atomic_sub(num_wrbs, &txq->used);
2294                                 timeo = 0;
2295                         }
2296                         if (!be_is_tx_compl_pending(txo))
2297                                 pending_txqs--;
2298                 }
2299
2300                 if (pending_txqs == 0 || ++timeo > 10 || be_hw_error(adapter))
2301                         break;
2302
2303                 mdelay(1);
2304         } while (true);
2305
2306         /* Free enqueued TX that was never notified to HW */
2307         for_all_tx_queues(adapter, txo, i) {
2308                 txq = &txo->q;
2309
2310                 if (atomic_read(&txq->used)) {
2311                         dev_info(dev, "txq%d: cleaning %d pending tx-wrbs\n",
2312                                  i, atomic_read(&txq->used));
2313                         notified_idx = txq->tail;
2314                         end_idx = txq->tail;
2315                         index_adv(&end_idx, atomic_read(&txq->used) - 1,
2316                                   txq->len);
2317                         /* Use the tx-compl process logic to handle requests
2318                          * that were not sent to the HW.
2319                          */
2320                         num_wrbs = be_tx_compl_process(adapter, txo, end_idx);
2321                         atomic_sub(num_wrbs, &txq->used);
2322                         BUG_ON(atomic_read(&txq->used));
2323                         txo->pend_wrb_cnt = 0;
2324                         /* Since hw was never notified of these requests,
2325                          * reset TXQ indices
2326                          */
2327                         txq->head = notified_idx;
2328                         txq->tail = notified_idx;
2329                 }
2330         }
2331 }
2332
2333 static void be_evt_queues_destroy(struct be_adapter *adapter)
2334 {
2335         struct be_eq_obj *eqo;
2336         int i;
2337
2338         for_all_evt_queues(adapter, eqo, i) {
2339                 if (eqo->q.created) {
2340                         be_eq_clean(eqo);
2341                         be_cmd_q_destroy(adapter, &eqo->q, QTYPE_EQ);
2342                         napi_hash_del(&eqo->napi);
2343                         netif_napi_del(&eqo->napi);
2344                 }
2345                 be_queue_free(adapter, &eqo->q);
2346         }
2347 }
2348
2349 static int be_evt_queues_create(struct be_adapter *adapter)
2350 {
2351         struct be_queue_info *eq;
2352         struct be_eq_obj *eqo;
2353         struct be_aic_obj *aic;
2354         int i, rc;
2355
2356         adapter->num_evt_qs = min_t(u16, num_irqs(adapter),
2357                                     adapter->cfg_num_qs);
2358
2359         for_all_evt_queues(adapter, eqo, i) {
2360                 netif_napi_add(adapter->netdev, &eqo->napi, be_poll,
2361                                BE_NAPI_WEIGHT);
2362                 napi_hash_add(&eqo->napi);
2363                 aic = &adapter->aic_obj[i];
2364                 eqo->adapter = adapter;
2365                 eqo->idx = i;
2366                 aic->max_eqd = BE_MAX_EQD;
2367                 aic->enable = true;
2368
2369                 eq = &eqo->q;
2370                 rc = be_queue_alloc(adapter, eq, EVNT_Q_LEN,
2371                                     sizeof(struct be_eq_entry));
2372                 if (rc)
2373                         return rc;
2374
2375                 rc = be_cmd_eq_create(adapter, eqo);
2376                 if (rc)
2377                         return rc;
2378         }
2379         return 0;
2380 }
2381
2382 static void be_mcc_queues_destroy(struct be_adapter *adapter)
2383 {
2384         struct be_queue_info *q;
2385
2386         q = &adapter->mcc_obj.q;
2387         if (q->created)
2388                 be_cmd_q_destroy(adapter, q, QTYPE_MCCQ);
2389         be_queue_free(adapter, q);
2390
2391         q = &adapter->mcc_obj.cq;
2392         if (q->created)
2393                 be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2394         be_queue_free(adapter, q);
2395 }
2396
2397 /* Must be called only after TX qs are created as MCC shares TX EQ */
2398 static int be_mcc_queues_create(struct be_adapter *adapter)
2399 {
2400         struct be_queue_info *q, *cq;
2401
2402         cq = &adapter->mcc_obj.cq;
2403         if (be_queue_alloc(adapter, cq, MCC_CQ_LEN,
2404                            sizeof(struct be_mcc_compl)))
2405                 goto err;
2406
2407         /* Use the default EQ for MCC completions */
2408         if (be_cmd_cq_create(adapter, cq, &mcc_eqo(adapter)->q, true, 0))
2409                 goto mcc_cq_free;
2410
2411         q = &adapter->mcc_obj.q;
2412         if (be_queue_alloc(adapter, q, MCC_Q_LEN, sizeof(struct be_mcc_wrb)))
2413                 goto mcc_cq_destroy;
2414
2415         if (be_cmd_mccq_create(adapter, q, cq))
2416                 goto mcc_q_free;
2417
2418         return 0;
2419
2420 mcc_q_free:
2421         be_queue_free(adapter, q);
2422 mcc_cq_destroy:
2423         be_cmd_q_destroy(adapter, cq, QTYPE_CQ);
2424 mcc_cq_free:
2425         be_queue_free(adapter, cq);
2426 err:
2427         return -1;
2428 }
2429
2430 static void be_tx_queues_destroy(struct be_adapter *adapter)
2431 {
2432         struct be_queue_info *q;
2433         struct be_tx_obj *txo;
2434         u8 i;
2435
2436         for_all_tx_queues(adapter, txo, i) {
2437                 q = &txo->q;
2438                 if (q->created)
2439                         be_cmd_q_destroy(adapter, q, QTYPE_TXQ);
2440                 be_queue_free(adapter, q);
2441
2442                 q = &txo->cq;
2443                 if (q->created)
2444                         be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2445                 be_queue_free(adapter, q);
2446         }
2447 }
2448
2449 static int be_tx_qs_create(struct be_adapter *adapter)
2450 {
2451         struct be_queue_info *cq, *eq;
2452         struct be_tx_obj *txo;
2453         int status, i;
2454
2455         adapter->num_tx_qs = min(adapter->num_evt_qs, be_max_txqs(adapter));
2456
2457         for_all_tx_queues(adapter, txo, i) {
2458                 cq = &txo->cq;
2459                 status = be_queue_alloc(adapter, cq, TX_CQ_LEN,
2460                                         sizeof(struct be_eth_tx_compl));
2461                 if (status)
2462                         return status;
2463
2464                 u64_stats_init(&txo->stats.sync);
2465                 u64_stats_init(&txo->stats.sync_compl);
2466
2467                 /* If num_evt_qs is less than num_tx_qs, then more than
2468                  * one txq share an eq
2469                  */
2470                 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2471                 status = be_cmd_cq_create(adapter, cq, eq, false, 3);
2472                 if (status)
2473                         return status;
2474
2475                 status = be_queue_alloc(adapter, &txo->q, TX_Q_LEN,
2476                                         sizeof(struct be_eth_wrb));
2477                 if (status)
2478                         return status;
2479
2480                 status = be_cmd_txq_create(adapter, txo);
2481                 if (status)
2482                         return status;
2483         }
2484
2485         dev_info(&adapter->pdev->dev, "created %d TX queue(s)\n",
2486                  adapter->num_tx_qs);
2487         return 0;
2488 }
2489
2490 static void be_rx_cqs_destroy(struct be_adapter *adapter)
2491 {
2492         struct be_queue_info *q;
2493         struct be_rx_obj *rxo;
2494         int i;
2495
2496         for_all_rx_queues(adapter, rxo, i) {
2497                 q = &rxo->cq;
2498                 if (q->created)
2499                         be_cmd_q_destroy(adapter, q, QTYPE_CQ);
2500                 be_queue_free(adapter, q);
2501         }
2502 }
2503
2504 static int be_rx_cqs_create(struct be_adapter *adapter)
2505 {
2506         struct be_queue_info *eq, *cq;
2507         struct be_rx_obj *rxo;
2508         int rc, i;
2509
2510         /* We can create as many RSS rings as there are EQs. */
2511         adapter->num_rss_qs = adapter->num_evt_qs;
2512
2513         /* We'll use RSS only if atleast 2 RSS rings are supported. */
2514         if (adapter->num_rss_qs <= 1)
2515                 adapter->num_rss_qs = 0;
2516
2517         adapter->num_rx_qs = adapter->num_rss_qs + adapter->need_def_rxq;
2518
2519         /* When the interface is not capable of RSS rings (and there is no
2520          * need to create a default RXQ) we'll still need one RXQ
2521          */
2522         if (adapter->num_rx_qs == 0)
2523                 adapter->num_rx_qs = 1;
2524
2525         adapter->big_page_size = (1 << get_order(rx_frag_size)) * PAGE_SIZE;
2526         for_all_rx_queues(adapter, rxo, i) {
2527                 rxo->adapter = adapter;
2528                 cq = &rxo->cq;
2529                 rc = be_queue_alloc(adapter, cq, RX_CQ_LEN,
2530                                     sizeof(struct be_eth_rx_compl));
2531                 if (rc)
2532                         return rc;
2533
2534                 u64_stats_init(&rxo->stats.sync);
2535                 eq = &adapter->eq_obj[i % adapter->num_evt_qs].q;
2536                 rc = be_cmd_cq_create(adapter, cq, eq, false, 3);
2537                 if (rc)
2538                         return rc;
2539         }
2540
2541         dev_info(&adapter->pdev->dev,
2542                  "created %d RX queue(s)\n", adapter->num_rx_qs);
2543         return 0;
2544 }
2545
2546 static irqreturn_t be_intx(int irq, void *dev)
2547 {
2548         struct be_eq_obj *eqo = dev;
2549         struct be_adapter *adapter = eqo->adapter;
2550         int num_evts = 0;
2551
2552         /* IRQ is not expected when NAPI is scheduled as the EQ
2553          * will not be armed.
2554          * But, this can happen on Lancer INTx where it takes
2555          * a while to de-assert INTx or in BE2 where occasionaly
2556          * an interrupt may be raised even when EQ is unarmed.
2557          * If NAPI is already scheduled, then counting & notifying
2558          * events will orphan them.
2559          */
2560         if (napi_schedule_prep(&eqo->napi)) {
2561                 num_evts = events_get(eqo);
2562                 __napi_schedule(&eqo->napi);
2563                 if (num_evts)
2564                         eqo->spurious_intr = 0;
2565         }
2566         be_eq_notify(adapter, eqo->q.id, false, true, num_evts);
2567
2568         /* Return IRQ_HANDLED only for the the first spurious intr
2569          * after a valid intr to stop the kernel from branding
2570          * this irq as a bad one!
2571          */
2572         if (num_evts || eqo->spurious_intr++ == 0)
2573                 return IRQ_HANDLED;
2574         else
2575                 return IRQ_NONE;
2576 }
2577
2578 static irqreturn_t be_msix(int irq, void *dev)
2579 {
2580         struct be_eq_obj *eqo = dev;
2581
2582         be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
2583         napi_schedule(&eqo->napi);
2584         return IRQ_HANDLED;
2585 }
2586
2587 static inline bool do_gro(struct be_rx_compl_info *rxcp)
2588 {
2589         return (rxcp->tcpf && !rxcp->err && rxcp->l4_csum) ? true : false;
2590 }
2591
2592 static int be_process_rx(struct be_rx_obj *rxo, struct napi_struct *napi,
2593                          int budget, int polling)
2594 {
2595         struct be_adapter *adapter = rxo->adapter;
2596         struct be_queue_info *rx_cq = &rxo->cq;
2597         struct be_rx_compl_info *rxcp;
2598         u32 work_done;
2599         u32 frags_consumed = 0;
2600
2601         for (work_done = 0; work_done < budget; work_done++) {
2602                 rxcp = be_rx_compl_get(rxo);
2603                 if (!rxcp)
2604                         break;
2605
2606                 /* Is it a flush compl that has no data */
2607                 if (unlikely(rxcp->num_rcvd == 0))
2608                         goto loop_continue;
2609
2610                 /* Discard compl with partial DMA Lancer B0 */
2611                 if (unlikely(!rxcp->pkt_size)) {
2612                         be_rx_compl_discard(rxo, rxcp);
2613                         goto loop_continue;
2614                 }
2615
2616                 /* On BE drop pkts that arrive due to imperfect filtering in
2617                  * promiscuous mode on some skews
2618                  */
2619                 if (unlikely(rxcp->port != adapter->port_num &&
2620                              !lancer_chip(adapter))) {
2621                         be_rx_compl_discard(rxo, rxcp);
2622                         goto loop_continue;
2623                 }
2624
2625                 /* Don't do gro when we're busy_polling */
2626                 if (do_gro(rxcp) && polling != BUSY_POLLING)
2627                         be_rx_compl_process_gro(rxo, napi, rxcp);
2628                 else
2629                         be_rx_compl_process(rxo, napi, rxcp);
2630
2631 loop_continue:
2632                 frags_consumed += rxcp->num_rcvd;
2633                 be_rx_stats_update(rxo, rxcp);
2634         }
2635
2636         if (work_done) {
2637                 be_cq_notify(adapter, rx_cq->id, true, work_done);
2638
2639                 /* When an rx-obj gets into post_starved state, just
2640                  * let be_worker do the posting.
2641                  */
2642                 if (atomic_read(&rxo->q.used) < RX_FRAGS_REFILL_WM &&
2643                     !rxo->rx_post_starved)
2644                         be_post_rx_frags(rxo, GFP_ATOMIC,
2645                                          max_t(u32, MAX_RX_POST,
2646                                                frags_consumed));
2647         }
2648
2649         return work_done;
2650 }
2651
2652 static inline void be_update_tx_err(struct be_tx_obj *txo, u8 status)
2653 {
2654         switch (status) {
2655         case BE_TX_COMP_HDR_PARSE_ERR:
2656                 tx_stats(txo)->tx_hdr_parse_err++;
2657                 break;
2658         case BE_TX_COMP_NDMA_ERR:
2659                 tx_stats(txo)->tx_dma_err++;
2660                 break;
2661         case BE_TX_COMP_ACL_ERR:
2662                 tx_stats(txo)->tx_spoof_check_err++;
2663                 break;
2664         }
2665 }
2666
2667 static inline void lancer_update_tx_err(struct be_tx_obj *txo, u8 status)
2668 {
2669         switch (status) {
2670         case LANCER_TX_COMP_LSO_ERR:
2671                 tx_stats(txo)->tx_tso_err++;
2672                 break;
2673         case LANCER_TX_COMP_HSW_DROP_MAC_ERR:
2674         case LANCER_TX_COMP_HSW_DROP_VLAN_ERR:
2675                 tx_stats(txo)->tx_spoof_check_err++;
2676                 break;
2677         case LANCER_TX_COMP_QINQ_ERR:
2678                 tx_stats(txo)->tx_qinq_err++;
2679                 break;
2680         case LANCER_TX_COMP_PARITY_ERR:
2681                 tx_stats(txo)->tx_internal_parity_err++;
2682                 break;
2683         case LANCER_TX_COMP_DMA_ERR:
2684                 tx_stats(txo)->tx_dma_err++;
2685                 break;
2686         }
2687 }
2688
2689 static void be_process_tx(struct be_adapter *adapter, struct be_tx_obj *txo,
2690                           int idx)
2691 {
2692         int num_wrbs = 0, work_done = 0;
2693         struct be_tx_compl_info *txcp;
2694
2695         while ((txcp = be_tx_compl_get(txo))) {
2696                 num_wrbs += be_tx_compl_process(adapter, txo, txcp->end_index);
2697                 work_done++;
2698
2699                 if (txcp->status) {
2700                         if (lancer_chip(adapter))
2701                                 lancer_update_tx_err(txo, txcp->status);
2702                         else
2703                                 be_update_tx_err(txo, txcp->status);
2704                 }
2705         }
2706
2707         if (work_done) {
2708                 be_cq_notify(adapter, txo->cq.id, true, work_done);
2709                 atomic_sub(num_wrbs, &txo->q.used);
2710
2711                 /* As Tx wrbs have been freed up, wake up netdev queue
2712                  * if it was stopped due to lack of tx wrbs.  */
2713                 if (__netif_subqueue_stopped(adapter->netdev, idx) &&
2714                     be_can_txq_wake(txo)) {
2715                         netif_wake_subqueue(adapter->netdev, idx);
2716                 }
2717
2718                 u64_stats_update_begin(&tx_stats(txo)->sync_compl);
2719                 tx_stats(txo)->tx_compl += work_done;
2720                 u64_stats_update_end(&tx_stats(txo)->sync_compl);
2721         }
2722 }
2723
2724 #ifdef CONFIG_NET_RX_BUSY_POLL
2725 static inline bool be_lock_napi(struct be_eq_obj *eqo)
2726 {
2727         bool status = true;
2728
2729         spin_lock(&eqo->lock); /* BH is already disabled */
2730         if (eqo->state & BE_EQ_LOCKED) {
2731                 WARN_ON(eqo->state & BE_EQ_NAPI);
2732                 eqo->state |= BE_EQ_NAPI_YIELD;
2733                 status = false;
2734         } else {
2735                 eqo->state = BE_EQ_NAPI;
2736         }
2737         spin_unlock(&eqo->lock);
2738         return status;
2739 }
2740
2741 static inline void be_unlock_napi(struct be_eq_obj *eqo)
2742 {
2743         spin_lock(&eqo->lock); /* BH is already disabled */
2744
2745         WARN_ON(eqo->state & (BE_EQ_POLL | BE_EQ_NAPI_YIELD));
2746         eqo->state = BE_EQ_IDLE;
2747
2748         spin_unlock(&eqo->lock);
2749 }
2750
2751 static inline bool be_lock_busy_poll(struct be_eq_obj *eqo)
2752 {
2753         bool status = true;
2754
2755         spin_lock_bh(&eqo->lock);
2756         if (eqo->state & BE_EQ_LOCKED) {
2757                 eqo->state |= BE_EQ_POLL_YIELD;
2758                 status = false;
2759         } else {
2760                 eqo->state |= BE_EQ_POLL;
2761         }
2762         spin_unlock_bh(&eqo->lock);
2763         return status;
2764 }
2765
2766 static inline void be_unlock_busy_poll(struct be_eq_obj *eqo)
2767 {
2768         spin_lock_bh(&eqo->lock);
2769
2770         WARN_ON(eqo->state & (BE_EQ_NAPI));
2771         eqo->state = BE_EQ_IDLE;
2772
2773         spin_unlock_bh(&eqo->lock);
2774 }
2775
2776 static inline void be_enable_busy_poll(struct be_eq_obj *eqo)
2777 {
2778         spin_lock_init(&eqo->lock);
2779         eqo->state = BE_EQ_IDLE;
2780 }
2781
2782 static inline void be_disable_busy_poll(struct be_eq_obj *eqo)
2783 {
2784         local_bh_disable();
2785
2786         /* It's enough to just acquire napi lock on the eqo to stop
2787          * be_busy_poll() from processing any queueus.
2788          */
2789         while (!be_lock_napi(eqo))
2790                 mdelay(1);
2791
2792         local_bh_enable();
2793 }
2794
2795 #else /* CONFIG_NET_RX_BUSY_POLL */
2796
2797 static inline bool be_lock_napi(struct be_eq_obj *eqo)
2798 {
2799         return true;
2800 }
2801
2802 static inline void be_unlock_napi(struct be_eq_obj *eqo)
2803 {
2804 }
2805
2806 static inline bool be_lock_busy_poll(struct be_eq_obj *eqo)
2807 {
2808         return false;
2809 }
2810
2811 static inline void be_unlock_busy_poll(struct be_eq_obj *eqo)
2812 {
2813 }
2814
2815 static inline void be_enable_busy_poll(struct be_eq_obj *eqo)
2816 {
2817 }
2818
2819 static inline void be_disable_busy_poll(struct be_eq_obj *eqo)
2820 {
2821 }
2822 #endif /* CONFIG_NET_RX_BUSY_POLL */
2823
2824 int be_poll(struct napi_struct *napi, int budget)
2825 {
2826         struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2827         struct be_adapter *adapter = eqo->adapter;
2828         int max_work = 0, work, i, num_evts;
2829         struct be_rx_obj *rxo;
2830         struct be_tx_obj *txo;
2831
2832         num_evts = events_get(eqo);
2833
2834         for_all_tx_queues_on_eq(adapter, eqo, txo, i)
2835                 be_process_tx(adapter, txo, i);
2836
2837         if (be_lock_napi(eqo)) {
2838                 /* This loop will iterate twice for EQ0 in which
2839                  * completions of the last RXQ (default one) are also processed
2840                  * For other EQs the loop iterates only once
2841                  */
2842                 for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2843                         work = be_process_rx(rxo, napi, budget, NAPI_POLLING);
2844                         max_work = max(work, max_work);
2845                 }
2846                 be_unlock_napi(eqo);
2847         } else {
2848                 max_work = budget;
2849         }
2850
2851         if (is_mcc_eqo(eqo))
2852                 be_process_mcc(adapter);
2853
2854         if (max_work < budget) {
2855                 napi_complete(napi);
2856                 be_eq_notify(adapter, eqo->q.id, true, false, num_evts);
2857         } else {
2858                 /* As we'll continue in polling mode, count and clear events */
2859                 be_eq_notify(adapter, eqo->q.id, false, false, num_evts);
2860         }
2861         return max_work;
2862 }
2863
2864 #ifdef CONFIG_NET_RX_BUSY_POLL
2865 static int be_busy_poll(struct napi_struct *napi)
2866 {
2867         struct be_eq_obj *eqo = container_of(napi, struct be_eq_obj, napi);
2868         struct be_adapter *adapter = eqo->adapter;
2869         struct be_rx_obj *rxo;
2870         int i, work = 0;
2871
2872         if (!be_lock_busy_poll(eqo))
2873                 return LL_FLUSH_BUSY;
2874
2875         for_all_rx_queues_on_eq(adapter, eqo, rxo, i) {
2876                 work = be_process_rx(rxo, napi, 4, BUSY_POLLING);
2877                 if (work)
2878                         break;
2879         }
2880
2881         be_unlock_busy_poll(eqo);
2882         return work;
2883 }
2884 #endif
2885
2886 void be_detect_error(struct be_adapter *adapter)
2887 {
2888         u32 ue_lo = 0, ue_hi = 0, ue_lo_mask = 0, ue_hi_mask = 0;
2889         u32 sliport_status = 0, sliport_err1 = 0, sliport_err2 = 0;
2890         u32 i;
2891         bool error_detected = false;
2892         struct device *dev = &adapter->pdev->dev;
2893         struct net_device *netdev = adapter->netdev;
2894
2895         if (be_hw_error(adapter))
2896                 return;
2897
2898         if (lancer_chip(adapter)) {
2899                 sliport_status = ioread32(adapter->db + SLIPORT_STATUS_OFFSET);
2900                 if (sliport_status & SLIPORT_STATUS_ERR_MASK) {
2901                         sliport_err1 = ioread32(adapter->db +
2902                                                 SLIPORT_ERROR1_OFFSET);
2903                         sliport_err2 = ioread32(adapter->db +
2904                                                 SLIPORT_ERROR2_OFFSET);
2905                         adapter->hw_error = true;
2906                         error_detected = true;
2907                         /* Do not log error messages if its a FW reset */
2908                         if (sliport_err1 == SLIPORT_ERROR_FW_RESET1 &&
2909                             sliport_err2 == SLIPORT_ERROR_FW_RESET2) {
2910                                 dev_info(dev, "Firmware update in progress\n");
2911                         } else {
2912                                 dev_err(dev, "Error detected in the card\n");
2913                                 dev_err(dev, "ERR: sliport status 0x%x\n",
2914                                         sliport_status);
2915                                 dev_err(dev, "ERR: sliport error1 0x%x\n",
2916                                         sliport_err1);
2917                                 dev_err(dev, "ERR: sliport error2 0x%x\n",
2918                                         sliport_err2);
2919                         }
2920                 }
2921         } else {
2922                 ue_lo = ioread32(adapter->pcicfg + PCICFG_UE_STATUS_LOW);
2923                 ue_hi = ioread32(adapter->pcicfg + PCICFG_UE_STATUS_HIGH);
2924                 ue_lo_mask = ioread32(adapter->pcicfg +
2925                                       PCICFG_UE_STATUS_LOW_MASK);
2926                 ue_hi_mask = ioread32(adapter->pcicfg +
2927                                       PCICFG_UE_STATUS_HI_MASK);
2928
2929                 ue_lo = (ue_lo & ~ue_lo_mask);
2930                 ue_hi = (ue_hi & ~ue_hi_mask);
2931
2932                 /* On certain platforms BE hardware can indicate spurious UEs.
2933                  * Allow HW to stop working completely in case of a real UE.
2934                  * Hence not setting the hw_error for UE detection.
2935                  */
2936
2937                 if (ue_lo || ue_hi) {
2938                         error_detected = true;
2939                         dev_err(dev,
2940                                 "Unrecoverable Error detected in the adapter");
2941                         dev_err(dev, "Please reboot server to recover");
2942                         if (skyhawk_chip(adapter))
2943                                 adapter->hw_error = true;
2944                         for (i = 0; ue_lo; ue_lo >>= 1, i++) {
2945                                 if (ue_lo & 1)
2946                                         dev_err(dev, "UE: %s bit set\n",
2947                                                 ue_status_low_desc[i]);
2948                         }
2949                         for (i = 0; ue_hi; ue_hi >>= 1, i++) {
2950                                 if (ue_hi & 1)
2951                                         dev_err(dev, "UE: %s bit set\n",
2952                                                 ue_status_hi_desc[i]);
2953                         }
2954                 }
2955         }
2956         if (error_detected)
2957                 netif_carrier_off(netdev);
2958 }
2959
2960 static void be_msix_disable(struct be_adapter *adapter)
2961 {
2962         if (msix_enabled(adapter)) {
2963                 pci_disable_msix(adapter->pdev);
2964                 adapter->num_msix_vec = 0;
2965                 adapter->num_msix_roce_vec = 0;
2966         }
2967 }
2968
2969 static int be_msix_enable(struct be_adapter *adapter)
2970 {
2971         int i, num_vec;
2972         struct device *dev = &adapter->pdev->dev;
2973
2974         /* If RoCE is supported, program the max number of NIC vectors that
2975          * may be configured via set-channels, along with vectors needed for
2976          * RoCe. Else, just program the number we'll use initially.
2977          */
2978         if (be_roce_supported(adapter))
2979                 num_vec = min_t(int, 2 * be_max_eqs(adapter),
2980                                 2 * num_online_cpus());
2981         else
2982                 num_vec = adapter->cfg_num_qs;
2983
2984         for (i = 0; i < num_vec; i++)
2985                 adapter->msix_entries[i].entry = i;
2986
2987         num_vec = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
2988                                         MIN_MSIX_VECTORS, num_vec);
2989         if (num_vec < 0)
2990                 goto fail;
2991
2992         if (be_roce_supported(adapter) && num_vec > MIN_MSIX_VECTORS) {
2993                 adapter->num_msix_roce_vec = num_vec / 2;
2994                 dev_info(dev, "enabled %d MSI-x vector(s) for RoCE\n",
2995                          adapter->num_msix_roce_vec);
2996         }
2997
2998         adapter->num_msix_vec = num_vec - adapter->num_msix_roce_vec;
2999
3000         dev_info(dev, "enabled %d MSI-x vector(s) for NIC\n",
3001                  adapter->num_msix_vec);
3002         return 0;
3003
3004 fail:
3005         dev_warn(dev, "MSIx enable failed\n");
3006
3007         /* INTx is not supported in VFs, so fail probe if enable_msix fails */
3008         if (!be_physfn(adapter))
3009                 return num_vec;
3010         return 0;
3011 }
3012
3013 static inline int be_msix_vec_get(struct be_adapter *adapter,
3014                                   struct be_eq_obj *eqo)
3015 {
3016         return adapter->msix_entries[eqo->msix_idx].vector;
3017 }
3018
3019 static int be_msix_register(struct be_adapter *adapter)
3020 {
3021         struct net_device *netdev = adapter->netdev;
3022         struct be_eq_obj *eqo;
3023         int status, i, vec;
3024
3025         for_all_evt_queues(adapter, eqo, i) {
3026                 sprintf(eqo->desc, "%s-q%d", netdev->name, i);
3027                 vec = be_msix_vec_get(adapter, eqo);
3028                 status = request_irq(vec, be_msix, 0, eqo->desc, eqo);
3029                 if (status)
3030                         goto err_msix;
3031         }
3032
3033         return 0;
3034 err_msix:
3035         for (i--, eqo = &adapter->eq_obj[i]; i >= 0; i--, eqo--)
3036                 free_irq(be_msix_vec_get(adapter, eqo), eqo);
3037         dev_warn(&adapter->pdev->dev, "MSIX Request IRQ failed - err %d\n",
3038                  status);
3039         be_msix_disable(adapter);
3040         return status;
3041 }
3042
3043 static int be_irq_register(struct be_adapter *adapter)
3044 {
3045         struct net_device *netdev = adapter->netdev;
3046         int status;
3047
3048         if (msix_enabled(adapter)) {
3049                 status = be_msix_register(adapter);
3050                 if (status == 0)
3051                         goto done;
3052                 /* INTx is not supported for VF */
3053                 if (!be_physfn(adapter))
3054                         return status;
3055         }
3056
3057         /* INTx: only the first EQ is used */
3058         netdev->irq = adapter->pdev->irq;
3059         status = request_irq(netdev->irq, be_intx, IRQF_SHARED, netdev->name,
3060                              &adapter->eq_obj[0]);
3061         if (status) {
3062                 dev_err(&adapter->pdev->dev,
3063                         "INTx request IRQ failed - err %d\n", status);
3064                 return status;
3065         }
3066 done:
3067         adapter->isr_registered = true;
3068         return 0;
3069 }
3070
3071 static void be_irq_unregister(struct be_adapter *adapter)
3072 {
3073         struct net_device *netdev = adapter->netdev;
3074         struct be_eq_obj *eqo;
3075         int i;
3076
3077         if (!adapter->isr_registered)
3078                 return;
3079
3080         /* INTx */
3081         if (!msix_enabled(adapter)) {
3082                 free_irq(netdev->irq, &adapter->eq_obj[0]);
3083                 goto done;
3084         }
3085
3086         /* MSIx */
3087         for_all_evt_queues(adapter, eqo, i)
3088                 free_irq(be_msix_vec_get(adapter, eqo), eqo);
3089
3090 done:
3091         adapter->isr_registered = false;
3092 }
3093
3094 static void be_rx_qs_destroy(struct be_adapter *adapter)
3095 {
3096         struct be_queue_info *q;
3097         struct be_rx_obj *rxo;
3098         int i;
3099
3100         for_all_rx_queues(adapter, rxo, i) {
3101                 q = &rxo->q;
3102                 if (q->created) {
3103                         be_cmd_rxq_destroy(adapter, q);
3104                         be_rx_cq_clean(rxo);
3105                 }
3106                 be_queue_free(adapter, q);
3107         }
3108 }
3109
3110 static int be_close(struct net_device *netdev)
3111 {
3112         struct be_adapter *adapter = netdev_priv(netdev);
3113         struct be_eq_obj *eqo;
3114         int i;
3115
3116         /* This protection is needed as be_close() may be called even when the
3117          * adapter is in cleared state (after eeh perm failure)
3118          */
3119         if (!(adapter->flags & BE_FLAGS_SETUP_DONE))
3120                 return 0;
3121
3122         be_roce_dev_close(adapter);
3123
3124         if (adapter->flags & BE_FLAGS_NAPI_ENABLED) {
3125                 for_all_evt_queues(adapter, eqo, i) {
3126                         napi_disable(&eqo->napi);
3127                         be_disable_busy_poll(eqo);
3128                 }
3129                 adapter->flags &= ~BE_FLAGS_NAPI_ENABLED;
3130         }
3131
3132         be_async_mcc_disable(adapter);
3133
3134         /* Wait for all pending tx completions to arrive so that
3135          * all tx skbs are freed.
3136          */
3137         netif_tx_disable(netdev);
3138         be_tx_compl_clean(adapter);
3139
3140         be_rx_qs_destroy(adapter);
3141         be_clear_uc_list(adapter);
3142
3143         for_all_evt_queues(adapter, eqo, i) {
3144                 if (msix_enabled(adapter))
3145                         synchronize_irq(be_msix_vec_get(adapter, eqo));
3146                 else
3147                         synchronize_irq(netdev->irq);
3148                 be_eq_clean(eqo);
3149         }
3150
3151         be_irq_unregister(adapter);
3152
3153         return 0;
3154 }
3155
3156 static int be_rx_qs_create(struct be_adapter *adapter)
3157 {
3158         struct rss_info *rss = &adapter->rss_info;
3159         u8 rss_key[RSS_HASH_KEY_LEN];
3160         struct be_rx_obj *rxo;
3161         int rc, i, j;
3162
3163         for_all_rx_queues(adapter, rxo, i) {
3164                 rc = be_queue_alloc(adapter, &rxo->q, RX_Q_LEN,
3165                                     sizeof(struct be_eth_rx_d));
3166                 if (rc)
3167                         return rc;
3168         }
3169
3170         if (adapter->need_def_rxq || !adapter->num_rss_qs) {
3171                 rxo = default_rxo(adapter);
3172                 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
3173                                        rx_frag_size, adapter->if_handle,
3174                                        false, &rxo->rss_id);
3175                 if (rc)
3176                         return rc;
3177         }
3178
3179         for_all_rss_queues(adapter, rxo, i) {
3180                 rc = be_cmd_rxq_create(adapter, &rxo->q, rxo->cq.id,
3181                                        rx_frag_size, adapter->if_handle,
3182                                        true, &rxo->rss_id);
3183                 if (rc)
3184                         return rc;
3185         }
3186
3187         if (be_multi_rxq(adapter)) {
3188                 for (j = 0; j < RSS_INDIR_TABLE_LEN; j += adapter->num_rss_qs) {
3189                         for_all_rss_queues(adapter, rxo, i) {
3190                                 if ((j + i) >= RSS_INDIR_TABLE_LEN)
3191                                         break;
3192                                 rss->rsstable[j + i] = rxo->rss_id;
3193                                 rss->rss_queue[j + i] = i;
3194                         }
3195                 }
3196                 rss->rss_flags = RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4 |
3197                         RSS_ENABLE_TCP_IPV6 | RSS_ENABLE_IPV6;
3198
3199                 if (!BEx_chip(adapter))
3200                         rss->rss_flags |= RSS_ENABLE_UDP_IPV4 |
3201                                 RSS_ENABLE_UDP_IPV6;
3202         } else {
3203                 /* Disable RSS, if only default RX Q is created */
3204                 rss->rss_flags = RSS_ENABLE_NONE;
3205         }
3206
3207         netdev_rss_key_fill(rss_key, RSS_HASH_KEY_LEN);
3208         rc = be_cmd_rss_config(adapter, rss->rsstable, rss->rss_flags,
3209                                128, rss_key);
3210         if (rc) {
3211                 rss->rss_flags = RSS_ENABLE_NONE;
3212                 return rc;
3213         }
3214
3215         memcpy(rss->rss_hkey, rss_key, RSS_HASH_KEY_LEN);
3216
3217         /* First time posting */
3218         for_all_rx_queues(adapter, rxo, i)
3219                 be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
3220         return 0;
3221 }
3222
3223 static int be_open(struct net_device *netdev)
3224 {
3225         struct be_adapter *adapter = netdev_priv(netdev);
3226         struct be_eq_obj *eqo;
3227         struct be_rx_obj *rxo;
3228         struct be_tx_obj *txo;
3229         u8 link_status;
3230         int status, i;
3231
3232         status = be_rx_qs_create(adapter);
3233         if (status)
3234                 goto err;
3235
3236         status = be_irq_register(adapter);
3237         if (status)
3238                 goto err;
3239
3240         for_all_rx_queues(adapter, rxo, i)
3241                 be_cq_notify(adapter, rxo->cq.id, true, 0);
3242
3243         for_all_tx_queues(adapter, txo, i)
3244                 be_cq_notify(adapter, txo->cq.id, true, 0);
3245
3246         be_async_mcc_enable(adapter);
3247
3248         for_all_evt_queues(adapter, eqo, i) {
3249                 napi_enable(&eqo->napi);
3250                 be_enable_busy_poll(eqo);
3251                 be_eq_notify(adapter, eqo->q.id, true, true, 0);
3252         }
3253         adapter->flags |= BE_FLAGS_NAPI_ENABLED;
3254
3255         status = be_cmd_link_status_query(adapter, NULL, &link_status, 0);
3256         if (!status)
3257                 be_link_status_update(adapter, link_status);
3258
3259         netif_tx_start_all_queues(netdev);
3260         be_roce_dev_open(adapter);
3261
3262 #ifdef CONFIG_BE2NET_VXLAN
3263         if (skyhawk_chip(adapter))
3264                 vxlan_get_rx_port(netdev);
3265 #endif
3266
3267         return 0;
3268 err:
3269         be_close(adapter->netdev);
3270         return -EIO;
3271 }
3272
3273 static int be_setup_wol(struct be_adapter *adapter, bool enable)
3274 {
3275         struct be_dma_mem cmd;
3276         int status = 0;
3277         u8 mac[ETH_ALEN];
3278
3279         eth_zero_addr(mac);
3280
3281         cmd.size = sizeof(struct be_cmd_req_acpi_wol_magic_config);
3282         cmd.va = dma_zalloc_coherent(&adapter->pdev->dev, cmd.size, &cmd.dma,
3283                                      GFP_KERNEL);
3284         if (!cmd.va)
3285                 return -ENOMEM;
3286
3287         if (enable) {
3288                 status = pci_write_config_dword(adapter->pdev,
3289                                                 PCICFG_PM_CONTROL_OFFSET,
3290                                                 PCICFG_PM_CONTROL_MASK);
3291                 if (status) {
3292                         dev_err(&adapter->pdev->dev,
3293                                 "Could not enable Wake-on-lan\n");
3294                         dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va,
3295                                           cmd.dma);
3296                         return status;
3297                 }
3298                 status = be_cmd_enable_magic_wol(adapter,
3299                                                  adapter->netdev->dev_addr,
3300                                                  &cmd);
3301                 pci_enable_wake(adapter->pdev, PCI_D3hot, 1);
3302                 pci_enable_wake(adapter->pdev, PCI_D3cold, 1);
3303         } else {
3304                 status = be_cmd_enable_magic_wol(adapter, mac, &cmd);
3305                 pci_enable_wake(adapter->pdev, PCI_D3hot, 0);
3306                 pci_enable_wake(adapter->pdev, PCI_D3cold, 0);
3307         }
3308
3309         dma_free_coherent(&adapter->pdev->dev, cmd.size, cmd.va, cmd.dma);
3310         return status;
3311 }
3312
3313 static void be_vf_eth_addr_generate(struct be_adapter *adapter, u8 *mac)
3314 {
3315         u32 addr;
3316
3317         addr = jhash(adapter->netdev->dev_addr, ETH_ALEN, 0);
3318
3319         mac[5] = (u8)(addr & 0xFF);
3320         mac[4] = (u8)((addr >> 8) & 0xFF);
3321         mac[3] = (u8)((addr >> 16) & 0xFF);
3322         /* Use the OUI from the current MAC address */
3323         memcpy(mac, adapter->netdev->dev_addr, 3);
3324 }
3325
3326 /*
3327  * Generate a seed MAC address from the PF MAC Address using jhash.
3328  * MAC Address for VFs are assigned incrementally starting from the seed.
3329  * These addresses are programmed in the ASIC by the PF and the VF driver
3330  * queries for the MAC address during its probe.
3331  */
3332 static int be_vf_eth_addr_config(struct be_adapter *adapter)
3333 {
3334         u32 vf;
3335         int status = 0;
3336         u8 mac[ETH_ALEN];
3337         struct be_vf_cfg *vf_cfg;
3338
3339         be_vf_eth_addr_generate(adapter, mac);
3340
3341         for_all_vfs(adapter, vf_cfg, vf) {
3342                 if (BEx_chip(adapter))
3343                         status = be_cmd_pmac_add(adapter, mac,
3344                                                  vf_cfg->if_handle,
3345                                                  &vf_cfg->pmac_id, vf + 1);
3346                 else
3347                         status = be_cmd_set_mac(adapter, mac, vf_cfg->if_handle,
3348                                                 vf + 1);
3349
3350                 if (status)
3351                         dev_err(&adapter->pdev->dev,
3352                                 "Mac address assignment failed for VF %d\n",
3353                                 vf);
3354                 else
3355                         memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3356
3357                 mac[5] += 1;
3358         }
3359         return status;
3360 }
3361
3362 static int be_vfs_mac_query(struct be_adapter *adapter)
3363 {
3364         int status, vf;
3365         u8 mac[ETH_ALEN];
3366         struct be_vf_cfg *vf_cfg;
3367
3368         for_all_vfs(adapter, vf_cfg, vf) {
3369                 status = be_cmd_get_active_mac(adapter, vf_cfg->pmac_id,
3370                                                mac, vf_cfg->if_handle,
3371                                                false, vf+1);
3372                 if (status)
3373                         return status;
3374                 memcpy(vf_cfg->mac_addr, mac, ETH_ALEN);
3375         }
3376         return 0;
3377 }
3378
3379 static void be_vf_clear(struct be_adapter *adapter)
3380 {
3381         struct be_vf_cfg *vf_cfg;
3382         u32 vf;
3383
3384         if (pci_vfs_assigned(adapter->pdev)) {
3385                 dev_warn(&adapter->pdev->dev,
3386                          "VFs are assigned to VMs: not disabling VFs\n");
3387                 goto done;
3388         }
3389
3390         pci_disable_sriov(adapter->pdev);
3391
3392         for_all_vfs(adapter, vf_cfg, vf) {
3393                 if (BEx_chip(adapter))
3394                         be_cmd_pmac_del(adapter, vf_cfg->if_handle,
3395                                         vf_cfg->pmac_id, vf + 1);
3396                 else
3397                         be_cmd_set_mac(adapter, NULL, vf_cfg->if_handle,
3398                                        vf + 1);
3399
3400                 be_cmd_if_destroy(adapter, vf_cfg->if_handle, vf + 1);
3401         }
3402 done:
3403         kfree(adapter->vf_cfg);
3404         adapter->num_vfs = 0;
3405         adapter->flags &= ~BE_FLAGS_SRIOV_ENABLED;
3406 }
3407
3408 static void be_clear_queues(struct be_adapter *adapter)
3409 {
3410         be_mcc_queues_destroy(adapter);
3411         be_rx_cqs_destroy(adapter);
3412         be_tx_queues_destroy(adapter);
3413         be_evt_queues_destroy(adapter);
3414 }
3415
3416 static void be_cancel_worker(struct be_adapter *adapter)
3417 {
3418         if (adapter->flags & BE_FLAGS_WORKER_SCHEDULED) {
3419                 cancel_delayed_work_sync(&adapter->work);
3420                 adapter->flags &= ~BE_FLAGS_WORKER_SCHEDULED;
3421         }
3422 }
3423
3424 static void be_cancel_err_detection(struct be_adapter *adapter)
3425 {
3426         if (adapter->flags & BE_FLAGS_ERR_DETECTION_SCHEDULED) {
3427                 cancel_delayed_work_sync(&adapter->be_err_detection_work);
3428                 adapter->flags &= ~BE_FLAGS_ERR_DETECTION_SCHEDULED;
3429         }
3430 }
3431
3432 static void be_mac_clear(struct be_adapter *adapter)
3433 {
3434         if (adapter->pmac_id) {
3435                 be_cmd_pmac_del(adapter, adapter->if_handle,
3436                                 adapter->pmac_id[0], 0);
3437                 kfree(adapter->pmac_id);
3438                 adapter->pmac_id = NULL;
3439         }
3440 }
3441
3442 #ifdef CONFIG_BE2NET_VXLAN
3443 static void be_disable_vxlan_offloads(struct be_adapter *adapter)
3444 {
3445         struct net_device *netdev = adapter->netdev;
3446
3447         if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS)
3448                 be_cmd_manage_iface(adapter, adapter->if_handle,
3449                                     OP_CONVERT_TUNNEL_TO_NORMAL);
3450
3451         if (adapter->vxlan_port)
3452                 be_cmd_set_vxlan_port(adapter, 0);
3453
3454         adapter->flags &= ~BE_FLAGS_VXLAN_OFFLOADS;
3455         adapter->vxlan_port = 0;
3456
3457         netdev->hw_enc_features = 0;
3458         netdev->hw_features &= ~(NETIF_F_GSO_UDP_TUNNEL);
3459         netdev->features &= ~(NETIF_F_GSO_UDP_TUNNEL);
3460 }
3461 #endif
3462
3463 static u16 be_calculate_vf_qs(struct be_adapter *adapter, u16 num_vfs)
3464 {
3465         struct be_resources res = adapter->pool_res;
3466         u16 num_vf_qs = 1;
3467
3468         /* Distribute the queue resources equally among the PF and it's VFs
3469          * Do not distribute queue resources in multi-channel configuration.
3470          */
3471         if (num_vfs && !be_is_mc(adapter)) {
3472                 /* If number of VFs requested is 8 less than max supported,
3473                  * assign 8 queue pairs to the PF and divide the remaining
3474                  * resources evenly among the VFs
3475                  */
3476                 if (num_vfs < (be_max_vfs(adapter) - 8))
3477                         num_vf_qs = (res.max_rss_qs - 8) / num_vfs;
3478                 else
3479                         num_vf_qs = res.max_rss_qs / num_vfs;
3480
3481                 /* Skyhawk-R chip supports only MAX_RSS_IFACES RSS capable
3482                  * interfaces per port. Provide RSS on VFs, only if number
3483                  * of VFs requested is less than MAX_RSS_IFACES limit.
3484                  */
3485                 if (num_vfs >= MAX_RSS_IFACES)
3486                         num_vf_qs = 1;
3487         }
3488         return num_vf_qs;
3489 }
3490
3491 static int be_clear(struct be_adapter *adapter)
3492 {
3493         struct pci_dev *pdev = adapter->pdev;
3494         u16 num_vf_qs;
3495
3496         be_cancel_worker(adapter);
3497
3498         if (sriov_enabled(adapter))
3499                 be_vf_clear(adapter);
3500
3501         /* Re-configure FW to distribute resources evenly across max-supported
3502          * number of VFs, only when VFs are not already enabled.
3503          */
3504         if (skyhawk_chip(adapter) && be_physfn(adapter) &&
3505             !pci_vfs_assigned(pdev)) {
3506                 num_vf_qs = be_calculate_vf_qs(adapter,
3507                                                pci_sriov_get_totalvfs(pdev));
3508                 be_cmd_set_sriov_config(adapter, adapter->pool_res,
3509                                         pci_sriov_get_totalvfs(pdev),
3510                                         num_vf_qs);
3511         }
3512
3513 #ifdef CONFIG_BE2NET_VXLAN
3514         be_disable_vxlan_offloads(adapter);
3515 #endif
3516         /* delete the primary mac along with the uc-mac list */
3517         be_mac_clear(adapter);
3518
3519         be_cmd_if_destroy(adapter, adapter->if_handle,  0);
3520
3521         be_clear_queues(adapter);
3522
3523         be_msix_disable(adapter);
3524         adapter->flags &= ~BE_FLAGS_SETUP_DONE;
3525         return 0;
3526 }
3527
3528 static int be_if_create(struct be_adapter *adapter, u32 *if_handle,
3529                         u32 cap_flags, u32 vf)
3530 {
3531         u32 en_flags;
3532
3533         en_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3534                    BE_IF_FLAGS_MULTICAST | BE_IF_FLAGS_PASS_L3L4_ERRORS |
3535                    BE_IF_FLAGS_RSS | BE_IF_FLAGS_DEFQ_RSS;
3536
3537         en_flags &= cap_flags;
3538
3539         return be_cmd_if_create(adapter, cap_flags, en_flags, if_handle, vf);
3540 }
3541
3542 static int be_vfs_if_create(struct be_adapter *adapter)
3543 {
3544         struct be_resources res = {0};
3545         struct be_vf_cfg *vf_cfg;
3546         u32 cap_flags, vf;
3547         int status;
3548
3549         /* If a FW profile exists, then cap_flags are updated */
3550         cap_flags = BE_IF_FLAGS_UNTAGGED | BE_IF_FLAGS_BROADCAST |
3551                     BE_IF_FLAGS_MULTICAST;
3552
3553         for_all_vfs(adapter, vf_cfg, vf) {
3554                 if (!BE3_chip(adapter)) {
3555                         status = be_cmd_get_profile_config(adapter, &res,
3556                                                            RESOURCE_LIMITS,
3557                                                            vf + 1);
3558                         if (!status) {
3559                                 cap_flags = res.if_cap_flags;
3560                                 /* Prevent VFs from enabling VLAN promiscuous
3561                                  * mode
3562                                  */
3563                                 cap_flags &= ~BE_IF_FLAGS_VLAN_PROMISCUOUS;
3564                         }
3565                 }
3566
3567                 status = be_if_create(adapter, &vf_cfg->if_handle,
3568                                       cap_flags, vf + 1);
3569                 if (status)
3570                         return status;
3571         }
3572
3573         return 0;
3574 }
3575
3576 static int be_vf_setup_init(struct be_adapter *adapter)
3577 {
3578         struct be_vf_cfg *vf_cfg;
3579         int vf;
3580
3581         adapter->vf_cfg = kcalloc(adapter->num_vfs, sizeof(*vf_cfg),
3582                                   GFP_KERNEL);
3583         if (!adapter->vf_cfg)
3584                 return -ENOMEM;
3585
3586         for_all_vfs(adapter, vf_cfg, vf) {
3587                 vf_cfg->if_handle = -1;
3588                 vf_cfg->pmac_id = -1;
3589         }
3590         return 0;
3591 }
3592
3593 static int be_vf_setup(struct be_adapter *adapter)
3594 {
3595         struct device *dev = &adapter->pdev->dev;
3596         struct be_vf_cfg *vf_cfg;
3597         int status, old_vfs, vf;
3598
3599         old_vfs = pci_num_vf(adapter->pdev);
3600
3601         status = be_vf_setup_init(adapter);
3602         if (status)
3603                 goto err;
3604
3605         if (old_vfs) {
3606                 for_all_vfs(adapter, vf_cfg, vf) {
3607                         status = be_cmd_get_if_id(adapter, vf_cfg, vf);
3608                         if (status)
3609                                 goto err;
3610                 }
3611
3612                 status = be_vfs_mac_query(adapter);
3613                 if (status)
3614                         goto err;
3615         } else {
3616                 status = be_vfs_if_create(adapter);
3617                 if (status)
3618                         goto err;
3619
3620                 status = be_vf_eth_addr_config(adapter);
3621                 if (status)
3622                         goto err;
3623         }
3624
3625         for_all_vfs(adapter, vf_cfg, vf) {
3626                 /* Allow VFs to programs MAC/VLAN filters */
3627                 status = be_cmd_get_fn_privileges(adapter, &vf_cfg->privileges,
3628                                                   vf + 1);
3629                 if (!status && !(vf_cfg->privileges & BE_PRIV_FILTMGMT)) {
3630                         status = be_cmd_set_fn_privileges(adapter,
3631                                                           vf_cfg->privileges |
3632                                                           BE_PRIV_FILTMGMT,
3633                                                           vf + 1);
3634                         if (!status) {
3635                                 vf_cfg->privileges |= BE_PRIV_FILTMGMT;
3636                                 dev_info(dev, "VF%d has FILTMGMT privilege\n",
3637                                          vf);
3638                         }
3639                 }
3640
3641                 /* Allow full available bandwidth */
3642                 if (!old_vfs)
3643                         be_cmd_config_qos(adapter, 0, 0, vf + 1);
3644
3645                 if (!old_vfs) {
3646                         be_cmd_enable_vf(adapter, vf + 1);
3647                         be_cmd_set_logical_link_config(adapter,
3648                                                        IFLA_VF_LINK_STATE_AUTO,
3649                                                        vf+1);
3650                 }
3651         }
3652
3653         if (!old_vfs) {
3654                 status = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
3655                 if (status) {
3656                         dev_err(dev, "SRIOV enable failed\n");
3657                         adapter->num_vfs = 0;
3658                         goto err;
3659                 }
3660         }
3661
3662         adapter->flags |= BE_FLAGS_SRIOV_ENABLED;
3663         return 0;
3664 err:
3665         dev_err(dev, "VF setup failed\n");
3666         be_vf_clear(adapter);
3667         return status;
3668 }
3669
3670 /* Converting function_mode bits on BE3 to SH mc_type enums */
3671
3672 static u8 be_convert_mc_type(u32 function_mode)
3673 {
3674         if (function_mode & VNIC_MODE && function_mode & QNQ_MODE)
3675                 return vNIC1;
3676         else if (function_mode & QNQ_MODE)
3677                 return FLEX10;
3678         else if (function_mode & VNIC_MODE)
3679                 return vNIC2;
3680         else if (function_mode & UMC_ENABLED)
3681                 return UMC;
3682         else
3683                 return MC_NONE;
3684 }
3685
3686 /* On BE2/BE3 FW does not suggest the supported limits */
3687 static void BEx_get_resources(struct be_adapter *adapter,
3688                               struct be_resources *res)
3689 {
3690         bool use_sriov = adapter->num_vfs ? 1 : 0;
3691
3692         if (be_physfn(adapter))
3693                 res->max_uc_mac = BE_UC_PMAC_COUNT;
3694         else
3695                 res->max_uc_mac = BE_VF_UC_PMAC_COUNT;
3696
3697         adapter->mc_type = be_convert_mc_type(adapter->function_mode);
3698
3699         if (be_is_mc(adapter)) {
3700                 /* Assuming that there are 4 channels per port,
3701                  * when multi-channel is enabled
3702                  */
3703                 if (be_is_qnq_mode(adapter))
3704                         res->max_vlans = BE_NUM_VLANS_SUPPORTED/8;
3705                 else
3706                         /* In a non-qnq multichannel mode, the pvid
3707                          * takes up one vlan entry
3708                          */
3709                         res->max_vlans = (BE_NUM_VLANS_SUPPORTED / 4) - 1;
3710         } else {
3711                 res->max_vlans = BE_NUM_VLANS_SUPPORTED;
3712         }
3713
3714         res->max_mcast_mac = BE_MAX_MC;
3715
3716         /* 1) For BE3 1Gb ports, FW does not support multiple TXQs
3717          * 2) Create multiple TX rings on a BE3-R multi-channel interface
3718          *    *only* if it is RSS-capable.
3719          */
3720         if (BE2_chip(adapter) || use_sriov ||  (adapter->port_num > 1) ||
3721             !be_physfn(adapter) || (be_is_mc(adapter) &&
3722             !(adapter->function_caps & BE_FUNCTION_CAPS_RSS))) {
3723                 res->max_tx_qs = 1;
3724         } else if (adapter->function_caps & BE_FUNCTION_CAPS_SUPER_NIC) {
3725                 struct be_resources super_nic_res = {0};
3726
3727                 /* On a SuperNIC profile, the driver needs to use the
3728                  * GET_PROFILE_CONFIG cmd to query the per-function TXQ limits
3729                  */
3730                 be_cmd_get_profile_config(adapter, &super_nic_res,
3731                                           RESOURCE_LIMITS, 0);
3732                 /* Some old versions of BE3 FW don't report max_tx_qs value */
3733                 res->max_tx_qs = super_nic_res.max_tx_qs ? : BE3_MAX_TX_QS;
3734         } else {
3735                 res->max_tx_qs = BE3_MAX_TX_QS;
3736         }
3737
3738         if ((adapter->function_caps & BE_FUNCTION_CAPS_RSS) &&
3739             !use_sriov && be_physfn(adapter))
3740                 res->max_rss_qs = (adapter->be3_native) ?
3741                                            BE3_MAX_RSS_QS : BE2_MAX_RSS_QS;
3742         res->max_rx_qs = res->max_rss_qs + 1;
3743
3744         if (be_physfn(adapter))
3745                 res->max_evt_qs = (be_max_vfs(adapter) > 0) ?
3746                                         BE3_SRIOV_MAX_EVT_QS : BE3_MAX_EVT_QS;
3747         else
3748                 res->max_evt_qs = 1;
3749
3750         res->if_cap_flags = BE_IF_CAP_FLAGS_WANT;
3751         res->if_cap_flags &= ~BE_IF_FLAGS_DEFQ_RSS;
3752         if (!(adapter->function_caps & BE_FUNCTION_CAPS_RSS))
3753                 res->if_cap_flags &= ~BE_IF_FLAGS_RSS;
3754 }
3755
3756 static void be_setup_init(struct be_adapter *adapter)
3757 {
3758         adapter->vlan_prio_bmap = 0xff;
3759         adapter->phy.link_speed = -1;
3760         adapter->if_handle = -1;
3761         adapter->be3_native = false;
3762         adapter->if_flags = 0;
3763         if (be_physfn(adapter))
3764                 adapter->cmd_privileges = MAX_PRIVILEGES;
3765         else
3766                 adapter->cmd_privileges = MIN_PRIVILEGES;
3767 }
3768
3769 static int be_get_sriov_config(struct be_adapter *adapter)
3770 {
3771         struct be_resources res = {0};
3772         int max_vfs, old_vfs;
3773
3774         be_cmd_get_profile_config(adapter, &res, RESOURCE_LIMITS, 0);
3775
3776         /* Some old versions of BE3 FW don't report max_vfs value */
3777         if (BE3_chip(adapter) && !res.max_vfs) {
3778                 max_vfs = pci_sriov_get_totalvfs(adapter->pdev);
3779                 res.max_vfs = max_vfs > 0 ? min(MAX_VFS, max_vfs) : 0;
3780         }
3781
3782         adapter->pool_res = res;
3783
3784         /* If during previous unload of the driver, the VFs were not disabled,
3785          * then we cannot rely on the PF POOL limits for the TotalVFs value.
3786          * Instead use the TotalVFs value stored in the pci-dev struct.
3787          */
3788         old_vfs = pci_num_vf(adapter->pdev);
3789         if (old_vfs) {
3790                 dev_info(&adapter->pdev->dev, "%d VFs are already enabled\n",
3791                          old_vfs);
3792
3793                 adapter->pool_res.max_vfs =
3794                         pci_sriov_get_totalvfs(adapter->pdev);
3795                 adapter->num_vfs = old_vfs;
3796         }
3797
3798         return 0;
3799 }
3800
3801 static void be_alloc_sriov_res(struct be_adapter *adapter)
3802 {
3803         int old_vfs = pci_num_vf(adapter->pdev);
3804         u16 num_vf_qs;
3805         int status;
3806
3807         be_get_sriov_config(adapter);
3808
3809         if (!old_vfs)
3810                 pci_sriov_set_totalvfs(adapter->pdev, be_max_vfs(adapter));
3811
3812         /* When the HW is in SRIOV capable configuration, the PF-pool
3813          * resources are given to PF during driver load, if there are no
3814          * old VFs. This facility is not available in BE3 FW.
3815          * Also, this is done by FW in Lancer chip.
3816          */
3817         if (skyhawk_chip(adapter) && be_max_vfs(adapter) && !old_vfs) {
3818                 num_vf_qs = be_calculate_vf_qs(adapter, 0);
3819                 status = be_cmd_set_sriov_config(adapter, adapter->pool_res, 0,
3820                                                  num_vf_qs);
3821                 if (status)
3822                         dev_err(&adapter->pdev->dev,
3823                                 "Failed to optimize SRIOV resources\n");
3824         }
3825 }
3826
3827 static int be_get_resources(struct be_adapter *adapter)
3828 {
3829         struct device *dev = &adapter->pdev->dev;
3830         struct be_resources res = {0};
3831         int status;
3832
3833         if (BEx_chip(adapter)) {
3834                 BEx_get_resources(adapter, &res);
3835                 adapter->res = res;
3836         }
3837
3838         /* For Lancer, SH etc read per-function resource limits from FW.
3839          * GET_FUNC_CONFIG returns per function guaranteed limits.
3840          * GET_PROFILE_CONFIG returns PCI-E related limits PF-pool limits
3841          */
3842         if (!BEx_chip(adapter)) {
3843                 status = be_cmd_get_func_config(adapter, &res);
3844                 if (status)
3845                         return status;
3846
3847                 /* If a deafault RXQ must be created, we'll use up one RSSQ*/
3848                 if (res.max_rss_qs && res.max_rss_qs == res.max_rx_qs &&
3849                     !(res.if_cap_flags & BE_IF_FLAGS_DEFQ_RSS))
3850                         res.max_rss_qs -= 1;
3851
3852                 /* If RoCE may be enabled stash away half the EQs for RoCE */
3853                 if (be_roce_supported(adapter))
3854                         res.max_evt_qs /= 2;
3855                 adapter->res = res;
3856         }
3857
3858         /* If FW supports RSS default queue, then skip creating non-RSS
3859          * queue for non-IP traffic.
3860          */
3861         adapter->need_def_rxq = (be_if_cap_flags(adapter) &
3862                                  BE_IF_FLAGS_DEFQ_RSS) ? 0 : 1;
3863
3864         dev_info(dev, "Max: txqs %d, rxqs %d, rss %d, eqs %d, vfs %d\n",
3865                  be_max_txqs(adapter), be_max_rxqs(adapter),
3866                  be_max_rss(adapter), be_max_eqs(adapter),
3867                  be_max_vfs(adapter));
3868         dev_info(dev, "Max: uc-macs %d, mc-macs %d, vlans %d\n",
3869                  be_max_uc(adapter), be_max_mc(adapter),
3870                  be_max_vlans(adapter));
3871
3872         /* Sanitize cfg_num_qs based on HW and platform limits */
3873         adapter->cfg_num_qs = min_t(u16, netif_get_num_default_rss_queues(),
3874                                     be_max_qs(adapter));
3875         return 0;
3876 }
3877
3878 static int be_get_config(struct be_adapter *adapter)
3879 {
3880         int status, level;
3881         u16 profile_id;
3882
3883         status = be_cmd_get_cntl_attributes(adapter);
3884         if (status)
3885                 return status;
3886
3887         status = be_cmd_query_fw_cfg(adapter);
3888         if (status)
3889                 return status;
3890
3891         if (BEx_chip(adapter)) {
3892                 level = be_cmd_get_fw_log_level(adapter);
3893                 adapter->msg_enable =
3894                         level <= FW_LOG_LEVEL_DEFAULT ? NETIF_MSG_HW : 0;
3895         }
3896
3897         be_cmd_get_acpi_wol_cap(adapter);
3898
3899         be_cmd_query_port_name(adapter);
3900
3901         if (be_physfn(adapter)) {
3902                 status = be_cmd_get_active_profile(adapter, &profile_id);
3903                 if (!status)
3904                         dev_info(&adapter->pdev->dev,
3905                                  "Using profile 0x%x\n", profile_id);
3906         }
3907
3908         status = be_get_resources(adapter);
3909         if (status)
3910                 return status;
3911
3912         adapter->pmac_id = kcalloc(be_max_uc(adapter),
3913                                    sizeof(*adapter->pmac_id), GFP_KERNEL);
3914         if (!adapter->pmac_id)
3915                 return -ENOMEM;
3916
3917         return 0;
3918 }
3919
3920 static int be_mac_setup(struct be_adapter *adapter)
3921 {
3922         u8 mac[ETH_ALEN];
3923         int status;
3924
3925         if (is_zero_ether_addr(adapter->netdev->dev_addr)) {
3926                 status = be_cmd_get_perm_mac(adapter, mac);
3927                 if (status)
3928                         return status;
3929
3930                 memcpy(adapter->netdev->dev_addr, mac, ETH_ALEN);
3931                 memcpy(adapter->netdev->perm_addr, mac, ETH_ALEN);
3932         } else {
3933                 /* Maybe the HW was reset; dev_addr must be re-programmed */
3934                 memcpy(mac, adapter->netdev->dev_addr, ETH_ALEN);
3935         }
3936
3937         /* For BE3-R VFs, the PF programs the initial MAC address */
3938         if (!(BEx_chip(adapter) && be_virtfn(adapter)))
3939                 be_cmd_pmac_add(adapter, mac, adapter->if_handle,
3940                                 &adapter->pmac_id[0], 0);
3941         return 0;
3942 }
3943
3944 static void be_schedule_worker(struct be_adapter *adapter)
3945 {
3946         schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
3947         adapter->flags |= BE_FLAGS_WORKER_SCHEDULED;
3948 }
3949
3950 static void be_schedule_err_detection(struct be_adapter *adapter)
3951 {
3952         schedule_delayed_work(&adapter->be_err_detection_work,
3953                               msecs_to_jiffies(1000));
3954         adapter->flags |= BE_FLAGS_ERR_DETECTION_SCHEDULED;
3955 }
3956
3957 static int be_setup_queues(struct be_adapter *adapter)
3958 {
3959         struct net_device *netdev = adapter->netdev;
3960         int status;
3961
3962         status = be_evt_queues_create(adapter);
3963         if (status)
3964                 goto err;
3965
3966         status = be_tx_qs_create(adapter);
3967         if (status)
3968                 goto err;
3969
3970         status = be_rx_cqs_create(adapter);
3971         if (status)
3972                 goto err;
3973
3974         status = be_mcc_queues_create(adapter);
3975         if (status)
3976                 goto err;
3977
3978         status = netif_set_real_num_rx_queues(netdev, adapter->num_rx_qs);
3979         if (status)
3980                 goto err;
3981
3982         status = netif_set_real_num_tx_queues(netdev, adapter->num_tx_qs);
3983         if (status)
3984                 goto err;
3985
3986         return 0;
3987 err:
3988         dev_err(&adapter->pdev->dev, "queue_setup failed\n");
3989         return status;
3990 }
3991
3992 int be_update_queues(struct be_adapter *adapter)
3993 {
3994         struct net_device *netdev = adapter->netdev;
3995         int status;
3996
3997         if (netif_running(netdev))
3998                 be_close(netdev);
3999
4000         be_cancel_worker(adapter);
4001
4002         /* If any vectors have been shared with RoCE we cannot re-program
4003          * the MSIx table.
4004          */
4005         if (!adapter->num_msix_roce_vec)
4006                 be_msix_disable(adapter);
4007
4008         be_clear_queues(adapter);
4009
4010         if (!msix_enabled(adapter)) {
4011                 status = be_msix_enable(adapter);
4012                 if (status)
4013                         return status;
4014         }
4015
4016         status = be_setup_queues(adapter);
4017         if (status)
4018                 return status;
4019
4020         be_schedule_worker(adapter);
4021
4022         if (netif_running(netdev))
4023                 status = be_open(netdev);
4024
4025         return status;
4026 }
4027
4028 static inline int fw_major_num(const char *fw_ver)
4029 {
4030         int fw_major = 0, i;
4031
4032         i = sscanf(fw_ver, "%d.", &fw_major);
4033         if (i != 1)
4034                 return 0;
4035
4036         return fw_major;
4037 }
4038
4039 /* If any VFs are already enabled don't FLR the PF */
4040 static bool be_reset_required(struct be_adapter *adapter)
4041 {
4042         return pci_num_vf(adapter->pdev) ? false : true;
4043 }
4044
4045 /* Wait for the FW to be ready and perform the required initialization */
4046 static int be_func_init(struct be_adapter *adapter)
4047 {
4048         int status;
4049
4050         status = be_fw_wait_ready(adapter);
4051         if (status)
4052                 return status;
4053
4054         if (be_reset_required(adapter)) {
4055                 status = be_cmd_reset_function(adapter);
4056                 if (status)
4057                         return status;
4058
4059                 /* Wait for interrupts to quiesce after an FLR */
4060                 msleep(100);
4061
4062                 /* We can clear all errors when function reset succeeds */
4063                 be_clear_all_error(adapter);
4064         }
4065
4066         /* Tell FW we're ready to fire cmds */
4067         status = be_cmd_fw_init(adapter);
4068         if (status)
4069                 return status;
4070
4071         /* Allow interrupts for other ULPs running on NIC function */
4072         be_intr_set(adapter, true);
4073
4074         return 0;
4075 }
4076
4077 static int be_setup(struct be_adapter *adapter)
4078 {
4079         struct device *dev = &adapter->pdev->dev;
4080         int status;
4081
4082         status = be_func_init(adapter);
4083         if (status)
4084                 return status;
4085
4086         be_setup_init(adapter);
4087
4088         if (!lancer_chip(adapter))
4089                 be_cmd_req_native_mode(adapter);
4090
4091         if (!BE2_chip(adapter) && be_physfn(adapter))
4092                 be_alloc_sriov_res(adapter);
4093
4094         status = be_get_config(adapter);
4095         if (status)
4096                 goto err;
4097
4098         status = be_msix_enable(adapter);
4099         if (status)
4100                 goto err;
4101
4102         status = be_if_create(adapter, &adapter->if_handle,
4103                               be_if_cap_flags(adapter), 0);
4104         if (status)
4105                 goto err;
4106
4107         /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
4108         rtnl_lock();
4109         status = be_setup_queues(adapter);
4110         rtnl_unlock();
4111         if (status)
4112                 goto err;
4113
4114         be_cmd_get_fn_privileges(adapter, &adapter->cmd_privileges, 0);
4115
4116         status = be_mac_setup(adapter);
4117         if (status)
4118                 goto err;
4119
4120         be_cmd_get_fw_ver(adapter);
4121         dev_info(dev, "FW version is %s\n", adapter->fw_ver);
4122
4123         if (BE2_chip(adapter) && fw_major_num(adapter->fw_ver) < 4) {
4124                 dev_err(dev, "Firmware on card is old(%s), IRQs may not work",
4125                         adapter->fw_ver);
4126                 dev_err(dev, "Please upgrade firmware to version >= 4.0\n");
4127         }
4128
4129         if (adapter->vlans_added)
4130                 be_vid_config(adapter);
4131
4132         be_set_rx_mode(adapter->netdev);
4133
4134         status = be_cmd_set_flow_control(adapter, adapter->tx_fc,
4135                                          adapter->rx_fc);
4136         if (status)
4137                 be_cmd_get_flow_control(adapter, &adapter->tx_fc,
4138                                         &adapter->rx_fc);
4139
4140         dev_info(&adapter->pdev->dev, "HW Flow control - TX:%d RX:%d\n",
4141                  adapter->tx_fc, adapter->rx_fc);
4142
4143         if (be_physfn(adapter))
4144                 be_cmd_set_logical_link_config(adapter,
4145                                                IFLA_VF_LINK_STATE_AUTO, 0);
4146
4147         if (adapter->num_vfs)
4148                 be_vf_setup(adapter);
4149
4150         status = be_cmd_get_phy_info(adapter);
4151         if (!status && be_pause_supported(adapter))
4152                 adapter->phy.fc_autoneg = 1;
4153
4154         be_schedule_worker(adapter);
4155         adapter->flags |= BE_FLAGS_SETUP_DONE;
4156         return 0;
4157 err:
4158         be_clear(adapter);
4159         return status;
4160 }
4161
4162 #ifdef CONFIG_NET_POLL_CONTROLLER
4163 static void be_netpoll(struct net_device *netdev)
4164 {
4165         struct be_adapter *adapter = netdev_priv(netdev);
4166         struct be_eq_obj *eqo;
4167         int i;
4168
4169         for_all_evt_queues(adapter, eqo, i) {
4170                 be_eq_notify(eqo->adapter, eqo->q.id, false, true, 0);
4171                 napi_schedule(&eqo->napi);
4172         }
4173 }
4174 #endif
4175
4176 static char flash_cookie[2][16] = {"*** SE FLAS", "H DIRECTORY *** "};
4177
4178 static bool phy_flashing_required(struct be_adapter *adapter)
4179 {
4180         return (adapter->phy.phy_type == PHY_TYPE_TN_8022 &&
4181                 adapter->phy.interface_type == PHY_TYPE_BASET_10GB);
4182 }
4183
4184 static bool is_comp_in_ufi(struct be_adapter *adapter,
4185                            struct flash_section_info *fsec, int type)
4186 {
4187         int i = 0, img_type = 0;
4188         struct flash_section_info_g2 *fsec_g2 = NULL;
4189
4190         if (BE2_chip(adapter))
4191                 fsec_g2 = (struct flash_section_info_g2 *)fsec;
4192
4193         for (i = 0; i < MAX_FLASH_COMP; i++) {
4194                 if (fsec_g2)
4195                         img_type = le32_to_cpu(fsec_g2->fsec_entry[i].type);
4196                 else
4197                         img_type = le32_to_cpu(fsec->fsec_entry[i].type);
4198
4199                 if (img_type == type)
4200                         return true;
4201         }
4202         return false;
4203
4204 }
4205
4206 static struct flash_section_info *get_fsec_info(struct be_adapter *adapter,
4207                                                 int header_size,
4208                                                 const struct firmware *fw)
4209 {
4210         struct flash_section_info *fsec = NULL;
4211         const u8 *p = fw->data;
4212
4213         p += header_size;
4214         while (p < (fw->data + fw->size)) {
4215                 fsec = (struct flash_section_info *)p;
4216                 if (!memcmp(flash_cookie, fsec->cookie, sizeof(flash_cookie)))
4217                         return fsec;
4218                 p += 32;
4219         }
4220         return NULL;
4221 }
4222
4223 static int be_check_flash_crc(struct be_adapter *adapter, const u8 *p,
4224                               u32 img_offset, u32 img_size, int hdr_size,
4225                               u16 img_optype, bool *crc_match)
4226 {
4227         u32 crc_offset;
4228         int status;
4229         u8 crc[4];
4230
4231         status = be_cmd_get_flash_crc(adapter, crc, img_optype, img_offset,
4232                                       img_size - 4);
4233         if (status)
4234                 return status;
4235
4236         crc_offset = hdr_size + img_offset + img_size - 4;
4237
4238         /* Skip flashing, if crc of flashed region matches */
4239         if (!memcmp(crc, p + crc_offset, 4))
4240                 *crc_match = true;
4241         else
4242                 *crc_match = false;
4243
4244         return status;
4245 }
4246
4247 static int be_flash(struct be_adapter *adapter, const u8 *img,
4248                     struct be_dma_mem *flash_cmd, int optype, int img_size,
4249                     u32 img_offset)
4250 {
4251         u32 flash_op, num_bytes, total_bytes = img_size, bytes_sent = 0;
4252         struct be_cmd_write_flashrom *req = flash_cmd->va;
4253         int status;
4254
4255         while (total_bytes) {
4256                 num_bytes = min_t(u32, 32*1024, total_bytes);
4257
4258                 total_bytes -= num_bytes;
4259
4260                 if (!total_bytes) {
4261                         if (optype == OPTYPE_PHY_FW)
4262                                 flash_op = FLASHROM_OPER_PHY_FLASH;
4263                         else
4264                                 flash_op = FLASHROM_OPER_FLASH;
4265                 } else {
4266                         if (optype == OPTYPE_PHY_FW)
4267                                 flash_op = FLASHROM_OPER_PHY_SAVE;
4268                         else
4269                                 flash_op = FLASHROM_OPER_SAVE;
4270                 }
4271
4272                 memcpy(req->data_buf, img, num_bytes);
4273                 img += num_bytes;
4274                 status = be_cmd_write_flashrom(adapter, flash_cmd, optype,
4275                                                flash_op, img_offset +
4276                                                bytes_sent, num_bytes);
4277                 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST &&
4278                     optype == OPTYPE_PHY_FW)
4279                         break;
4280                 else if (status)
4281                         return status;
4282
4283                 bytes_sent += num_bytes;
4284         }
4285         return 0;
4286 }
4287
4288 /* For BE2, BE3 and BE3-R */
4289 static int be_flash_BEx(struct be_adapter *adapter,
4290                         const struct firmware *fw,
4291                         struct be_dma_mem *flash_cmd, int num_of_images)
4292 {
4293         int img_hdrs_size = (num_of_images * sizeof(struct image_hdr));
4294         struct device *dev = &adapter->pdev->dev;
4295         struct flash_section_info *fsec = NULL;
4296         int status, i, filehdr_size, num_comp;
4297         const struct flash_comp *pflashcomp;
4298         bool crc_match;
4299         const u8 *p;
4300
4301         struct flash_comp gen3_flash_types[] = {
4302                 { FLASH_iSCSI_PRIMARY_IMAGE_START_g3, OPTYPE_ISCSI_ACTIVE,
4303                         FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_iSCSI},
4304                 { FLASH_REDBOOT_START_g3, OPTYPE_REDBOOT,
4305                         FLASH_REDBOOT_IMAGE_MAX_SIZE_g3, IMAGE_BOOT_CODE},
4306                 { FLASH_iSCSI_BIOS_START_g3, OPTYPE_BIOS,
4307                         FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_ISCSI},
4308                 { FLASH_PXE_BIOS_START_g3, OPTYPE_PXE_BIOS,
4309                         FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_PXE},
4310                 { FLASH_FCoE_BIOS_START_g3, OPTYPE_FCOE_BIOS,
4311                         FLASH_BIOS_IMAGE_MAX_SIZE_g3, IMAGE_OPTION_ROM_FCoE},
4312                 { FLASH_iSCSI_BACKUP_IMAGE_START_g3, OPTYPE_ISCSI_BACKUP,
4313                         FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_iSCSI},
4314                 { FLASH_FCoE_PRIMARY_IMAGE_START_g3, OPTYPE_FCOE_FW_ACTIVE,
4315                         FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_FCoE},
4316                 { FLASH_FCoE_BACKUP_IMAGE_START_g3, OPTYPE_FCOE_FW_BACKUP,
4317                         FLASH_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_BACKUP_FCoE},
4318                 { FLASH_NCSI_START_g3, OPTYPE_NCSI_FW,
4319                         FLASH_NCSI_IMAGE_MAX_SIZE_g3, IMAGE_NCSI},
4320                 { FLASH_PHY_FW_START_g3, OPTYPE_PHY_FW,
4321                         FLASH_PHY_FW_IMAGE_MAX_SIZE_g3, IMAGE_FIRMWARE_PHY}
4322         };
4323
4324         struct flash_comp gen2_flash_types[] = {
4325                 { FLASH_iSCSI_PRIMARY_IMAGE_START_g2, OPTYPE_ISCSI_ACTIVE,
4326                         FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_iSCSI},
4327                 { FLASH_REDBOOT_START_g2, OPTYPE_REDBOOT,
4328                         FLASH_REDBOOT_IMAGE_MAX_SIZE_g2, IMAGE_BOOT_CODE},
4329                 { FLASH_iSCSI_BIOS_START_g2, OPTYPE_BIOS,
4330                         FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_ISCSI},
4331                 { FLASH_PXE_BIOS_START_g2, OPTYPE_PXE_BIOS,
4332                         FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_PXE},
4333                 { FLASH_FCoE_BIOS_START_g2, OPTYPE_FCOE_BIOS,
4334                         FLASH_BIOS_IMAGE_MAX_SIZE_g2, IMAGE_OPTION_ROM_FCoE},
4335                 { FLASH_iSCSI_BACKUP_IMAGE_START_g2, OPTYPE_ISCSI_BACKUP,
4336                         FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_iSCSI},
4337                 { FLASH_FCoE_PRIMARY_IMAGE_START_g2, OPTYPE_FCOE_FW_ACTIVE,
4338                         FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_FCoE},
4339                 { FLASH_FCoE_BACKUP_IMAGE_START_g2, OPTYPE_FCOE_FW_BACKUP,
4340                          FLASH_IMAGE_MAX_SIZE_g2, IMAGE_FIRMWARE_BACKUP_FCoE}
4341         };
4342
4343         if (BE3_chip(adapter)) {
4344                 pflashcomp = gen3_flash_types;
4345                 filehdr_size = sizeof(struct flash_file_hdr_g3);
4346                 num_comp = ARRAY_SIZE(gen3_flash_types);
4347         } else {
4348                 pflashcomp = gen2_flash_types;
4349                 filehdr_size = sizeof(struct flash_file_hdr_g2);
4350                 num_comp = ARRAY_SIZE(gen2_flash_types);
4351                 img_hdrs_size = 0;
4352         }
4353
4354         /* Get flash section info*/
4355         fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4356         if (!fsec) {
4357                 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4358                 return -1;
4359         }
4360         for (i = 0; i < num_comp; i++) {
4361                 if (!is_comp_in_ufi(adapter, fsec, pflashcomp[i].img_type))
4362                         continue;
4363
4364                 if ((pflashcomp[i].optype == OPTYPE_NCSI_FW) &&
4365                     memcmp(adapter->fw_ver, "3.102.148.0", 11) < 0)
4366                         continue;
4367
4368                 if (pflashcomp[i].optype == OPTYPE_PHY_FW  &&
4369                     !phy_flashing_required(adapter))
4370                                 continue;
4371
4372                 if (pflashcomp[i].optype == OPTYPE_REDBOOT) {
4373                         status = be_check_flash_crc(adapter, fw->data,
4374                                                     pflashcomp[i].offset,
4375                                                     pflashcomp[i].size,
4376                                                     filehdr_size +
4377                                                     img_hdrs_size,
4378                                                     OPTYPE_REDBOOT, &crc_match);
4379                         if (status) {
4380                                 dev_err(dev,
4381                                         "Could not get CRC for 0x%x region\n",
4382                                         pflashcomp[i].optype);
4383                                 continue;
4384                         }
4385
4386                         if (crc_match)
4387                                 continue;
4388                 }
4389
4390                 p = fw->data + filehdr_size + pflashcomp[i].offset +
4391                         img_hdrs_size;
4392                 if (p + pflashcomp[i].size > fw->data + fw->size)
4393                         return -1;
4394
4395                 status = be_flash(adapter, p, flash_cmd, pflashcomp[i].optype,
4396                                   pflashcomp[i].size, 0);
4397                 if (status) {
4398                         dev_err(dev, "Flashing section type 0x%x failed\n",
4399                                 pflashcomp[i].img_type);
4400                         return status;
4401                 }
4402         }
4403         return 0;
4404 }
4405
4406 static u16 be_get_img_optype(struct flash_section_entry fsec_entry)
4407 {
4408         u32 img_type = le32_to_cpu(fsec_entry.type);
4409         u16 img_optype = le16_to_cpu(fsec_entry.optype);
4410
4411         if (img_optype != 0xFFFF)
4412                 return img_optype;
4413
4414         switch (img_type) {
4415         case IMAGE_FIRMWARE_iSCSI:
4416                 img_optype = OPTYPE_ISCSI_ACTIVE;
4417                 break;
4418         case IMAGE_BOOT_CODE:
4419                 img_optype = OPTYPE_REDBOOT;
4420                 break;
4421         case IMAGE_OPTION_ROM_ISCSI:
4422                 img_optype = OPTYPE_BIOS;
4423                 break;
4424         case IMAGE_OPTION_ROM_PXE:
4425                 img_optype = OPTYPE_PXE_BIOS;
4426                 break;
4427         case IMAGE_OPTION_ROM_FCoE:
4428                 img_optype = OPTYPE_FCOE_BIOS;
4429                 break;
4430         case IMAGE_FIRMWARE_BACKUP_iSCSI:
4431                 img_optype = OPTYPE_ISCSI_BACKUP;
4432                 break;
4433         case IMAGE_NCSI:
4434                 img_optype = OPTYPE_NCSI_FW;
4435                 break;
4436         case IMAGE_FLASHISM_JUMPVECTOR:
4437                 img_optype = OPTYPE_FLASHISM_JUMPVECTOR;
4438                 break;
4439         case IMAGE_FIRMWARE_PHY:
4440                 img_optype = OPTYPE_SH_PHY_FW;
4441                 break;
4442         case IMAGE_REDBOOT_DIR:
4443                 img_optype = OPTYPE_REDBOOT_DIR;
4444                 break;
4445         case IMAGE_REDBOOT_CONFIG:
4446                 img_optype = OPTYPE_REDBOOT_CONFIG;
4447                 break;
4448         case IMAGE_UFI_DIR:
4449                 img_optype = OPTYPE_UFI_DIR;
4450                 break;
4451         default:
4452                 break;
4453         }
4454
4455         return img_optype;
4456 }
4457
4458 static int be_flash_skyhawk(struct be_adapter *adapter,
4459                             const struct firmware *fw,
4460                             struct be_dma_mem *flash_cmd, int num_of_images)
4461 {
4462         int img_hdrs_size = num_of_images * sizeof(struct image_hdr);
4463         bool crc_match, old_fw_img, flash_offset_support = true;
4464         struct device *dev = &adapter->pdev->dev;
4465         struct flash_section_info *fsec = NULL;
4466         u32 img_offset, img_size, img_type;
4467         u16 img_optype, flash_optype;
4468         int status, i, filehdr_size;
4469         const u8 *p;
4470
4471         filehdr_size = sizeof(struct flash_file_hdr_g3);
4472         fsec = get_fsec_info(adapter, filehdr_size + img_hdrs_size, fw);
4473         if (!fsec) {
4474                 dev_err(dev, "Invalid Cookie. FW image may be corrupted\n");
4475                 return -EINVAL;
4476         }
4477
4478 retry_flash:
4479         for (i = 0; i < le32_to_cpu(fsec->fsec_hdr.num_images); i++) {
4480                 img_offset = le32_to_cpu(fsec->fsec_entry[i].offset);
4481                 img_size   = le32_to_cpu(fsec->fsec_entry[i].pad_size);
4482                 img_type   = le32_to_cpu(fsec->fsec_entry[i].type);
4483                 img_optype = be_get_img_optype(fsec->fsec_entry[i]);
4484                 old_fw_img = fsec->fsec_entry[i].optype == 0xFFFF;
4485
4486                 if (img_optype == 0xFFFF)
4487                         continue;
4488
4489                 if (flash_offset_support)
4490                         flash_optype = OPTYPE_OFFSET_SPECIFIED;
4491                 else
4492                         flash_optype = img_optype;
4493
4494                 /* Don't bother verifying CRC if an old FW image is being
4495                  * flashed
4496                  */
4497                 if (old_fw_img)
4498                         goto flash;
4499
4500                 status = be_check_flash_crc(adapter, fw->data, img_offset,
4501                                             img_size, filehdr_size +
4502                                             img_hdrs_size, flash_optype,
4503                                             &crc_match);
4504                 if (base_status(status) == MCC_STATUS_ILLEGAL_REQUEST ||
4505                     base_status(status) == MCC_STATUS_ILLEGAL_FIELD) {
4506                         /* The current FW image on the card does not support
4507                          * OFFSET based flashing. Retry using older mechanism
4508                          * of OPTYPE based flashing
4509                          */
4510                         if (flash_optype == OPTYPE_OFFSET_SPECIFIED) {
4511                                 flash_offset_support = false;
4512                                 goto retry_flash;
4513                         }
4514
4515                         /* The current FW image on the card does not recognize
4516                          * the new FLASH op_type. The FW download is partially
4517                          * complete. Reboot the server now to enable FW image
4518                          * to recognize the new FLASH op_type. To complete the
4519                          * remaining process, download the same FW again after
4520                          * the reboot.
4521                          */
4522                         dev_err(dev, "Flash incomplete. Reset the server\n");
4523                         dev_err(dev, "Download FW image again after reset\n");
4524                         return -EAGAIN;
4525                 } else if (status) {
4526                         dev_err(dev, "Could not get CRC for 0x%x region\n",
4527                                 img_optype);
4528                         return -EFAULT;
4529                 }
4530
4531                 if (crc_match)
4532                         continue;
4533
4534 flash:
4535                 p = fw->data + filehdr_size + img_offset + img_hdrs_size;
4536                 if (p + img_size > fw->data + fw->size)
4537                         return -1;
4538
4539                 status = be_flash(adapter, p, flash_cmd, flash_optype, img_size,
4540                                   img_offset);
4541
4542                 /* The current FW image on the card does not support OFFSET
4543                  * based flashing. Retry using older mechanism of OPTYPE based
4544                  * flashing
4545                  */
4546                 if (base_status(status) == MCC_STATUS_ILLEGAL_FIELD &&
4547                     flash_optype == OPTYPE_OFFSET_SPECIFIED) {
4548                         flash_offset_support = false;
4549                         goto retry_flash;
4550                 }
4551
4552                 /* For old FW images ignore ILLEGAL_FIELD error or errors on
4553                  * UFI_DIR region
4554                  */
4555                 if (old_fw_img &&
4556                     (base_status(status) == MCC_STATUS_ILLEGAL_FIELD ||
4557                      (img_optype == OPTYPE_UFI_DIR &&
4558                       base_status(status) == MCC_STATUS_FAILED))) {
4559                         continue;
4560                 } else if (status) {
4561                         dev_err(dev, "Flashing section type 0x%x failed\n",
4562                                 img_type);
4563                         return -EFAULT;
4564                 }
4565         }
4566         return 0;
4567 }
4568
4569 static int lancer_fw_download(struct be_adapter *adapter,
4570                               const struct firmware *fw)
4571 {
4572 #define LANCER_FW_DOWNLOAD_CHUNK      (32 * 1024)
4573 #define LANCER_FW_DOWNLOAD_LOCATION   "/prg"
4574         struct device *dev = &adapter->pdev->dev;
4575         struct be_dma_mem flash_cmd;
4576         const u8 *data_ptr = NULL;
4577         u8 *dest_image_ptr = NULL;
4578         size_t image_size = 0;
4579         u32 chunk_size = 0;
4580         u32 data_written = 0;
4581         u32 offset = 0;
4582         int status = 0;
4583         u8 add_status = 0;
4584         u8 change_status;
4585
4586         if (!IS_ALIGNED(fw->size, sizeof(u32))) {
4587                 dev_err(dev, "FW image size should be multiple of 4\n");
4588                 return -EINVAL;
4589         }
4590
4591         flash_cmd.size = sizeof(struct lancer_cmd_req_write_object)
4592                                 + LANCER_FW_DOWNLOAD_CHUNK;
4593         flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size,
4594                                           &flash_cmd.dma, GFP_KERNEL);
4595         if (!flash_cmd.va)
4596                 return -ENOMEM;
4597
4598         dest_image_ptr = flash_cmd.va +
4599                                 sizeof(struct lancer_cmd_req_write_object);
4600         image_size = fw->size;
4601         data_ptr = fw->data;
4602
4603         while (image_size) {
4604                 chunk_size = min_t(u32, image_size, LANCER_FW_DOWNLOAD_CHUNK);
4605
4606                 /* Copy the image chunk content. */
4607                 memcpy(dest_image_ptr, data_ptr, chunk_size);
4608
4609                 status = lancer_cmd_write_object(adapter, &flash_cmd,
4610                                                  chunk_size, offset,
4611                                                  LANCER_FW_DOWNLOAD_LOCATION,
4612                                                  &data_written, &change_status,
4613                                                  &add_status);
4614                 if (status)
4615                         break;
4616
4617                 offset += data_written;
4618                 data_ptr += data_written;
4619                 image_size -= data_written;
4620         }
4621
4622         if (!status) {
4623                 /* Commit the FW written */
4624                 status = lancer_cmd_write_object(adapter, &flash_cmd,
4625                                                  0, offset,
4626                                                  LANCER_FW_DOWNLOAD_LOCATION,
4627                                                  &data_written, &change_status,
4628                                                  &add_status);
4629         }
4630
4631         dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4632         if (status) {
4633                 dev_err(dev, "Firmware load error\n");
4634                 return be_cmd_status(status);
4635         }
4636
4637         dev_info(dev, "Firmware flashed successfully\n");
4638
4639         if (change_status == LANCER_FW_RESET_NEEDED) {
4640                 dev_info(dev, "Resetting adapter to activate new FW\n");
4641                 status = lancer_physdev_ctrl(adapter,
4642                                              PHYSDEV_CONTROL_FW_RESET_MASK);
4643                 if (status) {
4644                         dev_err(dev, "Adapter busy, could not reset FW\n");
4645                         dev_err(dev, "Reboot server to activate new FW\n");
4646                 }
4647         } else if (change_status != LANCER_NO_RESET_NEEDED) {
4648                 dev_info(dev, "Reboot server to activate new FW\n");
4649         }
4650
4651         return 0;
4652 }
4653
4654 #define BE2_UFI         2
4655 #define BE3_UFI         3
4656 #define BE3R_UFI        10
4657 #define SH_UFI          4
4658 #define SH_P2_UFI       11
4659
4660 static int be_get_ufi_type(struct be_adapter *adapter,
4661                            struct flash_file_hdr_g3 *fhdr)
4662 {
4663         if (!fhdr) {
4664                 dev_err(&adapter->pdev->dev, "Invalid FW UFI file");
4665                 return -1;
4666         }
4667
4668         /* First letter of the build version is used to identify
4669          * which chip this image file is meant for.
4670          */
4671         switch (fhdr->build[0]) {
4672         case BLD_STR_UFI_TYPE_SH:
4673                 return (fhdr->asic_type_rev == ASIC_REV_P2) ? SH_P2_UFI :
4674                                                                 SH_UFI;
4675         case BLD_STR_UFI_TYPE_BE3:
4676                 return (fhdr->asic_type_rev == ASIC_REV_B0) ? BE3R_UFI :
4677                                                                 BE3_UFI;
4678         case BLD_STR_UFI_TYPE_BE2:
4679                 return BE2_UFI;
4680         default:
4681                 return -1;
4682         }
4683 }
4684
4685 /* Check if the flash image file is compatible with the adapter that
4686  * is being flashed.
4687  * BE3 chips with asic-rev B0 must be flashed only with BE3R_UFI type.
4688  * Skyhawk chips with asic-rev P2 must be flashed only with SH_P2_UFI type.
4689  */
4690 static bool be_check_ufi_compatibility(struct be_adapter *adapter,
4691                                        struct flash_file_hdr_g3 *fhdr)
4692 {
4693         int ufi_type = be_get_ufi_type(adapter, fhdr);
4694
4695         switch (ufi_type) {
4696         case SH_P2_UFI:
4697                 return skyhawk_chip(adapter);
4698         case SH_UFI:
4699                 return (skyhawk_chip(adapter) &&
4700                         adapter->asic_rev < ASIC_REV_P2);
4701         case BE3R_UFI:
4702                 return BE3_chip(adapter);
4703         case BE3_UFI:
4704                 return (BE3_chip(adapter) && adapter->asic_rev < ASIC_REV_B0);
4705         case BE2_UFI:
4706                 return BE2_chip(adapter);
4707         default:
4708                 return false;
4709         }
4710 }
4711
4712 static int be_fw_download(struct be_adapter *adapter, const struct firmware* fw)
4713 {
4714         struct device *dev = &adapter->pdev->dev;
4715         struct flash_file_hdr_g3 *fhdr3;
4716         struct image_hdr *img_hdr_ptr;
4717         int status = 0, i, num_imgs;
4718         struct be_dma_mem flash_cmd;
4719
4720         fhdr3 = (struct flash_file_hdr_g3 *)fw->data;
4721         if (!be_check_ufi_compatibility(adapter, fhdr3)) {
4722                 dev_err(dev, "Flash image is not compatible with adapter\n");
4723                 return -EINVAL;
4724         }
4725
4726         flash_cmd.size = sizeof(struct be_cmd_write_flashrom);
4727         flash_cmd.va = dma_alloc_coherent(dev, flash_cmd.size, &flash_cmd.dma,
4728                                           GFP_KERNEL);
4729         if (!flash_cmd.va)
4730                 return -ENOMEM;
4731
4732         num_imgs = le32_to_cpu(fhdr3->num_imgs);
4733         for (i = 0; i < num_imgs; i++) {
4734                 img_hdr_ptr = (struct image_hdr *)(fw->data +
4735                                 (sizeof(struct flash_file_hdr_g3) +
4736                                  i * sizeof(struct image_hdr)));
4737                 if (!BE2_chip(adapter) &&
4738                     le32_to_cpu(img_hdr_ptr->imageid) != 1)
4739                         continue;
4740
4741                 if (skyhawk_chip(adapter))
4742                         status = be_flash_skyhawk(adapter, fw, &flash_cmd,
4743                                                   num_imgs);
4744                 else
4745                         status = be_flash_BEx(adapter, fw, &flash_cmd,
4746                                               num_imgs);
4747         }
4748
4749         dma_free_coherent(dev, flash_cmd.size, flash_cmd.va, flash_cmd.dma);
4750         if (!status)
4751                 dev_info(dev, "Firmware flashed successfully\n");
4752
4753         return status;
4754 }
4755
4756 int be_load_fw(struct be_adapter *adapter, u8 *fw_file)
4757 {
4758         const struct firmware *fw;
4759         int status;
4760
4761         if (!netif_running(adapter->netdev)) {
4762                 dev_err(&adapter->pdev->dev,
4763                         "Firmware load not allowed (interface is down)\n");
4764                 return -ENETDOWN;
4765         }
4766
4767         status = request_firmware(&fw, fw_file, &adapter->pdev->dev);
4768         if (status)
4769                 goto fw_exit;
4770
4771         dev_info(&adapter->pdev->dev, "Flashing firmware file %s\n", fw_file);
4772
4773         if (lancer_chip(adapter))
4774                 status = lancer_fw_download(adapter, fw);
4775         else
4776                 status = be_fw_download(adapter, fw);
4777
4778         if (!status)
4779                 be_cmd_get_fw_ver(adapter);
4780
4781 fw_exit:
4782         release_firmware(fw);
4783         return status;
4784 }
4785
4786 static int be_ndo_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
4787                                  u16 flags)
4788 {
4789         struct be_adapter *adapter = netdev_priv(dev);
4790         struct nlattr *attr, *br_spec;
4791         int rem;
4792         int status = 0;
4793         u16 mode = 0;
4794
4795         if (!sriov_enabled(adapter))
4796                 return -EOPNOTSUPP;
4797
4798         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4799         if (!br_spec)
4800                 return -EINVAL;
4801
4802         nla_for_each_nested(attr, br_spec, rem) {
4803                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4804                         continue;
4805
4806                 if (nla_len(attr) < sizeof(mode))
4807                         return -EINVAL;
4808
4809                 mode = nla_get_u16(attr);
4810                 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4811                         return -EINVAL;
4812
4813                 status = be_cmd_set_hsw_config(adapter, 0, 0,
4814                                                adapter->if_handle,
4815                                                mode == BRIDGE_MODE_VEPA ?
4816                                                PORT_FWD_TYPE_VEPA :
4817                                                PORT_FWD_TYPE_VEB);
4818                 if (status)
4819                         goto err;
4820
4821                 dev_info(&adapter->pdev->dev, "enabled switch mode: %s\n",
4822                          mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4823
4824                 return status;
4825         }
4826 err:
4827         dev_err(&adapter->pdev->dev, "Failed to set switch mode %s\n",
4828                 mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
4829
4830         return status;
4831 }
4832
4833 static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4834                                  struct net_device *dev, u32 filter_mask)
4835 {
4836         struct be_adapter *adapter = netdev_priv(dev);
4837         int status = 0;
4838         u8 hsw_mode;
4839
4840         if (!sriov_enabled(adapter))
4841                 return 0;
4842
4843         /* BE and Lancer chips support VEB mode only */
4844         if (BEx_chip(adapter) || lancer_chip(adapter)) {
4845                 hsw_mode = PORT_FWD_TYPE_VEB;
4846         } else {
4847                 status = be_cmd_get_hsw_config(adapter, NULL, 0,
4848                                                adapter->if_handle, &hsw_mode);
4849                 if (status)
4850                         return 0;
4851         }
4852
4853         return ndo_dflt_bridge_getlink(skb, pid, seq, dev,
4854                                        hsw_mode == PORT_FWD_TYPE_VEPA ?
4855                                        BRIDGE_MODE_VEPA : BRIDGE_MODE_VEB,
4856                                        0, 0);
4857 }
4858
4859 #ifdef CONFIG_BE2NET_VXLAN
4860 /* VxLAN offload Notes:
4861  *
4862  * The stack defines tunnel offload flags (hw_enc_features) for IP and doesn't
4863  * distinguish various types of transports (VxLAN, GRE, NVGRE ..). So, offload
4864  * is expected to work across all types of IP tunnels once exported. Skyhawk
4865  * supports offloads for either VxLAN or NVGRE, exclusively. So we export VxLAN
4866  * offloads in hw_enc_features only when a VxLAN port is added. If other (non
4867  * VxLAN) tunnels are configured while VxLAN offloads are enabled, offloads for
4868  * those other tunnels are unexported on the fly through ndo_features_check().
4869  *
4870  * Skyhawk supports VxLAN offloads only for one UDP dport. So, if the stack
4871  * adds more than one port, disable offloads and don't re-enable them again
4872  * until after all the tunnels are removed.
4873  */
4874 static void be_add_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4875                               __be16 port)
4876 {
4877         struct be_adapter *adapter = netdev_priv(netdev);
4878         struct device *dev = &adapter->pdev->dev;
4879         int status;
4880
4881         if (lancer_chip(adapter) || BEx_chip(adapter))
4882                 return;
4883
4884         if (adapter->flags & BE_FLAGS_VXLAN_OFFLOADS) {
4885                 dev_info(dev,
4886                          "Only one UDP port supported for VxLAN offloads\n");
4887                 dev_info(dev, "Disabling VxLAN offloads\n");
4888                 adapter->vxlan_port_count++;
4889                 goto err;
4890         }
4891
4892         if (adapter->vxlan_port_count++ >= 1)
4893                 return;
4894
4895         status = be_cmd_manage_iface(adapter, adapter->if_handle,
4896                                      OP_CONVERT_NORMAL_TO_TUNNEL);
4897         if (status) {
4898                 dev_warn(dev, "Failed to convert normal interface to tunnel\n");
4899                 goto err;
4900         }
4901
4902         status = be_cmd_set_vxlan_port(adapter, port);
4903         if (status) {
4904                 dev_warn(dev, "Failed to add VxLAN port\n");
4905                 goto err;
4906         }
4907         adapter->flags |= BE_FLAGS_VXLAN_OFFLOADS;
4908         adapter->vxlan_port = port;
4909
4910         netdev->hw_enc_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
4911                                    NETIF_F_TSO | NETIF_F_TSO6 |
4912                                    NETIF_F_GSO_UDP_TUNNEL;
4913         netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
4914         netdev->features |= NETIF_F_GSO_UDP_TUNNEL;
4915
4916         dev_info(dev, "Enabled VxLAN offloads for UDP port %d\n",
4917                  be16_to_cpu(port));
4918         return;
4919 err:
4920         be_disable_vxlan_offloads(adapter);
4921 }
4922
4923 static void be_del_vxlan_port(struct net_device *netdev, sa_family_t sa_family,
4924                               __be16 port)
4925 {
4926         struct be_adapter *adapter = netdev_priv(netdev);
4927
4928         if (lancer_chip(adapter) || BEx_chip(adapter))
4929                 return;
4930
4931         if (adapter->vxlan_port != port)
4932                 goto done;
4933
4934         be_disable_vxlan_offloads(adapter);
4935
4936         dev_info(&adapter->pdev->dev,
4937                  "Disabled VxLAN offloads for UDP port %d\n",
4938                  be16_to_cpu(port));
4939 done:
4940         adapter->vxlan_port_count--;
4941 }
4942
4943 static netdev_features_t be_features_check(struct sk_buff *skb,
4944                                            struct net_device *dev,
4945                                            netdev_features_t features)
4946 {
4947         struct be_adapter *adapter = netdev_priv(dev);
4948         u8 l4_hdr = 0;
4949
4950         /* The code below restricts offload features for some tunneled packets.
4951          * Offload features for normal (non tunnel) packets are unchanged.
4952          */
4953         if (!skb->encapsulation ||
4954             !(adapter->flags & BE_FLAGS_VXLAN_OFFLOADS))
4955                 return features;
4956
4957         /* It's an encapsulated packet and VxLAN offloads are enabled. We
4958          * should disable tunnel offload features if it's not a VxLAN packet,
4959          * as tunnel offloads have been enabled only for VxLAN. This is done to
4960          * allow other tunneled traffic like GRE work fine while VxLAN
4961          * offloads are configured in Skyhawk-R.
4962          */
4963         switch (vlan_get_protocol(skb)) {
4964         case htons(ETH_P_IP):
4965                 l4_hdr = ip_hdr(skb)->protocol;
4966                 break;
4967         case htons(ETH_P_IPV6):
4968                 l4_hdr = ipv6_hdr(skb)->nexthdr;
4969                 break;
4970         default:
4971                 return features;
4972         }
4973
4974         if (l4_hdr != IPPROTO_UDP ||
4975             skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
4976             skb->inner_protocol != htons(ETH_P_TEB) ||
4977             skb_inner_mac_header(skb) - skb_transport_header(skb) !=
4978             sizeof(struct udphdr) + sizeof(struct vxlanhdr))
4979                 return features & ~(NETIF_F_ALL_CSUM | NETIF_F_GSO_MASK);
4980
4981         return features;
4982 }
4983 #endif
4984
4985 static const struct net_device_ops be_netdev_ops = {
4986         .ndo_open               = be_open,
4987         .ndo_stop               = be_close,
4988         .ndo_start_xmit         = be_xmit,
4989         .ndo_set_rx_mode        = be_set_rx_mode,
4990         .ndo_set_mac_address    = be_mac_addr_set,
4991         .ndo_change_mtu         = be_change_mtu,
4992         .ndo_get_stats64        = be_get_stats64,
4993         .ndo_validate_addr      = eth_validate_addr,
4994         .ndo_vlan_rx_add_vid    = be_vlan_add_vid,
4995         .ndo_vlan_rx_kill_vid   = be_vlan_rem_vid,
4996         .ndo_set_vf_mac         = be_set_vf_mac,
4997         .ndo_set_vf_vlan        = be_set_vf_vlan,
4998         .ndo_set_vf_rate        = be_set_vf_tx_rate,
4999         .ndo_get_vf_config      = be_get_vf_config,
5000         .ndo_set_vf_link_state  = be_set_vf_link_state,
5001 #ifdef CONFIG_NET_POLL_CONTROLLER
5002         .ndo_poll_controller    = be_netpoll,
5003 #endif
5004         .ndo_bridge_setlink     = be_ndo_bridge_setlink,
5005         .ndo_bridge_getlink     = be_ndo_bridge_getlink,
5006 #ifdef CONFIG_NET_RX_BUSY_POLL
5007         .ndo_busy_poll          = be_busy_poll,
5008 #endif
5009 #ifdef CONFIG_BE2NET_VXLAN
5010         .ndo_add_vxlan_port     = be_add_vxlan_port,
5011         .ndo_del_vxlan_port     = be_del_vxlan_port,
5012         .ndo_features_check     = be_features_check,
5013 #endif
5014 };
5015
5016 static void be_netdev_init(struct net_device *netdev)
5017 {
5018         struct be_adapter *adapter = netdev_priv(netdev);
5019
5020         netdev->hw_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
5021                 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM |
5022                 NETIF_F_HW_VLAN_CTAG_TX;
5023         if (be_multi_rxq(adapter))
5024                 netdev->hw_features |= NETIF_F_RXHASH;
5025
5026         netdev->features |= netdev->hw_features |
5027                 NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_FILTER;
5028
5029         netdev->vlan_features |= NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6 |
5030                 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
5031
5032         netdev->priv_flags |= IFF_UNICAST_FLT;
5033
5034         netdev->flags |= IFF_MULTICAST;
5035
5036         netif_set_gso_max_size(netdev, 65535 - ETH_HLEN);
5037
5038         netdev->netdev_ops = &be_netdev_ops;
5039
5040         netdev->ethtool_ops = &be_ethtool_ops;
5041 }
5042
5043 static void be_cleanup(struct be_adapter *adapter)
5044 {
5045         struct net_device *netdev = adapter->netdev;
5046
5047         rtnl_lock();
5048         netif_device_detach(netdev);
5049         if (netif_running(netdev))
5050                 be_close(netdev);
5051         rtnl_unlock();
5052
5053         be_clear(adapter);
5054 }
5055
5056 static int be_resume(struct be_adapter *adapter)
5057 {
5058         struct net_device *netdev = adapter->netdev;
5059         int status;
5060
5061         status = be_setup(adapter);
5062         if (status)
5063                 return status;
5064
5065         if (netif_running(netdev)) {
5066                 status = be_open(netdev);
5067                 if (status)
5068                         return status;
5069         }
5070
5071         netif_device_attach(netdev);
5072
5073         return 0;
5074 }
5075
5076 static int be_err_recover(struct be_adapter *adapter)
5077 {
5078         struct device *dev = &adapter->pdev->dev;
5079         int status;
5080
5081         status = be_resume(adapter);
5082         if (status)
5083                 goto err;
5084
5085         dev_info(dev, "Adapter recovery successful\n");
5086         return 0;
5087 err:
5088         if (be_physfn(adapter))
5089                 dev_err(dev, "Adapter recovery failed\n");
5090         else
5091                 dev_err(dev, "Re-trying adapter recovery\n");
5092
5093         return status;
5094 }
5095
5096 static void be_err_detection_task(struct work_struct *work)
5097 {
5098         struct be_adapter *adapter =
5099                                 container_of(work, struct be_adapter,
5100                                              be_err_detection_work.work);
5101         int status = 0;
5102
5103         be_detect_error(adapter);
5104
5105         if (adapter->hw_error) {
5106                 be_cleanup(adapter);
5107
5108                 /* As of now error recovery support is in Lancer only */
5109                 if (lancer_chip(adapter))
5110                         status = be_err_recover(adapter);
5111         }
5112
5113         /* Always attempt recovery on VFs */
5114         if (!status || be_virtfn(adapter))
5115                 be_schedule_err_detection(adapter);
5116 }
5117
5118 static void be_log_sfp_info(struct be_adapter *adapter)
5119 {
5120         int status;
5121
5122         status = be_cmd_query_sfp_info(adapter);
5123         if (!status) {
5124                 dev_err(&adapter->pdev->dev,
5125                         "Unqualified SFP+ detected on %c from %s part no: %s",
5126                         adapter->port_name, adapter->phy.vendor_name,
5127                         adapter->phy.vendor_pn);
5128         }
5129         adapter->flags &= ~BE_FLAGS_EVT_INCOMPATIBLE_SFP;
5130 }
5131
5132 static void be_worker(struct work_struct *work)
5133 {
5134         struct be_adapter *adapter =
5135                 container_of(work, struct be_adapter, work.work);
5136         struct be_rx_obj *rxo;
5137         int i;
5138
5139         /* when interrupts are not yet enabled, just reap any pending
5140          * mcc completions
5141          */
5142         if (!netif_running(adapter->netdev)) {
5143                 local_bh_disable();
5144                 be_process_mcc(adapter);
5145                 local_bh_enable();
5146                 goto reschedule;
5147         }
5148
5149         if (!adapter->stats_cmd_sent) {
5150                 if (lancer_chip(adapter))
5151                         lancer_cmd_get_pport_stats(adapter,
5152                                                    &adapter->stats_cmd);
5153                 else
5154                         be_cmd_get_stats(adapter, &adapter->stats_cmd);
5155         }
5156
5157         if (be_physfn(adapter) &&
5158             MODULO(adapter->work_counter, adapter->be_get_temp_freq) == 0)
5159                 be_cmd_get_die_temperature(adapter);
5160
5161         for_all_rx_queues(adapter, rxo, i) {
5162                 /* Replenish RX-queues starved due to memory
5163                  * allocation failures.
5164                  */
5165                 if (rxo->rx_post_starved)
5166                         be_post_rx_frags(rxo, GFP_KERNEL, MAX_RX_POST);
5167         }
5168
5169         be_eqd_update(adapter);
5170
5171         if (adapter->flags & BE_FLAGS_EVT_INCOMPATIBLE_SFP)
5172                 be_log_sfp_info(adapter);
5173
5174 reschedule:
5175         adapter->work_counter++;
5176         schedule_delayed_work(&adapter->work, msecs_to_jiffies(1000));
5177 }
5178
5179 static void be_unmap_pci_bars(struct be_adapter *adapter)
5180 {
5181         if (adapter->csr)
5182                 pci_iounmap(adapter->pdev, adapter->csr);
5183         if (adapter->db)
5184                 pci_iounmap(adapter->pdev, adapter->db);
5185 }
5186
5187 static int db_bar(struct be_adapter *adapter)
5188 {
5189         if (lancer_chip(adapter) || !be_physfn(adapter))
5190                 return 0;
5191         else
5192                 return 4;
5193 }
5194
5195 static int be_roce_map_pci_bars(struct be_adapter *adapter)
5196 {
5197         if (skyhawk_chip(adapter)) {
5198                 adapter->roce_db.size = 4096;
5199                 adapter->roce_db.io_addr = pci_resource_start(adapter->pdev,
5200                                                               db_bar(adapter));
5201                 adapter->roce_db.total_size = pci_resource_len(adapter->pdev,
5202                                                                db_bar(adapter));
5203         }
5204         return 0;
5205 }
5206
5207 static int be_map_pci_bars(struct be_adapter *adapter)
5208 {
5209         struct pci_dev *pdev = adapter->pdev;
5210         u8 __iomem *addr;
5211         u32 sli_intf;
5212
5213         pci_read_config_dword(adapter->pdev, SLI_INTF_REG_OFFSET, &sli_intf);
5214         adapter->sli_family = (sli_intf & SLI_INTF_FAMILY_MASK) >>
5215                                 SLI_INTF_FAMILY_SHIFT;
5216         adapter->virtfn = (sli_intf & SLI_INTF_FT_MASK) ? 1 : 0;
5217
5218         if (BEx_chip(adapter) && be_physfn(adapter)) {
5219                 adapter->csr = pci_iomap(pdev, 2, 0);
5220                 if (!adapter->csr)
5221                         return -ENOMEM;
5222         }
5223
5224         addr = pci_iomap(pdev, db_bar(adapter), 0);
5225         if (!addr)
5226                 goto pci_map_err;
5227         adapter->db = addr;
5228
5229         if (skyhawk_chip(adapter) || BEx_chip(adapter)) {
5230                 if (be_physfn(adapter)) {
5231                         /* PCICFG is the 2nd BAR in BE2 */
5232                         addr = pci_iomap(pdev, BE2_chip(adapter) ? 1 : 0, 0);
5233                         if (!addr)
5234                                 goto pci_map_err;
5235                         adapter->pcicfg = addr;
5236                 } else {
5237                         adapter->pcicfg = adapter->db + SRIOV_VF_PCICFG_OFFSET;
5238                 }
5239         }
5240
5241         be_roce_map_pci_bars(adapter);
5242         return 0;
5243
5244 pci_map_err:
5245         dev_err(&pdev->dev, "Error in mapping PCI BARs\n");
5246         be_unmap_pci_bars(adapter);
5247         return -ENOMEM;
5248 }
5249
5250 static void be_drv_cleanup(struct be_adapter *adapter)
5251 {
5252         struct be_dma_mem *mem = &adapter->mbox_mem_alloced;
5253         struct device *dev = &adapter->pdev->dev;
5254
5255         if (mem->va)
5256                 dma_free_coherent(dev, mem->size, mem->va, mem->dma);
5257
5258         mem = &adapter->rx_filter;
5259         if (mem->va)
5260                 dma_free_coherent(dev, mem->size, mem->va, mem->dma);
5261
5262         mem = &adapter->stats_cmd;
5263         if (mem->va)
5264                 dma_free_coherent(dev, mem->size, mem->va, mem->dma);
5265 }
5266
5267 /* Allocate and initialize various fields in be_adapter struct */
5268 static int be_drv_init(struct be_adapter *adapter)
5269 {
5270         struct be_dma_mem *mbox_mem_alloc = &adapter->mbox_mem_alloced;
5271         struct be_dma_mem *mbox_mem_align = &adapter->mbox_mem;
5272         struct be_dma_mem *rx_filter = &adapter->rx_filter;
5273         struct be_dma_mem *stats_cmd = &adapter->stats_cmd;
5274         struct device *dev = &adapter->pdev->dev;
5275         int status = 0;
5276
5277         mbox_mem_alloc->size = sizeof(struct be_mcc_mailbox) + 16;
5278         mbox_mem_alloc->va = dma_alloc_coherent(dev, mbox_mem_alloc->size,
5279                                                 &mbox_mem_alloc->dma,
5280                                                 GFP_KERNEL);
5281         if (!mbox_mem_alloc->va)
5282                 return -ENOMEM;
5283
5284         mbox_mem_align->size = sizeof(struct be_mcc_mailbox);
5285         mbox_mem_align->va = PTR_ALIGN(mbox_mem_alloc->va, 16);
5286         mbox_mem_align->dma = PTR_ALIGN(mbox_mem_alloc->dma, 16);
5287         memset(mbox_mem_align->va, 0, sizeof(struct be_mcc_mailbox));
5288
5289         rx_filter->size = sizeof(struct be_cmd_req_rx_filter);
5290         rx_filter->va = dma_zalloc_coherent(dev, rx_filter->size,
5291                                             &rx_filter->dma, GFP_KERNEL);
5292         if (!rx_filter->va) {
5293                 status = -ENOMEM;
5294                 goto free_mbox;
5295         }
5296
5297         if (lancer_chip(adapter))
5298                 stats_cmd->size = sizeof(struct lancer_cmd_req_pport_stats);
5299         else if (BE2_chip(adapter))
5300                 stats_cmd->size = sizeof(struct be_cmd_req_get_stats_v0);
5301         else if (BE3_chip(adapter))
5302                 stats_cmd->size = sizeof(struct be_cmd_req_get_stats_v1);
5303         else
5304                 stats_cmd->size = sizeof(struct be_cmd_req_get_stats_v2);
5305         stats_cmd->va = dma_zalloc_coherent(dev, stats_cmd->size,
5306                                             &stats_cmd->dma, GFP_KERNEL);
5307         if (!stats_cmd->va) {
5308                 status = -ENOMEM;
5309                 goto free_rx_filter;
5310         }
5311
5312         mutex_init(&adapter->mbox_lock);
5313         spin_lock_init(&adapter->mcc_lock);
5314         spin_lock_init(&adapter->mcc_cq_lock);
5315         init_completion(&adapter->et_cmd_compl);
5316
5317         pci_save_state(adapter->pdev);
5318
5319         INIT_DELAYED_WORK(&adapter->work, be_worker);
5320         INIT_DELAYED_WORK(&adapter->be_err_detection_work,
5321                           be_err_detection_task);
5322
5323         adapter->rx_fc = true;
5324         adapter->tx_fc = true;
5325
5326         /* Must be a power of 2 or else MODULO will BUG_ON */
5327         adapter->be_get_temp_freq = 64;
5328
5329         return 0;
5330
5331 free_rx_filter:
5332         dma_free_coherent(dev, rx_filter->size, rx_filter->va, rx_filter->dma);
5333 free_mbox:
5334         dma_free_coherent(dev, mbox_mem_alloc->size, mbox_mem_alloc->va,
5335                           mbox_mem_alloc->dma);
5336         return status;
5337 }
5338
5339 static void be_remove(struct pci_dev *pdev)
5340 {
5341         struct be_adapter *adapter = pci_get_drvdata(pdev);
5342
5343         if (!adapter)
5344                 return;
5345
5346         be_roce_dev_remove(adapter);
5347         be_intr_set(adapter, false);
5348
5349         be_cancel_err_detection(adapter);
5350
5351         unregister_netdev(adapter->netdev);
5352
5353         be_clear(adapter);
5354
5355         /* tell fw we're done with firing cmds */
5356         be_cmd_fw_clean(adapter);
5357
5358         be_unmap_pci_bars(adapter);
5359         be_drv_cleanup(adapter);
5360
5361         pci_disable_pcie_error_reporting(pdev);
5362
5363         pci_release_regions(pdev);
5364         pci_disable_device(pdev);
5365
5366         free_netdev(adapter->netdev);
5367 }
5368
5369 static char *mc_name(struct be_adapter *adapter)
5370 {
5371         char *str = ""; /* default */
5372
5373         switch (adapter->mc_type) {
5374         case UMC:
5375                 str = "UMC";
5376                 break;
5377         case FLEX10:
5378                 str = "FLEX10";
5379                 break;
5380         case vNIC1:
5381                 str = "vNIC-1";
5382                 break;
5383         case nPAR:
5384                 str = "nPAR";
5385                 break;
5386         case UFP:
5387                 str = "UFP";
5388                 break;
5389         case vNIC2:
5390                 str = "vNIC-2";
5391                 break;
5392         default:
5393                 str = "";
5394         }
5395
5396         return str;
5397 }
5398
5399 static inline char *func_name(struct be_adapter *adapter)
5400 {
5401         return be_physfn(adapter) ? "PF" : "VF";
5402 }
5403
5404 static inline char *nic_name(struct pci_dev *pdev)
5405 {
5406         switch (pdev->device) {
5407         case OC_DEVICE_ID1:
5408                 return OC_NAME;
5409         case OC_DEVICE_ID2:
5410                 return OC_NAME_BE;
5411         case OC_DEVICE_ID3:
5412         case OC_DEVICE_ID4:
5413                 return OC_NAME_LANCER;
5414         case BE_DEVICE_ID2:
5415                 return BE3_NAME;
5416         case OC_DEVICE_ID5:
5417         case OC_DEVICE_ID6:
5418                 return OC_NAME_SH;
5419         default:
5420                 return BE_NAME;
5421         }
5422 }
5423
5424 static int be_probe(struct pci_dev *pdev, const struct pci_device_id *pdev_id)
5425 {
5426         struct be_adapter *adapter;
5427         struct net_device *netdev;
5428         int status = 0;
5429
5430         dev_info(&pdev->dev, "%s version is %s\n", DRV_NAME, DRV_VER);
5431
5432         status = pci_enable_device(pdev);
5433         if (status)
5434                 goto do_none;
5435
5436         status = pci_request_regions(pdev, DRV_NAME);
5437         if (status)
5438                 goto disable_dev;
5439         pci_set_master(pdev);
5440
5441         netdev = alloc_etherdev_mqs(sizeof(*adapter), MAX_TX_QS, MAX_RX_QS);
5442         if (!netdev) {
5443                 status = -ENOMEM;
5444                 goto rel_reg;
5445         }
5446         adapter = netdev_priv(netdev);
5447         adapter->pdev = pdev;
5448         pci_set_drvdata(pdev, adapter);
5449         adapter->netdev = netdev;
5450         SET_NETDEV_DEV(netdev, &pdev->dev);
5451
5452         status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
5453         if (!status) {
5454                 netdev->features |= NETIF_F_HIGHDMA;
5455         } else {
5456                 status = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
5457                 if (status) {
5458                         dev_err(&pdev->dev, "Could not set PCI DMA Mask\n");
5459                         goto free_netdev;
5460                 }
5461         }
5462
5463         status = pci_enable_pcie_error_reporting(pdev);
5464         if (!status)
5465                 dev_info(&pdev->dev, "PCIe error reporting enabled\n");
5466
5467         status = be_map_pci_bars(adapter);
5468         if (status)
5469                 goto free_netdev;
5470
5471         status = be_drv_init(adapter);
5472         if (status)
5473                 goto unmap_bars;
5474
5475         status = be_setup(adapter);
5476         if (status)
5477                 goto drv_cleanup;
5478
5479         be_netdev_init(netdev);
5480         status = register_netdev(netdev);
5481         if (status != 0)
5482                 goto unsetup;
5483
5484         be_roce_dev_add(adapter);
5485
5486         be_schedule_err_detection(adapter);
5487
5488         dev_info(&pdev->dev, "%s: %s %s port %c\n", nic_name(pdev),
5489                  func_name(adapter), mc_name(adapter), adapter->port_name);
5490
5491         return 0;
5492
5493 unsetup:
5494         be_clear(adapter);
5495 drv_cleanup:
5496         be_drv_cleanup(adapter);
5497 unmap_bars:
5498         be_unmap_pci_bars(adapter);
5499 free_netdev:
5500         free_netdev(netdev);
5501 rel_reg:
5502         pci_release_regions(pdev);
5503 disable_dev:
5504         pci_disable_device(pdev);
5505 do_none:
5506         dev_err(&pdev->dev, "%s initialization failed\n", nic_name(pdev));
5507         return status;
5508 }
5509
5510 static int be_suspend(struct pci_dev *pdev, pm_message_t state)
5511 {
5512         struct be_adapter *adapter = pci_get_drvdata(pdev);
5513
5514         if (adapter->wol_en)
5515                 be_setup_wol(adapter, true);
5516
5517         be_intr_set(adapter, false);
5518         be_cancel_err_detection(adapter);
5519
5520         be_cleanup(adapter);
5521
5522         pci_save_state(pdev);
5523         pci_disable_device(pdev);
5524         pci_set_power_state(pdev, pci_choose_state(pdev, state));
5525         return 0;
5526 }
5527
5528 static int be_pci_resume(struct pci_dev *pdev)
5529 {
5530         struct be_adapter *adapter = pci_get_drvdata(pdev);
5531         int status = 0;
5532
5533         status = pci_enable_device(pdev);
5534         if (status)
5535                 return status;
5536
5537         pci_set_power_state(pdev, PCI_D0);
5538         pci_restore_state(pdev);
5539
5540         status = be_resume(adapter);
5541         if (status)
5542                 return status;
5543
5544         be_schedule_err_detection(adapter);
5545
5546         if (adapter->wol_en)
5547                 be_setup_wol(adapter, false);
5548
5549         return 0;
5550 }
5551
5552 /*
5553  * An FLR will stop BE from DMAing any data.
5554  */
5555 static void be_shutdown(struct pci_dev *pdev)
5556 {
5557         struct be_adapter *adapter = pci_get_drvdata(pdev);
5558
5559         if (!adapter)
5560                 return;
5561
5562         be_roce_dev_shutdown(adapter);
5563         cancel_delayed_work_sync(&adapter->work);
5564         be_cancel_err_detection(adapter);
5565
5566         netif_device_detach(adapter->netdev);
5567
5568         be_cmd_reset_function(adapter);
5569
5570         pci_disable_device(pdev);
5571 }
5572
5573 static pci_ers_result_t be_eeh_err_detected(struct pci_dev *pdev,
5574                                             pci_channel_state_t state)
5575 {
5576         struct be_adapter *adapter = pci_get_drvdata(pdev);
5577
5578         dev_err(&adapter->pdev->dev, "EEH error detected\n");
5579
5580         if (!adapter->eeh_error) {
5581                 adapter->eeh_error = true;
5582
5583                 be_cancel_err_detection(adapter);
5584
5585                 be_cleanup(adapter);
5586         }
5587
5588         if (state == pci_channel_io_perm_failure)
5589                 return PCI_ERS_RESULT_DISCONNECT;
5590
5591         pci_disable_device(pdev);
5592
5593         /* The error could cause the FW to trigger a flash debug dump.
5594          * Resetting the card while flash dump is in progress
5595          * can cause it not to recover; wait for it to finish.
5596          * Wait only for first function as it is needed only once per
5597          * adapter.
5598          */
5599         if (pdev->devfn == 0)
5600                 ssleep(30);
5601
5602         return PCI_ERS_RESULT_NEED_RESET;
5603 }
5604
5605 static pci_ers_result_t be_eeh_reset(struct pci_dev *pdev)
5606 {
5607         struct be_adapter *adapter = pci_get_drvdata(pdev);
5608         int status;
5609
5610         dev_info(&adapter->pdev->dev, "EEH reset\n");
5611
5612         status = pci_enable_device(pdev);
5613         if (status)
5614                 return PCI_ERS_RESULT_DISCONNECT;
5615
5616         pci_set_master(pdev);
5617         pci_set_power_state(pdev, PCI_D0);
5618         pci_restore_state(pdev);
5619
5620         /* Check if card is ok and fw is ready */
5621         dev_info(&adapter->pdev->dev,
5622                  "Waiting for FW to be ready after EEH reset\n");
5623         status = be_fw_wait_ready(adapter);
5624         if (status)
5625                 return PCI_ERS_RESULT_DISCONNECT;
5626
5627         pci_cleanup_aer_uncorrect_error_status(pdev);
5628         be_clear_all_error(adapter);
5629         return PCI_ERS_RESULT_RECOVERED;
5630 }
5631
5632 static void be_eeh_resume(struct pci_dev *pdev)
5633 {
5634         int status = 0;
5635         struct be_adapter *adapter = pci_get_drvdata(pdev);
5636
5637         dev_info(&adapter->pdev->dev, "EEH resume\n");
5638
5639         pci_save_state(pdev);
5640
5641         status = be_resume(adapter);
5642         if (status)
5643                 goto err;
5644
5645         be_schedule_err_detection(adapter);
5646         return;
5647 err:
5648         dev_err(&adapter->pdev->dev, "EEH resume failed\n");
5649 }
5650
5651 static int be_pci_sriov_configure(struct pci_dev *pdev, int num_vfs)
5652 {
5653         struct be_adapter *adapter = pci_get_drvdata(pdev);
5654         u16 num_vf_qs;
5655         int status;
5656
5657         if (!num_vfs)
5658                 be_vf_clear(adapter);
5659
5660         adapter->num_vfs = num_vfs;
5661
5662         if (adapter->num_vfs == 0 && pci_vfs_assigned(pdev)) {
5663                 dev_warn(&pdev->dev,
5664                          "Cannot disable VFs while they are assigned\n");
5665                 return -EBUSY;
5666         }
5667
5668         /* When the HW is in SRIOV capable configuration, the PF-pool resources
5669          * are equally distributed across the max-number of VFs. The user may
5670          * request only a subset of the max-vfs to be enabled.
5671          * Based on num_vfs, redistribute the resources across num_vfs so that
5672          * each VF will have access to more number of resources.
5673          * This facility is not available in BE3 FW.
5674          * Also, this is done by FW in Lancer chip.
5675          */
5676         if (skyhawk_chip(adapter) && !pci_num_vf(pdev)) {
5677                 num_vf_qs = be_calculate_vf_qs(adapter, adapter->num_vfs);
5678                 status = be_cmd_set_sriov_config(adapter, adapter->pool_res,
5679                                                  adapter->num_vfs, num_vf_qs);
5680                 if (status)
5681                         dev_err(&pdev->dev,
5682                                 "Failed to optimize SR-IOV resources\n");
5683         }
5684
5685         status = be_get_resources(adapter);
5686         if (status)
5687                 return be_cmd_status(status);
5688
5689         /* Updating real_num_tx/rx_queues() requires rtnl_lock() */
5690         rtnl_lock();
5691         status = be_update_queues(adapter);
5692         rtnl_unlock();
5693         if (status)
5694                 return be_cmd_status(status);
5695
5696         if (adapter->num_vfs)
5697                 status = be_vf_setup(adapter);
5698
5699         if (!status)
5700                 return adapter->num_vfs;
5701
5702         return 0;
5703 }
5704
5705 static const struct pci_error_handlers be_eeh_handlers = {
5706         .error_detected = be_eeh_err_detected,
5707         .slot_reset = be_eeh_reset,
5708         .resume = be_eeh_resume,
5709 };
5710
5711 static struct pci_driver be_driver = {
5712         .name = DRV_NAME,
5713         .id_table = be_dev_ids,
5714         .probe = be_probe,
5715         .remove = be_remove,
5716         .suspend = be_suspend,
5717         .resume = be_pci_resume,
5718         .shutdown = be_shutdown,
5719         .sriov_configure = be_pci_sriov_configure,
5720         .err_handler = &be_eeh_handlers
5721 };
5722
5723 static int __init be_init_module(void)
5724 {
5725         if (rx_frag_size != 8192 && rx_frag_size != 4096 &&
5726             rx_frag_size != 2048) {
5727                 printk(KERN_WARNING DRV_NAME
5728                         " : Module param rx_frag_size must be 2048/4096/8192."
5729                         " Using 2048\n");
5730                 rx_frag_size = 2048;
5731         }
5732
5733         if (num_vfs > 0) {
5734                 pr_info(DRV_NAME " : Module param num_vfs is obsolete.");
5735                 pr_info(DRV_NAME " : Use sysfs method to enable VFs\n");
5736         }
5737
5738         return pci_register_driver(&be_driver);
5739 }
5740 module_init(be_init_module);
5741
5742 static void __exit be_exit_module(void)
5743 {
5744         pci_unregister_driver(&be_driver);
5745 }
5746 module_exit(be_exit_module);