drivers/net/ethernet/ibm/ibmvnic.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /**************************************************************************/
   3 /*                                                                        */
   4 /*  IBM System i and System p Virtual NIC Device Driver                   */
   5 /*  Copyright (C) 2014 IBM Corp.                                          */
   6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   9 /*                                                                        */
  10 /*                                                                        */
  11 /* This module contains the implementation of a virtual ethernet device   */
  12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  15 /*                                                                         */
  16 /* Messages are passed between the VNIC driver and the VNIC server using  */
  17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  18 /* issue and receive commands that initiate communication with the server */
  19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  20 /* are used by the driver to notify the server that a packet is           */
  21 /* ready for transmission or that a buffer has been added to receive a    */
  22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  23 /* driver that a packet transmission has been completed or that a packet  */
  24 /* has been received and placed in a waiting buffer.                      */
  25 /*                                                                        */
  26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  28 /* or receive has been completed, the VNIC driver is required to use      */
  29 /* "long term mapping". This entails that large, continuous DMA mapped    */
  30 /* buffers are allocated on driver initialization and these buffers are   */
  31 /* then continuously reused to pass skbs to and from the VNIC server.     */
  32 /*                                                                        */
  33 /**************************************************************************/
  34
  35 #include <linux/module.h>
  36 #include <linux/moduleparam.h>
  37 #include <linux/types.h>
  38 #include <linux/errno.h>
  39 #include <linux/completion.h>
  40 #include <linux/ioport.h>
  41 #include <linux/dma-mapping.h>
  42 #include <linux/kernel.h>
  43 #include <linux/netdevice.h>
  44 #include <linux/etherdevice.h>
  45 #include <linux/skbuff.h>
  46 #include <linux/init.h>
  47 #include <linux/delay.h>
  48 #include <linux/mm.h>
  49 #include <linux/ethtool.h>
  50 #include <linux/proc_fs.h>
  51 #include <linux/if_arp.h>
  52 #include <linux/in.h>
  53 #include <linux/ip.h>
  54 #include <linux/ipv6.h>
  55 #include <linux/irq.h>
  56 #include <linux/kthread.h>
  57 #include <linux/seq_file.h>
  58 #include <linux/interrupt.h>
  59 #include <net/net_namespace.h>
  60 #include <asm/hvcall.h>
  61 #include <linux/atomic.h>
  62 #include <asm/vio.h>
  63 #include <asm/iommu.h>
  64 #include <linux/uaccess.h>
  65 #include <asm/firmware.h>
  66 #include <linux/workqueue.h>
  67 #include <linux/if_vlan.h>
  68 #include <linux/utsname.h>
  69
  70 #include "ibmvnic.h"
  71
  72 static const char ibmvnic_driver_name[] = "ibmvnic";
  73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  74
  75 MODULE_AUTHOR("Santiago Leon");
  76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  77 MODULE_LICENSE("GPL");
  78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  79
  80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  81 static int ibmvnic_remove(struct vio_dev *);
  82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  88                        union sub_crq *sub_crq);
  89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
  90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
  91 static int enable_scrq_irq(struct ibmvnic_adapter *,
  92                            struct ibmvnic_sub_crq_queue *);
  93 static int disable_scrq_irq(struct ibmvnic_adapter *,
  94                             struct ibmvnic_sub_crq_queue *);
  95 static int pending_scrq(struct ibmvnic_adapter *,
  96                         struct ibmvnic_sub_crq_queue *);
  97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
  98                                         struct ibmvnic_sub_crq_queue *);
  99 static int ibmvnic_poll(struct napi_struct *napi, int data);
 100 static void send_map_query(struct ibmvnic_adapter *adapter);
 101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
 103 static int send_login(struct ibmvnic_adapter *adapter);
 104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 105 static int init_sub_crqs(struct ibmvnic_adapter *);
 106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 107 static int ibmvnic_init(struct ibmvnic_adapter *);
 108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
 109 static void release_crq_queue(struct ibmvnic_adapter *);
 110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
 111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
 113
 114 struct ibmvnic_stat {
 115         char name[ETH_GSTRING_LEN];
 116         int offset;
 117 };
 118
 119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 120                              offsetof(struct ibmvnic_statistics, stat))
 121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 122
 123 static const struct ibmvnic_stat ibmvnic_stats[] = {
 124         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 125         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 126         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 127         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 128         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 129         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 130         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 131         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 132         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 133         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 134         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 135         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 136         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 137         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 138         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 139         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 140         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 141         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 142         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 143         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 144         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 145         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 146 };
 147
 148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 149                           unsigned long length, unsigned long *number,
 150                           unsigned long *irq)
 151 {
 152         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 153         long rc;
 154
 155         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 156         *number = retbuf[0];
 157         *irq = retbuf[1];
 158
 159         return rc;
 160 }
 161
 162 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 163                                 struct ibmvnic_long_term_buff *ltb, int size)
 164 {
 165         struct device *dev = &adapter->vdev->dev;
 166         int rc;
 167
 168         ltb->size = size;
 169         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 170                                        GFP_KERNEL);
 171
 172         if (!ltb->buff) {
 173                 dev_err(dev, "Couldn't alloc long term buffer\n");
 174                 return -ENOMEM;
 175         }
 176         ltb->map_id = adapter->map_id;
 177         adapter->map_id++;
 178
 179         init_completion(&adapter->fw_done);
 180         rc = send_request_map(adapter, ltb->addr,
 181                               ltb->size, ltb->map_id);
 182         if (rc) {
 183                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 184                 return rc;
 185         }
 186         wait_for_completion(&adapter->fw_done);
 187
 188         if (adapter->fw_done_rc) {
 189                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 190                         adapter->fw_done_rc);
 191                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 192                 return -1;
 193         }
 194         return 0;
 195 }
 196
 197 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 198                                 struct ibmvnic_long_term_buff *ltb)
 199 {
 200         struct device *dev = &adapter->vdev->dev;
 201
 202         if (!ltb->buff)
 203                 return;
 204
 205         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 206             adapter->reset_reason != VNIC_RESET_MOBILITY)
 207                 send_request_unmap(adapter, ltb->map_id);
 208         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 209 }
 210
 211 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 212                                 struct ibmvnic_long_term_buff *ltb)
 213 {
 214         int rc;
 215
 216         memset(ltb->buff, 0, ltb->size);
 217
 218         init_completion(&adapter->fw_done);
 219         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 220         if (rc)
 221                 return rc;
 222         wait_for_completion(&adapter->fw_done);
 223
 224         if (adapter->fw_done_rc) {
 225                 dev_info(&adapter->vdev->dev,
 226                          "Reset failed, attempting to free and reallocate buffer\n");
 227                 free_long_term_buff(adapter, ltb);
 228                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 229         }
 230         return 0;
 231 }
 232
 233 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 234 {
 235         int i;
 236
 237         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 238              i++)
 239                 adapter->rx_pool[i].active = 0;
 240 }
 241
 242 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 243                               struct ibmvnic_rx_pool *pool)
 244 {
 245         int count = pool->size - atomic_read(&pool->available);
 246         struct device *dev = &adapter->vdev->dev;
 247         int buffers_added = 0;
 248         unsigned long lpar_rc;
 249         union sub_crq sub_crq;
 250         struct sk_buff *skb;
 251         unsigned int offset;
 252         dma_addr_t dma_addr;
 253         unsigned char *dst;
 254         u64 *handle_array;
 255         int shift = 0;
 256         int index;
 257         int i;
 258
 259         if (!pool->active)
 260                 return;
 261
 262         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 263                                       be32_to_cpu(adapter->login_rsp_buf->
 264                                       off_rxadd_subcrqs));
 265
 266         for (i = 0; i < count; ++i) {
 267                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 268                 if (!skb) {
 269                         dev_err(dev, "Couldn't replenish rx buff\n");
 270                         adapter->replenish_no_mem++;
 271                         break;
 272                 }
 273
 274                 index = pool->free_map[pool->next_free];
 275
 276                 if (pool->rx_buff[index].skb)
 277                         dev_err(dev, "Inconsistent free_map!\n");
 278
 279                 /* Copy the skb to the long term mapped DMA buffer */
 280                 offset = index * pool->buff_size;
 281                 dst = pool->long_term_buff.buff + offset;
 282                 memset(dst, 0, pool->buff_size);
 283                 dma_addr = pool->long_term_buff.addr + offset;
 284                 pool->rx_buff[index].data = dst;
 285
 286                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 287                 pool->rx_buff[index].dma = dma_addr;
 288                 pool->rx_buff[index].skb = skb;
 289                 pool->rx_buff[index].pool_index = pool->index;
 290                 pool->rx_buff[index].size = pool->buff_size;
 291
 292                 memset(&sub_crq, 0, sizeof(sub_crq));
 293                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 294                 sub_crq.rx_add.correlator =
 295                     cpu_to_be64((u64)&pool->rx_buff[index]);
 296                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 297                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 298
 299                 /* The length field of the sCRQ is defined to be 24 bits so the
 300                  * buffer size needs to be left shifted by a byte before it is
 301                  * converted to big endian to prevent the last byte from being
 302                  * truncated.
 303                  */
 304 #ifdef __LITTLE_ENDIAN__
 305                 shift = 8;
 306 #endif
 307                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 308
 309                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 310                                       &sub_crq);
 311                 if (lpar_rc != H_SUCCESS)
 312                         goto failure;
 313
 314                 buffers_added++;
 315                 adapter->replenish_add_buff_success++;
 316                 pool->next_free = (pool->next_free + 1) % pool->size;
 317         }
 318         atomic_add(buffers_added, &pool->available);
 319         return;
 320
 321 failure:
 322         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 323                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 324         pool->free_map[pool->next_free] = index;
 325         pool->rx_buff[index].skb = NULL;
 326
 327         dev_kfree_skb_any(skb);
 328         adapter->replenish_add_buff_failure++;
 329         atomic_add(buffers_added, &pool->available);
 330
 331         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 332                 /* Disable buffer pool replenishment and report carrier off if
 333                  * queue is closed or pending failover.
 334                  * Firmware guarantees that a signal will be sent to the
 335                  * driver, triggering a reset.
 336                  */
 337                 deactivate_rx_pools(adapter);
 338                 netif_carrier_off(adapter->netdev);
 339         }
 340 }
 341
 342 static void replenish_pools(struct ibmvnic_adapter *adapter)
 343 {
 344         int i;
 345
 346         adapter->replenish_task_cycles++;
 347         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 348              i++) {
 349                 if (adapter->rx_pool[i].active)
 350                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 351         }
 352 }
 353
 354 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 355 {
 356         kfree(adapter->tx_stats_buffers);
 357         kfree(adapter->rx_stats_buffers);
 358         adapter->tx_stats_buffers = NULL;
 359         adapter->rx_stats_buffers = NULL;
 360 }
 361
 362 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 363 {
 364         adapter->tx_stats_buffers =
 365                                 kcalloc(IBMVNIC_MAX_QUEUES,
 366                                         sizeof(struct ibmvnic_tx_queue_stats),
 367                                         GFP_KERNEL);
 368         if (!adapter->tx_stats_buffers)
 369                 return -ENOMEM;
 370
 371         adapter->rx_stats_buffers =
 372                                 kcalloc(IBMVNIC_MAX_QUEUES,
 373                                         sizeof(struct ibmvnic_rx_queue_stats),
 374                                         GFP_KERNEL);
 375         if (!adapter->rx_stats_buffers)
 376                 return -ENOMEM;
 377
 378         return 0;
 379 }
 380
 381 static void release_stats_token(struct ibmvnic_adapter *adapter)
 382 {
 383         struct device *dev = &adapter->vdev->dev;
 384
 385         if (!adapter->stats_token)
 386                 return;
 387
 388         dma_unmap_single(dev, adapter->stats_token,
 389                          sizeof(struct ibmvnic_statistics),
 390                          DMA_FROM_DEVICE);
 391         adapter->stats_token = 0;
 392 }
 393
 394 static int init_stats_token(struct ibmvnic_adapter *adapter)
 395 {
 396         struct device *dev = &adapter->vdev->dev;
 397         dma_addr_t stok;
 398
 399         stok = dma_map_single(dev, &adapter->stats,
 400                               sizeof(struct ibmvnic_statistics),
 401                               DMA_FROM_DEVICE);
 402         if (dma_mapping_error(dev, stok)) {
 403                 dev_err(dev, "Couldn't map stats buffer\n");
 404                 return -1;
 405         }
 406
 407         adapter->stats_token = stok;
 408         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 409         return 0;
 410 }
 411
 412 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 413 {
 414         struct ibmvnic_rx_pool *rx_pool;
 415         int rx_scrqs;
 416         int i, j, rc;
 417         u64 *size_array;
 418
 419         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 420                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 421
 422         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 423         for (i = 0; i < rx_scrqs; i++) {
 424                 rx_pool = &adapter->rx_pool[i];
 425
 426                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 427
 428                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 429                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 430                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 431                         rc = alloc_long_term_buff(adapter,
 432                                                   &rx_pool->long_term_buff,
 433                                                   rx_pool->size *
 434                                                   rx_pool->buff_size);
 435                 } else {
 436                         rc = reset_long_term_buff(adapter,
 437                                                   &rx_pool->long_term_buff);
 438                 }
 439
 440                 if (rc)
 441                         return rc;
 442
 443                 for (j = 0; j < rx_pool->size; j++)
 444                         rx_pool->free_map[j] = j;
 445
 446                 memset(rx_pool->rx_buff, 0,
 447                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 448
 449                 atomic_set(&rx_pool->available, 0);
 450                 rx_pool->next_alloc = 0;
 451                 rx_pool->next_free = 0;
 452                 rx_pool->active = 1;
 453         }
 454
 455         return 0;
 456 }
 457
 458 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 459 {
 460         struct ibmvnic_rx_pool *rx_pool;
 461         int i, j;
 462
 463         if (!adapter->rx_pool)
 464                 return;
 465
 466         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 467                 rx_pool = &adapter->rx_pool[i];
 468
 469                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 470
 471                 kfree(rx_pool->free_map);
 472                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 473
 474                 if (!rx_pool->rx_buff)
 475                         continue;
 476
 477                 for (j = 0; j < rx_pool->size; j++) {
 478                         if (rx_pool->rx_buff[j].skb) {
 479                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 480                                 rx_pool->rx_buff[j].skb = NULL;
 481                         }
 482                 }
 483
 484                 kfree(rx_pool->rx_buff);
 485         }
 486
 487         kfree(adapter->rx_pool);
 488         adapter->rx_pool = NULL;
 489         adapter->num_active_rx_pools = 0;
 490 }
 491
 492 static int init_rx_pools(struct net_device *netdev)
 493 {
 494         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 495         struct device *dev = &adapter->vdev->dev;
 496         struct ibmvnic_rx_pool *rx_pool;
 497         int rxadd_subcrqs;
 498         u64 *size_array;
 499         int i, j;
 500
 501         rxadd_subcrqs =
 502                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 503         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 504                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 505
 506         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 507                                    sizeof(struct ibmvnic_rx_pool),
 508                                    GFP_KERNEL);
 509         if (!adapter->rx_pool) {
 510                 dev_err(dev, "Failed to allocate rx pools\n");
 511                 return -1;
 512         }
 513
 514         adapter->num_active_rx_pools = rxadd_subcrqs;
 515
 516         for (i = 0; i < rxadd_subcrqs; i++) {
 517                 rx_pool = &adapter->rx_pool[i];
 518
 519                 netdev_dbg(adapter->netdev,
 520                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 521                            i, adapter->req_rx_add_entries_per_subcrq,
 522                            be64_to_cpu(size_array[i]));
 523
 524                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 525                 rx_pool->index = i;
 526                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 527                 rx_pool->active = 1;
 528
 529                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 530                                             GFP_KERNEL);
 531                 if (!rx_pool->free_map) {
 532                         release_rx_pools(adapter);
 533                         return -1;
 534                 }
 535
 536                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 537                                            sizeof(struct ibmvnic_rx_buff),
 538                                            GFP_KERNEL);
 539                 if (!rx_pool->rx_buff) {
 540                         dev_err(dev, "Couldn't alloc rx buffers\n");
 541                         release_rx_pools(adapter);
 542                         return -1;
 543                 }
 544
 545                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 546                                          rx_pool->size * rx_pool->buff_size)) {
 547                         release_rx_pools(adapter);
 548                         return -1;
 549                 }
 550
 551                 for (j = 0; j < rx_pool->size; ++j)
 552                         rx_pool->free_map[j] = j;
 553
 554                 atomic_set(&rx_pool->available, 0);
 555                 rx_pool->next_alloc = 0;
 556                 rx_pool->next_free = 0;
 557         }
 558
 559         return 0;
 560 }
 561
 562 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 563                              struct ibmvnic_tx_pool *tx_pool)
 564 {
 565         int rc, i;
 566
 567         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 568         if (rc)
 569                 return rc;
 570
 571         memset(tx_pool->tx_buff, 0,
 572                tx_pool->num_buffers *
 573                sizeof(struct ibmvnic_tx_buff));
 574
 575         for (i = 0; i < tx_pool->num_buffers; i++)
 576                 tx_pool->free_map[i] = i;
 577
 578         tx_pool->consumer_index = 0;
 579         tx_pool->producer_index = 0;
 580
 581         return 0;
 582 }
 583
 584 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 585 {
 586         int tx_scrqs;
 587         int i, rc;
 588
 589         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 590         for (i = 0; i < tx_scrqs; i++) {
 591                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 592                 if (rc)
 593                         return rc;
 594                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 595                 if (rc)
 596                         return rc;
 597         }
 598
 599         return 0;
 600 }
 601
 602 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 603 {
 604         if (!adapter->vpd)
 605                 return;
 606
 607         kfree(adapter->vpd->buff);
 608         kfree(adapter->vpd);
 609
 610         adapter->vpd = NULL;
 611 }
 612
 613 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 614                                 struct ibmvnic_tx_pool *tx_pool)
 615 {
 616         kfree(tx_pool->tx_buff);
 617         kfree(tx_pool->free_map);
 618         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 619 }
 620
 621 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 622 {
 623         int i;
 624
 625         if (!adapter->tx_pool)
 626                 return;
 627
 628         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 629                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 630                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 631         }
 632
 633         kfree(adapter->tx_pool);
 634         adapter->tx_pool = NULL;
 635         kfree(adapter->tso_pool);
 636         adapter->tso_pool = NULL;
 637         adapter->num_active_tx_pools = 0;
 638 }
 639
 640 static int init_one_tx_pool(struct net_device *netdev,
 641                             struct ibmvnic_tx_pool *tx_pool,
 642                             int num_entries, int buf_size)
 643 {
 644         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 645         int i;
 646
 647         tx_pool->tx_buff = kcalloc(num_entries,
 648                                    sizeof(struct ibmvnic_tx_buff),
 649                                    GFP_KERNEL);
 650         if (!tx_pool->tx_buff)
 651                 return -1;
 652
 653         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 654                                  num_entries * buf_size))
 655                 return -1;
 656
 657         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 658         if (!tx_pool->free_map)
 659                 return -1;
 660
 661         for (i = 0; i < num_entries; i++)
 662                 tx_pool->free_map[i] = i;
 663
 664         tx_pool->consumer_index = 0;
 665         tx_pool->producer_index = 0;
 666         tx_pool->num_buffers = num_entries;
 667         tx_pool->buf_size = buf_size;
 668
 669         return 0;
 670 }
 671
 672 static int init_tx_pools(struct net_device *netdev)
 673 {
 674         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 675         int tx_subcrqs;
 676         int i, rc;
 677
 678         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 679         adapter->tx_pool = kcalloc(tx_subcrqs,
 680                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 681         if (!adapter->tx_pool)
 682                 return -1;
 683
 684         adapter->tso_pool = kcalloc(tx_subcrqs,
 685                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 686         if (!adapter->tso_pool)
 687                 return -1;
 688
 689         adapter->num_active_tx_pools = tx_subcrqs;
 690
 691         for (i = 0; i < tx_subcrqs; i++) {
 692                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 693                                       adapter->req_tx_entries_per_subcrq,
 694                                       adapter->req_mtu + VLAN_HLEN);
 695                 if (rc) {
 696                         release_tx_pools(adapter);
 697                         return rc;
 698                 }
 699
 700                 rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
 701                                       IBMVNIC_TSO_BUFS,
 702                                       IBMVNIC_TSO_BUF_SZ);
 703                 if (rc) {
 704                         release_tx_pools(adapter);
 705                         return rc;
 706                 }
 707         }
 708
 709         return 0;
 710 }
 711
 712 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 713 {
 714         int i;
 715
 716         if (adapter->napi_enabled)
 717                 return;
 718
 719         for (i = 0; i < adapter->req_rx_queues; i++)
 720                 napi_enable(&adapter->napi[i]);
 721
 722         adapter->napi_enabled = true;
 723 }
 724
 725 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 726 {
 727         int i;
 728
 729         if (!adapter->napi_enabled)
 730                 return;
 731
 732         for (i = 0; i < adapter->req_rx_queues; i++) {
 733                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 734                 napi_disable(&adapter->napi[i]);
 735         }
 736
 737         adapter->napi_enabled = false;
 738 }
 739
 740 static int init_napi(struct ibmvnic_adapter *adapter)
 741 {
 742         int i;
 743
 744         adapter->napi = kcalloc(adapter->req_rx_queues,
 745                                 sizeof(struct napi_struct), GFP_KERNEL);
 746         if (!adapter->napi)
 747                 return -ENOMEM;
 748
 749         for (i = 0; i < adapter->req_rx_queues; i++) {
 750                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 751                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 752                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 753         }
 754
 755         adapter->num_active_rx_napi = adapter->req_rx_queues;
 756         return 0;
 757 }
 758
 759 static void release_napi(struct ibmvnic_adapter *adapter)
 760 {
 761         int i;
 762
 763         if (!adapter->napi)
 764                 return;
 765
 766         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 767                 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
 768                 netif_napi_del(&adapter->napi[i]);
 769         }
 770
 771         kfree(adapter->napi);
 772         adapter->napi = NULL;
 773         adapter->num_active_rx_napi = 0;
 774         adapter->napi_enabled = false;
 775 }
 776
 777 static int ibmvnic_login(struct net_device *netdev)
 778 {
 779         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 780         unsigned long timeout = msecs_to_jiffies(30000);
 781         int retry_count = 0;
 782         bool retry;
 783         int rc;
 784
 785         do {
 786                 retry = false;
 787                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 788                         netdev_warn(netdev, "Login attempts exceeded\n");
 789                         return -1;
 790                 }
 791
 792                 adapter->init_done_rc = 0;
 793                 reinit_completion(&adapter->init_done);
 794                 rc = send_login(adapter);
 795                 if (rc) {
 796                         netdev_warn(netdev, "Unable to login\n");
 797                         return rc;
 798                 }
 799
 800                 if (!wait_for_completion_timeout(&adapter->init_done,
 801                                                  timeout)) {
 802                         netdev_warn(netdev, "Login timed out\n");
 803                         return -1;
 804                 }
 805
 806                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 807                         retry_count++;
 808                         release_sub_crqs(adapter, 1);
 809
 810                         retry = true;
 811                         netdev_dbg(netdev,
 812                                    "Received partial success, retrying...\n");
 813                         adapter->init_done_rc = 0;
 814                         reinit_completion(&adapter->init_done);
 815                         send_cap_queries(adapter);
 816                         if (!wait_for_completion_timeout(&adapter->init_done,
 817                                                          timeout)) {
 818                                 netdev_warn(netdev,
 819                                             "Capabilities query timed out\n");
 820                                 return -1;
 821                         }
 822
 823                         rc = init_sub_crqs(adapter);
 824                         if (rc) {
 825                                 netdev_warn(netdev,
 826                                             "SCRQ initialization failed\n");
 827                                 return -1;
 828                         }
 829
 830                         rc = init_sub_crq_irqs(adapter);
 831                         if (rc) {
 832                                 netdev_warn(netdev,
 833                                             "SCRQ irq initialization failed\n");
 834                                 return -1;
 835                         }
 836                 } else if (adapter->init_done_rc) {
 837                         netdev_warn(netdev, "Adapter login failed\n");
 838                         return -1;
 839                 }
 840         } while (retry);
 841
 842         __ibmvnic_set_mac(netdev, adapter->mac_addr);
 843
 844         return 0;
 845 }
 846
 847 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 848 {
 849         kfree(adapter->login_buf);
 850         adapter->login_buf = NULL;
 851 }
 852
 853 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 854 {
 855         kfree(adapter->login_rsp_buf);
 856         adapter->login_rsp_buf = NULL;
 857 }
 858
 859 static void release_resources(struct ibmvnic_adapter *adapter)
 860 {
 861         release_vpd_data(adapter);
 862
 863         release_tx_pools(adapter);
 864         release_rx_pools(adapter);
 865
 866         release_napi(adapter);
 867         release_login_rsp_buffer(adapter);
 868 }
 869
 870 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 871 {
 872         struct net_device *netdev = adapter->netdev;
 873         unsigned long timeout = msecs_to_jiffies(30000);
 874         union ibmvnic_crq crq;
 875         bool resend;
 876         int rc;
 877
 878         netdev_dbg(netdev, "setting link state %d\n", link_state);
 879
 880         memset(&crq, 0, sizeof(crq));
 881         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 882         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 883         crq.logical_link_state.link_state = link_state;
 884
 885         do {
 886                 resend = false;
 887
 888                 reinit_completion(&adapter->init_done);
 889                 rc = ibmvnic_send_crq(adapter, &crq);
 890                 if (rc) {
 891                         netdev_err(netdev, "Failed to set link state\n");
 892                         return rc;
 893                 }
 894
 895                 if (!wait_for_completion_timeout(&adapter->init_done,
 896                                                  timeout)) {
 897                         netdev_err(netdev, "timeout setting link state\n");
 898                         return -1;
 899                 }
 900
 901                 if (adapter->init_done_rc == 1) {
 902                         /* Partuial success, delay and re-send */
 903                         mdelay(1000);
 904                         resend = true;
 905                 } else if (adapter->init_done_rc) {
 906                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 907                                     adapter->init_done_rc);
 908                         return adapter->init_done_rc;
 909                 }
 910         } while (resend);
 911
 912         return 0;
 913 }
 914
 915 static int set_real_num_queues(struct net_device *netdev)
 916 {
 917         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 918         int rc;
 919
 920         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 921                    adapter->req_tx_queues, adapter->req_rx_queues);
 922
 923         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 924         if (rc) {
 925                 netdev_err(netdev, "failed to set the number of tx queues\n");
 926                 return rc;
 927         }
 928
 929         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 930         if (rc)
 931                 netdev_err(netdev, "failed to set the number of rx queues\n");
 932
 933         return rc;
 934 }
 935
 936 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 937 {
 938         struct device *dev = &adapter->vdev->dev;
 939         union ibmvnic_crq crq;
 940         int len = 0;
 941         int rc;
 942
 943         if (adapter->vpd->buff)
 944                 len = adapter->vpd->len;
 945
 946         init_completion(&adapter->fw_done);
 947         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 948         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 949         rc = ibmvnic_send_crq(adapter, &crq);
 950         if (rc)
 951                 return rc;
 952         wait_for_completion(&adapter->fw_done);
 953
 954         if (!adapter->vpd->len)
 955                 return -ENODATA;
 956
 957         if (!adapter->vpd->buff)
 958                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 959         else if (adapter->vpd->len != len)
 960                 adapter->vpd->buff =
 961                         krealloc(adapter->vpd->buff,
 962                                  adapter->vpd->len, GFP_KERNEL);
 963
 964         if (!adapter->vpd->buff) {
 965                 dev_err(dev, "Could allocate VPD buffer\n");
 966                 return -ENOMEM;
 967         }
 968
 969         adapter->vpd->dma_addr =
 970                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 971                                DMA_FROM_DEVICE);
 972         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 973                 dev_err(dev, "Could not map VPD buffer\n");
 974                 kfree(adapter->vpd->buff);
 975                 adapter->vpd->buff = NULL;
 976                 return -ENOMEM;
 977         }
 978
 979         reinit_completion(&adapter->fw_done);
 980         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 981         crq.get_vpd.cmd = GET_VPD;
 982         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 983         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 984         rc = ibmvnic_send_crq(adapter, &crq);
 985         if (rc) {
 986                 kfree(adapter->vpd->buff);
 987                 adapter->vpd->buff = NULL;
 988                 return rc;
 989         }
 990         wait_for_completion(&adapter->fw_done);
 991
 992         return 0;
 993 }
 994
 995 static int init_resources(struct ibmvnic_adapter *adapter)
 996 {
 997         struct net_device *netdev = adapter->netdev;
 998         int rc;
 999
1000         rc = set_real_num_queues(netdev);
1001         if (rc)
1002                 return rc;
1003
1004         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1005         if (!adapter->vpd)
1006                 return -ENOMEM;
1007
1008         /* Vital Product Data (VPD) */
1009         rc = ibmvnic_get_vpd(adapter);
1010         if (rc) {
1011                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1012                 return rc;
1013         }
1014
1015         adapter->map_id = 1;
1016
1017         rc = init_napi(adapter);
1018         if (rc)
1019                 return rc;
1020
1021         send_map_query(adapter);
1022
1023         rc = init_rx_pools(netdev);
1024         if (rc)
1025                 return rc;
1026
1027         rc = init_tx_pools(netdev);
1028         return rc;
1029 }
1030
1031 static int __ibmvnic_open(struct net_device *netdev)
1032 {
1033         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1034         enum vnic_state prev_state = adapter->state;
1035         int i, rc;
1036
1037         adapter->state = VNIC_OPENING;
1038         replenish_pools(adapter);
1039         ibmvnic_napi_enable(adapter);
1040
1041         /* We're ready to receive frames, enable the sub-crq interrupts and
1042          * set the logical link state to up
1043          */
1044         for (i = 0; i < adapter->req_rx_queues; i++) {
1045                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1046                 if (prev_state == VNIC_CLOSED)
1047                         enable_irq(adapter->rx_scrq[i]->irq);
1048                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1049         }
1050
1051         for (i = 0; i < adapter->req_tx_queues; i++) {
1052                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1053                 if (prev_state == VNIC_CLOSED)
1054                         enable_irq(adapter->tx_scrq[i]->irq);
1055                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1056         }
1057
1058         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1059         if (rc) {
1060                 for (i = 0; i < adapter->req_rx_queues; i++)
1061                         napi_disable(&adapter->napi[i]);
1062                 release_resources(adapter);
1063                 return rc;
1064         }
1065
1066         netif_tx_start_all_queues(netdev);
1067
1068         if (prev_state == VNIC_CLOSED) {
1069                 for (i = 0; i < adapter->req_rx_queues; i++)
1070                         napi_schedule(&adapter->napi[i]);
1071         }
1072
1073         adapter->state = VNIC_OPEN;
1074         return rc;
1075 }
1076
1077 static int ibmvnic_open(struct net_device *netdev)
1078 {
1079         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1080         int rc;
1081
1082         /* If device failover is pending, just set device state and return.
1083          * Device operation will be handled by reset routine.
1084          */
1085         if (adapter->failover_pending) {
1086                 adapter->state = VNIC_OPEN;
1087                 return 0;
1088         }
1089
1090         if (adapter->state != VNIC_CLOSED) {
1091                 rc = ibmvnic_login(netdev);
1092                 if (rc)
1093                         return rc;
1094
1095                 rc = init_resources(adapter);
1096                 if (rc) {
1097                         netdev_err(netdev, "failed to initialize resources\n");
1098                         release_resources(adapter);
1099                         return rc;
1100                 }
1101         }
1102
1103         rc = __ibmvnic_open(netdev);
1104
1105         return rc;
1106 }
1107
1108 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1109 {
1110         struct ibmvnic_rx_pool *rx_pool;
1111         struct ibmvnic_rx_buff *rx_buff;
1112         u64 rx_entries;
1113         int rx_scrqs;
1114         int i, j;
1115
1116         if (!adapter->rx_pool)
1117                 return;
1118
1119         rx_scrqs = adapter->num_active_rx_pools;
1120         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1121
1122         /* Free any remaining skbs in the rx buffer pools */
1123         for (i = 0; i < rx_scrqs; i++) {
1124                 rx_pool = &adapter->rx_pool[i];
1125                 if (!rx_pool || !rx_pool->rx_buff)
1126                         continue;
1127
1128                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1129                 for (j = 0; j < rx_entries; j++) {
1130                         rx_buff = &rx_pool->rx_buff[j];
1131                         if (rx_buff && rx_buff->skb) {
1132                                 dev_kfree_skb_any(rx_buff->skb);
1133                                 rx_buff->skb = NULL;
1134                         }
1135                 }
1136         }
1137 }
1138
1139 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1140                               struct ibmvnic_tx_pool *tx_pool)
1141 {
1142         struct ibmvnic_tx_buff *tx_buff;
1143         u64 tx_entries;
1144         int i;
1145
1146         if (!tx_pool || !tx_pool->tx_buff)
1147                 return;
1148
1149         tx_entries = tx_pool->num_buffers;
1150
1151         for (i = 0; i < tx_entries; i++) {
1152                 tx_buff = &tx_pool->tx_buff[i];
1153                 if (tx_buff && tx_buff->skb) {
1154                         dev_kfree_skb_any(tx_buff->skb);
1155                         tx_buff->skb = NULL;
1156                 }
1157         }
1158 }
1159
1160 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1161 {
1162         int tx_scrqs;
1163         int i;
1164
1165         if (!adapter->tx_pool || !adapter->tso_pool)
1166                 return;
1167
1168         tx_scrqs = adapter->num_active_tx_pools;
1169
1170         /* Free any remaining skbs in the tx buffer pools */
1171         for (i = 0; i < tx_scrqs; i++) {
1172                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1173                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1174                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1175         }
1176 }
1177
1178 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1179 {
1180         struct net_device *netdev = adapter->netdev;
1181         int i;
1182
1183         if (adapter->tx_scrq) {
1184                 for (i = 0; i < adapter->req_tx_queues; i++)
1185                         if (adapter->tx_scrq[i]->irq) {
1186                                 netdev_dbg(netdev,
1187                                            "Disabling tx_scrq[%d] irq\n", i);
1188                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1189                                 disable_irq(adapter->tx_scrq[i]->irq);
1190                         }
1191         }
1192
1193         if (adapter->rx_scrq) {
1194                 for (i = 0; i < adapter->req_rx_queues; i++) {
1195                         if (adapter->rx_scrq[i]->irq) {
1196                                 netdev_dbg(netdev,
1197                                            "Disabling rx_scrq[%d] irq\n", i);
1198                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1199                                 disable_irq(adapter->rx_scrq[i]->irq);
1200                         }
1201                 }
1202         }
1203 }
1204
1205 static void ibmvnic_cleanup(struct net_device *netdev)
1206 {
1207         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1208
1209         /* ensure that transmissions are stopped if called by do_reset */
1210         if (adapter->resetting)
1211                 netif_tx_disable(netdev);
1212         else
1213                 netif_tx_stop_all_queues(netdev);
1214
1215         ibmvnic_napi_disable(adapter);
1216         ibmvnic_disable_irqs(adapter);
1217
1218         clean_rx_pools(adapter);
1219         clean_tx_pools(adapter);
1220 }
1221
1222 static int __ibmvnic_close(struct net_device *netdev)
1223 {
1224         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1225         int rc = 0;
1226
1227         adapter->state = VNIC_CLOSING;
1228         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1229         if (rc)
1230                 return rc;
1231         adapter->state = VNIC_CLOSED;
1232         return 0;
1233 }
1234
1235 static int ibmvnic_close(struct net_device *netdev)
1236 {
1237         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1238         int rc;
1239
1240         /* If device failover is pending, just set device state and return.
1241          * Device operation will be handled by reset routine.
1242          */
1243         if (adapter->failover_pending) {
1244                 adapter->state = VNIC_CLOSED;
1245                 return 0;
1246         }
1247
1248         rc = __ibmvnic_close(netdev);
1249         ibmvnic_cleanup(netdev);
1250
1251         return rc;
1252 }
1253
1254 /**
1255  * build_hdr_data - creates L2/L3/L4 header data buffer
1256  * @hdr_field - bitfield determining needed headers
1257  * @skb - socket buffer
1258  * @hdr_len - array of header lengths
1259  * @tot_len - total length of data
1260  *
1261  * Reads hdr_field to determine which headers are needed by firmware.
1262  * Builds a buffer containing these headers.  Saves individual header
1263  * lengths and total buffer length to be used to build descriptors.
1264  */
1265 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1266                           int *hdr_len, u8 *hdr_data)
1267 {
1268         int len = 0;
1269         u8 *hdr;
1270
1271         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1272                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1273         else
1274                 hdr_len[0] = sizeof(struct ethhdr);
1275
1276         if (skb->protocol == htons(ETH_P_IP)) {
1277                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1278                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1279                         hdr_len[2] = tcp_hdrlen(skb);
1280                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1281                         hdr_len[2] = sizeof(struct udphdr);
1282         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1283                 hdr_len[1] = sizeof(struct ipv6hdr);
1284                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1285                         hdr_len[2] = tcp_hdrlen(skb);
1286                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1287                         hdr_len[2] = sizeof(struct udphdr);
1288         } else if (skb->protocol == htons(ETH_P_ARP)) {
1289                 hdr_len[1] = arp_hdr_len(skb->dev);
1290                 hdr_len[2] = 0;
1291         }
1292
1293         memset(hdr_data, 0, 120);
1294         if ((hdr_field >> 6) & 1) {
1295                 hdr = skb_mac_header(skb);
1296                 memcpy(hdr_data, hdr, hdr_len[0]);
1297                 len += hdr_len[0];
1298         }
1299
1300         if ((hdr_field >> 5) & 1) {
1301                 hdr = skb_network_header(skb);
1302                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1303                 len += hdr_len[1];
1304         }
1305
1306         if ((hdr_field >> 4) & 1) {
1307                 hdr = skb_transport_header(skb);
1308                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1309                 len += hdr_len[2];
1310         }
1311         return len;
1312 }
1313
1314 /**
1315  * create_hdr_descs - create header and header extension descriptors
1316  * @hdr_field - bitfield determining needed headers
1317  * @data - buffer containing header data
1318  * @len - length of data buffer
1319  * @hdr_len - array of individual header lengths
1320  * @scrq_arr - descriptor array
1321  *
1322  * Creates header and, if needed, header extension descriptors and
1323  * places them in a descriptor array, scrq_arr
1324  */
1325
1326 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1327                             union sub_crq *scrq_arr)
1328 {
1329         union sub_crq hdr_desc;
1330         int tmp_len = len;
1331         int num_descs = 0;
1332         u8 *data, *cur;
1333         int tmp;
1334
1335         while (tmp_len > 0) {
1336                 cur = hdr_data + len - tmp_len;
1337
1338                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1339                 if (cur != hdr_data) {
1340                         data = hdr_desc.hdr_ext.data;
1341                         tmp = tmp_len > 29 ? 29 : tmp_len;
1342                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1343                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1344                         hdr_desc.hdr_ext.len = tmp;
1345                 } else {
1346                         data = hdr_desc.hdr.data;
1347                         tmp = tmp_len > 24 ? 24 : tmp_len;
1348                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1349                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1350                         hdr_desc.hdr.len = tmp;
1351                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1352                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1353                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1354                         hdr_desc.hdr.flag = hdr_field << 1;
1355                 }
1356                 memcpy(data, cur, tmp);
1357                 tmp_len -= tmp;
1358                 *scrq_arr = hdr_desc;
1359                 scrq_arr++;
1360                 num_descs++;
1361         }
1362
1363         return num_descs;
1364 }
1365
1366 /**
1367  * build_hdr_descs_arr - build a header descriptor array
1368  * @skb - socket buffer
1369  * @num_entries - number of descriptors to be sent
1370  * @subcrq - first TX descriptor
1371  * @hdr_field - bit field determining which headers will be sent
1372  *
1373  * This function will build a TX descriptor array with applicable
1374  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1375  */
1376
1377 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1378                                 int *num_entries, u8 hdr_field)
1379 {
1380         int hdr_len[3] = {0, 0, 0};
1381         int tot_len;
1382         u8 *hdr_data = txbuff->hdr_data;
1383
1384         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1385                                  txbuff->hdr_data);
1386         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1387                          txbuff->indir_arr + 1);
1388 }
1389
1390 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1391                                     struct net_device *netdev)
1392 {
1393         /* For some backing devices, mishandling of small packets
1394          * can result in a loss of connection or TX stall. Device
1395          * architects recommend that no packet should be smaller
1396          * than the minimum MTU value provided to the driver, so
1397          * pad any packets to that length
1398          */
1399         if (skb->len < netdev->min_mtu)
1400                 return skb_put_padto(skb, netdev->min_mtu);
1401
1402         return 0;
1403 }
1404
1405 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1406 {
1407         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1408         int queue_num = skb_get_queue_mapping(skb);
1409         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1410         struct device *dev = &adapter->vdev->dev;
1411         struct ibmvnic_tx_buff *tx_buff = NULL;
1412         struct ibmvnic_sub_crq_queue *tx_scrq;
1413         struct ibmvnic_tx_pool *tx_pool;
1414         unsigned int tx_send_failed = 0;
1415         unsigned int tx_map_failed = 0;
1416         unsigned int tx_dropped = 0;
1417         unsigned int tx_packets = 0;
1418         unsigned int tx_bytes = 0;
1419         dma_addr_t data_dma_addr;
1420         struct netdev_queue *txq;
1421         unsigned long lpar_rc;
1422         union sub_crq tx_crq;
1423         unsigned int offset;
1424         int num_entries = 1;
1425         unsigned char *dst;
1426         u64 *handle_array;
1427         int index = 0;
1428         u8 proto = 0;
1429         netdev_tx_t ret = NETDEV_TX_OK;
1430
1431         if (adapter->resetting) {
1432                 if (!netif_subqueue_stopped(netdev, skb))
1433                         netif_stop_subqueue(netdev, queue_num);
1434                 dev_kfree_skb_any(skb);
1435
1436                 tx_send_failed++;
1437                 tx_dropped++;
1438                 ret = NETDEV_TX_OK;
1439                 goto out;
1440         }
1441
1442         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1443                 tx_dropped++;
1444                 tx_send_failed++;
1445                 ret = NETDEV_TX_OK;
1446                 goto out;
1447         }
1448         if (skb_is_gso(skb))
1449                 tx_pool = &adapter->tso_pool[queue_num];
1450         else
1451                 tx_pool = &adapter->tx_pool[queue_num];
1452
1453         tx_scrq = adapter->tx_scrq[queue_num];
1454         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1455         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1456                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1457
1458         index = tx_pool->free_map[tx_pool->consumer_index];
1459
1460         if (index == IBMVNIC_INVALID_MAP) {
1461                 dev_kfree_skb_any(skb);
1462                 tx_send_failed++;
1463                 tx_dropped++;
1464                 ret = NETDEV_TX_OK;
1465                 goto out;
1466         }
1467
1468         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1469
1470         offset = index * tx_pool->buf_size;
1471         dst = tx_pool->long_term_buff.buff + offset;
1472         memset(dst, 0, tx_pool->buf_size);
1473         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1474
1475         if (skb_shinfo(skb)->nr_frags) {
1476                 int cur, i;
1477
1478                 /* Copy the head */
1479                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1480                 cur = skb_headlen(skb);
1481
1482                 /* Copy the frags */
1483                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1484                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1485
1486                         memcpy(dst + cur,
1487                                page_address(skb_frag_page(frag)) +
1488                                frag->page_offset, skb_frag_size(frag));
1489                         cur += skb_frag_size(frag);
1490                 }
1491         } else {
1492                 skb_copy_from_linear_data(skb, dst, skb->len);
1493         }
1494
1495         tx_pool->consumer_index =
1496             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1497
1498         tx_buff = &tx_pool->tx_buff[index];
1499         tx_buff->skb = skb;
1500         tx_buff->data_dma[0] = data_dma_addr;
1501         tx_buff->data_len[0] = skb->len;
1502         tx_buff->index = index;
1503         tx_buff->pool_index = queue_num;
1504         tx_buff->last_frag = true;
1505
1506         memset(&tx_crq, 0, sizeof(tx_crq));
1507         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1508         tx_crq.v1.type = IBMVNIC_TX_DESC;
1509         tx_crq.v1.n_crq_elem = 1;
1510         tx_crq.v1.n_sge = 1;
1511         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1512
1513         if (skb_is_gso(skb))
1514                 tx_crq.v1.correlator =
1515                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1516         else
1517                 tx_crq.v1.correlator = cpu_to_be32(index);
1518         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1519         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1520         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1521
1522         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1523                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1524                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1525         }
1526
1527         if (skb->protocol == htons(ETH_P_IP)) {
1528                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1529                 proto = ip_hdr(skb)->protocol;
1530         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1531                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1532                 proto = ipv6_hdr(skb)->nexthdr;
1533         }
1534
1535         if (proto == IPPROTO_TCP)
1536                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1537         else if (proto == IPPROTO_UDP)
1538                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1539
1540         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1541                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1542                 hdrs += 2;
1543         }
1544         if (skb_is_gso(skb)) {
1545                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1546                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1547                 hdrs += 2;
1548         }
1549         /* determine if l2/3/4 headers are sent to firmware */
1550         if ((*hdrs >> 7) & 1) {
1551                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1552                 tx_crq.v1.n_crq_elem = num_entries;
1553                 tx_buff->num_entries = num_entries;
1554                 tx_buff->indir_arr[0] = tx_crq;
1555                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1556                                                     sizeof(tx_buff->indir_arr),
1557                                                     DMA_TO_DEVICE);
1558                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1559                         dev_kfree_skb_any(skb);
1560                         tx_buff->skb = NULL;
1561                         if (!firmware_has_feature(FW_FEATURE_CMO))
1562                                 dev_err(dev, "tx: unable to map descriptor array\n");
1563                         tx_map_failed++;
1564                         tx_dropped++;
1565                         ret = NETDEV_TX_OK;
1566                         goto tx_err_out;
1567                 }
1568                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1569                                                (u64)tx_buff->indir_dma,
1570                                                (u64)num_entries);
1571                 dma_unmap_single(dev, tx_buff->indir_dma,
1572                                  sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1573         } else {
1574                 tx_buff->num_entries = num_entries;
1575                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1576                                       &tx_crq);
1577         }
1578         if (lpar_rc != H_SUCCESS) {
1579                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1580                         dev_err_ratelimited(dev, "tx: send failed\n");
1581                 dev_kfree_skb_any(skb);
1582                 tx_buff->skb = NULL;
1583
1584                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1585                         /* Disable TX and report carrier off if queue is closed
1586                          * or pending failover.
1587                          * Firmware guarantees that a signal will be sent to the
1588                          * driver, triggering a reset or some other action.
1589                          */
1590                         netif_tx_stop_all_queues(netdev);
1591                         netif_carrier_off(netdev);
1592                 }
1593
1594                 tx_send_failed++;
1595                 tx_dropped++;
1596                 ret = NETDEV_TX_OK;
1597                 goto tx_err_out;
1598         }
1599
1600         if (atomic_add_return(num_entries, &tx_scrq->used)
1601                                         >= adapter->req_tx_entries_per_subcrq) {
1602                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1603                 netif_stop_subqueue(netdev, queue_num);
1604         }
1605
1606         tx_packets++;
1607         tx_bytes += skb->len;
1608         txq->trans_start = jiffies;
1609         ret = NETDEV_TX_OK;
1610         goto out;
1611
1612 tx_err_out:
1613         /* roll back consumer index and map array*/
1614         if (tx_pool->consumer_index == 0)
1615                 tx_pool->consumer_index =
1616                         tx_pool->num_buffers - 1;
1617         else
1618                 tx_pool->consumer_index--;
1619         tx_pool->free_map[tx_pool->consumer_index] = index;
1620 out:
1621         netdev->stats.tx_dropped += tx_dropped;
1622         netdev->stats.tx_bytes += tx_bytes;
1623         netdev->stats.tx_packets += tx_packets;
1624         adapter->tx_send_failed += tx_send_failed;
1625         adapter->tx_map_failed += tx_map_failed;
1626         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1627         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1628         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1629
1630         return ret;
1631 }
1632
1633 static void ibmvnic_set_multi(struct net_device *netdev)
1634 {
1635         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1636         struct netdev_hw_addr *ha;
1637         union ibmvnic_crq crq;
1638
1639         memset(&crq, 0, sizeof(crq));
1640         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1641         crq.request_capability.cmd = REQUEST_CAPABILITY;
1642
1643         if (netdev->flags & IFF_PROMISC) {
1644                 if (!adapter->promisc_supported)
1645                         return;
1646         } else {
1647                 if (netdev->flags & IFF_ALLMULTI) {
1648                         /* Accept all multicast */
1649                         memset(&crq, 0, sizeof(crq));
1650                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1651                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1652                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1653                         ibmvnic_send_crq(adapter, &crq);
1654                 } else if (netdev_mc_empty(netdev)) {
1655                         /* Reject all multicast */
1656                         memset(&crq, 0, sizeof(crq));
1657                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1658                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1659                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1660                         ibmvnic_send_crq(adapter, &crq);
1661                 } else {
1662                         /* Accept one or more multicast(s) */
1663                         netdev_for_each_mc_addr(ha, netdev) {
1664                                 memset(&crq, 0, sizeof(crq));
1665                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1666                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1667                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1668                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1669                                                 ha->addr);
1670                                 ibmvnic_send_crq(adapter, &crq);
1671                         }
1672                 }
1673         }
1674 }
1675
1676 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1677 {
1678         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1679         union ibmvnic_crq crq;
1680         int rc;
1681
1682         if (!is_valid_ether_addr(dev_addr)) {
1683                 rc = -EADDRNOTAVAIL;
1684                 goto err;
1685         }
1686
1687         memset(&crq, 0, sizeof(crq));
1688         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1689         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1690         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1691
1692         init_completion(&adapter->fw_done);
1693         rc = ibmvnic_send_crq(adapter, &crq);
1694         if (rc) {
1695                 rc = -EIO;
1696                 goto err;
1697         }
1698
1699         wait_for_completion(&adapter->fw_done);
1700         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1701         if (adapter->fw_done_rc) {
1702                 rc = -EIO;
1703                 goto err;
1704         }
1705
1706         return 0;
1707 err:
1708         ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1709         return rc;
1710 }
1711
1712 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1713 {
1714         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1715         struct sockaddr *addr = p;
1716         int rc;
1717
1718         rc = 0;
1719         ether_addr_copy(adapter->mac_addr, addr->sa_data);
1720         if (adapter->state != VNIC_PROBED)
1721                 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1722
1723         return rc;
1724 }
1725
1726 /**
1727  * do_reset returns zero if we are able to keep processing reset events, or
1728  * non-zero if we hit a fatal error and must halt.
1729  */
1730 static int do_reset(struct ibmvnic_adapter *adapter,
1731                     struct ibmvnic_rwi *rwi, u32 reset_state)
1732 {
1733         u64 old_num_rx_queues, old_num_tx_queues;
1734         u64 old_num_rx_slots, old_num_tx_slots;
1735         struct net_device *netdev = adapter->netdev;
1736         int i, rc;
1737
1738         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1739                    rwi->reset_reason);
1740
1741         netif_carrier_off(netdev);
1742         adapter->reset_reason = rwi->reset_reason;
1743
1744         old_num_rx_queues = adapter->req_rx_queues;
1745         old_num_tx_queues = adapter->req_tx_queues;
1746         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1747         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1748
1749         ibmvnic_cleanup(netdev);
1750
1751         if (reset_state == VNIC_OPEN &&
1752             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1753             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1754                 rc = __ibmvnic_close(netdev);
1755                 if (rc)
1756                         return rc;
1757         }
1758
1759         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1760             adapter->wait_for_reset) {
1761                 release_resources(adapter);
1762                 release_sub_crqs(adapter, 1);
1763                 release_crq_queue(adapter);
1764         }
1765
1766         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1767                 /* remove the closed state so when we call open it appears
1768                  * we are coming from the probed state.
1769                  */
1770                 adapter->state = VNIC_PROBED;
1771
1772                 if (adapter->wait_for_reset) {
1773                         rc = init_crq_queue(adapter);
1774                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1775                         rc = ibmvnic_reenable_crq_queue(adapter);
1776                         release_sub_crqs(adapter, 1);
1777                 } else {
1778                         rc = ibmvnic_reset_crq(adapter);
1779                         if (!rc)
1780                                 rc = vio_enable_interrupts(adapter->vdev);
1781                 }
1782
1783                 if (rc) {
1784                         netdev_err(adapter->netdev,
1785                                    "Couldn't initialize crq. rc=%d\n", rc);
1786                         return rc;
1787                 }
1788
1789                 rc = ibmvnic_reset_init(adapter);
1790                 if (rc)
1791                         return IBMVNIC_INIT_FAILED;
1792
1793                 /* If the adapter was in PROBE state prior to the reset,
1794                  * exit here.
1795                  */
1796                 if (reset_state == VNIC_PROBED)
1797                         return 0;
1798
1799                 rc = ibmvnic_login(netdev);
1800                 if (rc) {
1801                         adapter->state = reset_state;
1802                         return rc;
1803                 }
1804
1805                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1806                     adapter->wait_for_reset) {
1807                         rc = init_resources(adapter);
1808                         if (rc)
1809                                 return rc;
1810                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1811                            adapter->req_tx_queues != old_num_tx_queues ||
1812                            adapter->req_rx_add_entries_per_subcrq !=
1813                                                         old_num_rx_slots ||
1814                            adapter->req_tx_entries_per_subcrq !=
1815                                                         old_num_tx_slots) {
1816                         release_rx_pools(adapter);
1817                         release_tx_pools(adapter);
1818                         release_napi(adapter);
1819                         release_vpd_data(adapter);
1820
1821                         rc = init_resources(adapter);
1822                         if (rc)
1823                                 return rc;
1824
1825                 } else {
1826                         rc = reset_tx_pools(adapter);
1827                         if (rc)
1828                                 return rc;
1829
1830                         rc = reset_rx_pools(adapter);
1831                         if (rc)
1832                                 return rc;
1833                 }
1834                 ibmvnic_disable_irqs(adapter);
1835         }
1836         adapter->state = VNIC_CLOSED;
1837
1838         if (reset_state == VNIC_CLOSED)
1839                 return 0;
1840
1841         rc = __ibmvnic_open(netdev);
1842         if (rc) {
1843                 if (list_empty(&adapter->rwi_list))
1844                         adapter->state = VNIC_CLOSED;
1845                 else
1846                         adapter->state = reset_state;
1847
1848                 return 0;
1849         }
1850
1851         /* refresh device's multicast list */
1852         ibmvnic_set_multi(netdev);
1853
1854         /* kick napi */
1855         for (i = 0; i < adapter->req_rx_queues; i++)
1856                 napi_schedule(&adapter->napi[i]);
1857
1858         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1859             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1860                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1861
1862         return 0;
1863 }
1864
1865 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1866                          struct ibmvnic_rwi *rwi, u32 reset_state)
1867 {
1868         struct net_device *netdev = adapter->netdev;
1869         int rc;
1870
1871         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1872                    rwi->reset_reason);
1873
1874         netif_carrier_off(netdev);
1875         adapter->reset_reason = rwi->reset_reason;
1876
1877         ibmvnic_cleanup(netdev);
1878         release_resources(adapter);
1879         release_sub_crqs(adapter, 0);
1880         release_crq_queue(adapter);
1881
1882         /* remove the closed state so when we call open it appears
1883          * we are coming from the probed state.
1884          */
1885         adapter->state = VNIC_PROBED;
1886
1887         reinit_completion(&adapter->init_done);
1888         rc = init_crq_queue(adapter);
1889         if (rc) {
1890                 netdev_err(adapter->netdev,
1891                            "Couldn't initialize crq. rc=%d\n", rc);
1892                 return rc;
1893         }
1894
1895         rc = ibmvnic_init(adapter);
1896         if (rc)
1897                 return rc;
1898
1899         /* If the adapter was in PROBE state prior to the reset,
1900          * exit here.
1901          */
1902         if (reset_state == VNIC_PROBED)
1903                 return 0;
1904
1905         rc = ibmvnic_login(netdev);
1906         if (rc) {
1907                 adapter->state = VNIC_PROBED;
1908                 return 0;
1909         }
1910
1911         rc = init_resources(adapter);
1912         if (rc)
1913                 return rc;
1914
1915         ibmvnic_disable_irqs(adapter);
1916         adapter->state = VNIC_CLOSED;
1917
1918         if (reset_state == VNIC_CLOSED)
1919                 return 0;
1920
1921         rc = __ibmvnic_open(netdev);
1922         if (rc) {
1923                 if (list_empty(&adapter->rwi_list))
1924                         adapter->state = VNIC_CLOSED;
1925                 else
1926                         adapter->state = reset_state;
1927
1928                 return 0;
1929         }
1930
1931         return 0;
1932 }
1933
1934 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1935 {
1936         struct ibmvnic_rwi *rwi;
1937         unsigned long flags;
1938
1939         spin_lock_irqsave(&adapter->rwi_lock, flags);
1940
1941         if (!list_empty(&adapter->rwi_list)) {
1942                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1943                                        list);
1944                 list_del(&rwi->list);
1945         } else {
1946                 rwi = NULL;
1947         }
1948
1949         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1950         return rwi;
1951 }
1952
1953 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1954 {
1955         struct ibmvnic_rwi *rwi;
1956
1957         rwi = get_next_rwi(adapter);
1958         while (rwi) {
1959                 kfree(rwi);
1960                 rwi = get_next_rwi(adapter);
1961         }
1962 }
1963
1964 static void __ibmvnic_reset(struct work_struct *work)
1965 {
1966         struct ibmvnic_rwi *rwi;
1967         struct ibmvnic_adapter *adapter;
1968         bool we_lock_rtnl = false;
1969         u32 reset_state;
1970         int rc = 0;
1971
1972         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1973
1974         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1975          * unless wait_for_reset is set, in which case the rtnl lock
1976          * has already been taken before initializing the reset
1977          */
1978         if (!adapter->wait_for_reset) {
1979                 rtnl_lock();
1980                 we_lock_rtnl = true;
1981         }
1982         reset_state = adapter->state;
1983
1984         rwi = get_next_rwi(adapter);
1985         while (rwi) {
1986                 if (adapter->force_reset_recovery) {
1987                         adapter->force_reset_recovery = false;
1988                         rc = do_hard_reset(adapter, rwi, reset_state);
1989                 } else {
1990                         rc = do_reset(adapter, rwi, reset_state);
1991                 }
1992                 kfree(rwi);
1993                 if (rc && rc != IBMVNIC_INIT_FAILED &&
1994                     !adapter->force_reset_recovery)
1995                         break;
1996
1997                 rwi = get_next_rwi(adapter);
1998         }
1999
2000         if (adapter->wait_for_reset) {
2001                 adapter->wait_for_reset = false;
2002                 adapter->reset_done_rc = rc;
2003                 complete(&adapter->reset_done);
2004         }
2005
2006         if (rc) {
2007                 netdev_dbg(adapter->netdev, "Reset failed\n");
2008                 free_all_rwi(adapter);
2009         }
2010
2011         adapter->resetting = false;
2012         if (we_lock_rtnl)
2013                 rtnl_unlock();
2014 }
2015
2016 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2017                          enum ibmvnic_reset_reason reason)
2018 {
2019         struct list_head *entry, *tmp_entry;
2020         struct ibmvnic_rwi *rwi, *tmp;
2021         struct net_device *netdev = adapter->netdev;
2022         unsigned long flags;
2023         int ret;
2024
2025         if (adapter->state == VNIC_REMOVING ||
2026             adapter->state == VNIC_REMOVED ||
2027             adapter->failover_pending) {
2028                 ret = EBUSY;
2029                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2030                 goto err;
2031         }
2032
2033         if (adapter->state == VNIC_PROBING) {
2034                 netdev_warn(netdev, "Adapter reset during probe\n");
2035                 ret = adapter->init_done_rc = EAGAIN;
2036                 goto err;
2037         }
2038
2039         spin_lock_irqsave(&adapter->rwi_lock, flags);
2040
2041         list_for_each(entry, &adapter->rwi_list) {
2042                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2043                 if (tmp->reset_reason == reason) {
2044                         netdev_dbg(netdev, "Skipping matching reset\n");
2045                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2046                         ret = EBUSY;
2047                         goto err;
2048                 }
2049         }
2050
2051         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2052         if (!rwi) {
2053                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2054                 ibmvnic_close(netdev);
2055                 ret = ENOMEM;
2056                 goto err;
2057         }
2058         /* if we just received a transport event,
2059          * flush reset queue and process this reset
2060          */
2061         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2062                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2063                         list_del(entry);
2064         }
2065         rwi->reset_reason = reason;
2066         list_add_tail(&rwi->list, &adapter->rwi_list);
2067         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2068         adapter->resetting = true;
2069         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2070         schedule_work(&adapter->ibmvnic_reset);
2071
2072         return 0;
2073 err:
2074         if (adapter->wait_for_reset)
2075                 adapter->wait_for_reset = false;
2076         return -ret;
2077 }
2078
2079 static void ibmvnic_tx_timeout(struct net_device *dev)
2080 {
2081         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2082
2083         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2084 }
2085
2086 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2087                                   struct ibmvnic_rx_buff *rx_buff)
2088 {
2089         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2090
2091         rx_buff->skb = NULL;
2092
2093         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2094         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2095
2096         atomic_dec(&pool->available);
2097 }
2098
2099 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2100 {
2101         struct net_device *netdev = napi->dev;
2102         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2103         int scrq_num = (int)(napi - adapter->napi);
2104         int frames_processed = 0;
2105
2106 restart_poll:
2107         while (frames_processed < budget) {
2108                 struct sk_buff *skb;
2109                 struct ibmvnic_rx_buff *rx_buff;
2110                 union sub_crq *next;
2111                 u32 length;
2112                 u16 offset;
2113                 u8 flags = 0;
2114
2115                 if (unlikely(adapter->resetting &&
2116                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2117                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2118                         napi_complete_done(napi, frames_processed);
2119                         return frames_processed;
2120                 }
2121
2122                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2123                         break;
2124                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2125                 rx_buff =
2126                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2127                                                           rx_comp.correlator);
2128                 /* do error checking */
2129                 if (next->rx_comp.rc) {
2130                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2131                                    be16_to_cpu(next->rx_comp.rc));
2132                         /* free the entry */
2133                         next->rx_comp.first = 0;
2134                         dev_kfree_skb_any(rx_buff->skb);
2135                         remove_buff_from_pool(adapter, rx_buff);
2136                         continue;
2137                 } else if (!rx_buff->skb) {
2138                         /* free the entry */
2139                         next->rx_comp.first = 0;
2140                         remove_buff_from_pool(adapter, rx_buff);
2141                         continue;
2142                 }
2143
2144                 length = be32_to_cpu(next->rx_comp.len);
2145                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2146                 flags = next->rx_comp.flags;
2147                 skb = rx_buff->skb;
2148                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2149                                         length);
2150
2151                 /* VLAN Header has been stripped by the system firmware and
2152                  * needs to be inserted by the driver
2153                  */
2154                 if (adapter->rx_vlan_header_insertion &&
2155                     (flags & IBMVNIC_VLAN_STRIPPED))
2156                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2157                                                ntohs(next->rx_comp.vlan_tci));
2158
2159                 /* free the entry */
2160                 next->rx_comp.first = 0;
2161                 remove_buff_from_pool(adapter, rx_buff);
2162
2163                 skb_put(skb, length);
2164                 skb->protocol = eth_type_trans(skb, netdev);
2165                 skb_record_rx_queue(skb, scrq_num);
2166
2167                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2168                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2169                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2170                 }
2171
2172                 length = skb->len;
2173                 napi_gro_receive(napi, skb); /* send it up */
2174                 netdev->stats.rx_packets++;
2175                 netdev->stats.rx_bytes += length;
2176                 adapter->rx_stats_buffers[scrq_num].packets++;
2177                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2178                 frames_processed++;
2179         }
2180
2181         if (adapter->state != VNIC_CLOSING)
2182                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2183
2184         if (frames_processed < budget) {
2185                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2186                 napi_complete_done(napi, frames_processed);
2187                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2188                     napi_reschedule(napi)) {
2189                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2190                         goto restart_poll;
2191                 }
2192         }
2193         return frames_processed;
2194 }
2195
2196 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2197 {
2198         int rc, ret;
2199
2200         adapter->fallback.mtu = adapter->req_mtu;
2201         adapter->fallback.rx_queues = adapter->req_rx_queues;
2202         adapter->fallback.tx_queues = adapter->req_tx_queues;
2203         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2204         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2205
2206         init_completion(&adapter->reset_done);
2207         adapter->wait_for_reset = true;
2208         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2209         if (rc)
2210                 return rc;
2211         wait_for_completion(&adapter->reset_done);
2212
2213         ret = 0;
2214         if (adapter->reset_done_rc) {
2215                 ret = -EIO;
2216                 adapter->desired.mtu = adapter->fallback.mtu;
2217                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2218                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2219                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2220                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2221
2222                 init_completion(&adapter->reset_done);
2223                 adapter->wait_for_reset = true;
2224                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2225                 if (rc)
2226                         return ret;
2227                 wait_for_completion(&adapter->reset_done);
2228         }
2229         adapter->wait_for_reset = false;
2230
2231         return ret;
2232 }
2233
2234 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2235 {
2236         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2237
2238         adapter->desired.mtu = new_mtu + ETH_HLEN;
2239
2240         return wait_for_reset(adapter);
2241 }
2242
2243 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2244                                                 struct net_device *dev,
2245                                                 netdev_features_t features)
2246 {
2247         /* Some backing hardware adapters can not
2248          * handle packets with a MSS less than 224
2249          * or with only one segment.
2250          */
2251         if (skb_is_gso(skb)) {
2252                 if (skb_shinfo(skb)->gso_size < 224 ||
2253                     skb_shinfo(skb)->gso_segs == 1)
2254                         features &= ~NETIF_F_GSO_MASK;
2255         }
2256
2257         return features;
2258 }
2259
2260 static const struct net_device_ops ibmvnic_netdev_ops = {
2261         .ndo_open               = ibmvnic_open,
2262         .ndo_stop               = ibmvnic_close,
2263         .ndo_start_xmit         = ibmvnic_xmit,
2264         .ndo_set_rx_mode        = ibmvnic_set_multi,
2265         .ndo_set_mac_address    = ibmvnic_set_mac,
2266         .ndo_validate_addr      = eth_validate_addr,
2267         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2268         .ndo_change_mtu         = ibmvnic_change_mtu,
2269         .ndo_features_check     = ibmvnic_features_check,
2270 };
2271
2272 /* ethtool functions */
2273
2274 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2275                                       struct ethtool_link_ksettings *cmd)
2276 {
2277         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2278         int rc;
2279
2280         rc = send_query_phys_parms(adapter);
2281         if (rc) {
2282                 adapter->speed = SPEED_UNKNOWN;
2283                 adapter->duplex = DUPLEX_UNKNOWN;
2284         }
2285         cmd->base.speed = adapter->speed;
2286         cmd->base.duplex = adapter->duplex;
2287         cmd->base.port = PORT_FIBRE;
2288         cmd->base.phy_address = 0;
2289         cmd->base.autoneg = AUTONEG_ENABLE;
2290
2291         return 0;
2292 }
2293
2294 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2295                                 struct ethtool_drvinfo *info)
2296 {
2297         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2298
2299         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2300         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2301         strlcpy(info->fw_version, adapter->fw_version,
2302                 sizeof(info->fw_version));
2303 }
2304
2305 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2306 {
2307         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2308
2309         return adapter->msg_enable;
2310 }
2311
2312 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2313 {
2314         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2315
2316         adapter->msg_enable = data;
2317 }
2318
2319 static u32 ibmvnic_get_link(struct net_device *netdev)
2320 {
2321         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2322
2323         /* Don't need to send a query because we request a logical link up at
2324          * init and then we wait for link state indications
2325          */
2326         return adapter->logical_link_state;
2327 }
2328
2329 static void ibmvnic_get_ringparam(struct net_device *netdev,
2330                                   struct ethtool_ringparam *ring)
2331 {
2332         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2333
2334         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2335                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2336                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2337         } else {
2338                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2339                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2340         }
2341         ring->rx_mini_max_pending = 0;
2342         ring->rx_jumbo_max_pending = 0;
2343         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2344         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2345         ring->rx_mini_pending = 0;
2346         ring->rx_jumbo_pending = 0;
2347 }
2348
2349 static int ibmvnic_set_ringparam(struct net_device *netdev,
2350                                  struct ethtool_ringparam *ring)
2351 {
2352         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2353         int ret;
2354
2355         ret = 0;
2356         adapter->desired.rx_entries = ring->rx_pending;
2357         adapter->desired.tx_entries = ring->tx_pending;
2358
2359         ret = wait_for_reset(adapter);
2360
2361         if (!ret &&
2362             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2363              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2364                 netdev_info(netdev,
2365                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2366                             ring->rx_pending, ring->tx_pending,
2367                             adapter->req_rx_add_entries_per_subcrq,
2368                             adapter->req_tx_entries_per_subcrq);
2369         return ret;
2370 }
2371
2372 static void ibmvnic_get_channels(struct net_device *netdev,
2373                                  struct ethtool_channels *channels)
2374 {
2375         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2376
2377         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2378                 channels->max_rx = adapter->max_rx_queues;
2379                 channels->max_tx = adapter->max_tx_queues;
2380         } else {
2381                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2382                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2383         }
2384
2385         channels->max_other = 0;
2386         channels->max_combined = 0;
2387         channels->rx_count = adapter->req_rx_queues;
2388         channels->tx_count = adapter->req_tx_queues;
2389         channels->other_count = 0;
2390         channels->combined_count = 0;
2391 }
2392
2393 static int ibmvnic_set_channels(struct net_device *netdev,
2394                                 struct ethtool_channels *channels)
2395 {
2396         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2397         int ret;
2398
2399         ret = 0;
2400         adapter->desired.rx_queues = channels->rx_count;
2401         adapter->desired.tx_queues = channels->tx_count;
2402
2403         ret = wait_for_reset(adapter);
2404
2405         if (!ret &&
2406             (adapter->req_rx_queues != channels->rx_count ||
2407              adapter->req_tx_queues != channels->tx_count))
2408                 netdev_info(netdev,
2409                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2410                             channels->rx_count, channels->tx_count,
2411                             adapter->req_rx_queues, adapter->req_tx_queues);
2412         return ret;
2413
2414 }
2415
2416 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2417 {
2418         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2419         int i;
2420
2421         switch (stringset) {
2422         case ETH_SS_STATS:
2423                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2424                                 i++, data += ETH_GSTRING_LEN)
2425                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2426
2427                 for (i = 0; i < adapter->req_tx_queues; i++) {
2428                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2429                         data += ETH_GSTRING_LEN;
2430
2431                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2432                         data += ETH_GSTRING_LEN;
2433
2434                         snprintf(data, ETH_GSTRING_LEN,
2435                                  "tx%d_dropped_packets", i);
2436                         data += ETH_GSTRING_LEN;
2437                 }
2438
2439                 for (i = 0; i < adapter->req_rx_queues; i++) {
2440                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2441                         data += ETH_GSTRING_LEN;
2442
2443                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2444                         data += ETH_GSTRING_LEN;
2445
2446                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2447                         data += ETH_GSTRING_LEN;
2448                 }
2449                 break;
2450
2451         case ETH_SS_PRIV_FLAGS:
2452                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2453                         strcpy(data + i * ETH_GSTRING_LEN,
2454                                ibmvnic_priv_flags[i]);
2455                 break;
2456         default:
2457                 return;
2458         }
2459 }
2460
2461 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2462 {
2463         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2464
2465         switch (sset) {
2466         case ETH_SS_STATS:
2467                 return ARRAY_SIZE(ibmvnic_stats) +
2468                        adapter->req_tx_queues * NUM_TX_STATS +
2469                        adapter->req_rx_queues * NUM_RX_STATS;
2470         case ETH_SS_PRIV_FLAGS:
2471                 return ARRAY_SIZE(ibmvnic_priv_flags);
2472         default:
2473                 return -EOPNOTSUPP;
2474         }
2475 }
2476
2477 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2478                                       struct ethtool_stats *stats, u64 *data)
2479 {
2480         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2481         union ibmvnic_crq crq;
2482         int i, j;
2483         int rc;
2484
2485         memset(&crq, 0, sizeof(crq));
2486         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2487         crq.request_statistics.cmd = REQUEST_STATISTICS;
2488         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2489         crq.request_statistics.len =
2490             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2491
2492         /* Wait for data to be written */
2493         init_completion(&adapter->stats_done);
2494         rc = ibmvnic_send_crq(adapter, &crq);
2495         if (rc)
2496                 return;
2497         wait_for_completion(&adapter->stats_done);
2498
2499         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2500                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2501                                                 ibmvnic_stats[i].offset));
2502
2503         for (j = 0; j < adapter->req_tx_queues; j++) {
2504                 data[i] = adapter->tx_stats_buffers[j].packets;
2505                 i++;
2506                 data[i] = adapter->tx_stats_buffers[j].bytes;
2507                 i++;
2508                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2509                 i++;
2510         }
2511
2512         for (j = 0; j < adapter->req_rx_queues; j++) {
2513                 data[i] = adapter->rx_stats_buffers[j].packets;
2514                 i++;
2515                 data[i] = adapter->rx_stats_buffers[j].bytes;
2516                 i++;
2517                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2518                 i++;
2519         }
2520 }
2521
2522 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2523 {
2524         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2525
2526         return adapter->priv_flags;
2527 }
2528
2529 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2530 {
2531         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2532         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2533
2534         if (which_maxes)
2535                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2536         else
2537                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2538
2539         return 0;
2540 }
2541 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2542         .get_drvinfo            = ibmvnic_get_drvinfo,
2543         .get_msglevel           = ibmvnic_get_msglevel,
2544         .set_msglevel           = ibmvnic_set_msglevel,
2545         .get_link               = ibmvnic_get_link,
2546         .get_ringparam          = ibmvnic_get_ringparam,
2547         .set_ringparam          = ibmvnic_set_ringparam,
2548         .get_channels           = ibmvnic_get_channels,
2549         .set_channels           = ibmvnic_set_channels,
2550         .get_strings            = ibmvnic_get_strings,
2551         .get_sset_count         = ibmvnic_get_sset_count,
2552         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2553         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2554         .get_priv_flags         = ibmvnic_get_priv_flags,
2555         .set_priv_flags         = ibmvnic_set_priv_flags,
2556 };
2557
2558 /* Routines for managing CRQs/sCRQs  */
2559
2560 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2561                                    struct ibmvnic_sub_crq_queue *scrq)
2562 {
2563         int rc;
2564
2565         if (scrq->irq) {
2566                 free_irq(scrq->irq, scrq);
2567                 irq_dispose_mapping(scrq->irq);
2568                 scrq->irq = 0;
2569         }
2570
2571         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2572         atomic_set(&scrq->used, 0);
2573         scrq->cur = 0;
2574
2575         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2576                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2577         return rc;
2578 }
2579
2580 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2581 {
2582         int i, rc;
2583
2584         for (i = 0; i < adapter->req_tx_queues; i++) {
2585                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2586                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2587                 if (rc)
2588                         return rc;
2589         }
2590
2591         for (i = 0; i < adapter->req_rx_queues; i++) {
2592                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2593                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2594                 if (rc)
2595                         return rc;
2596         }
2597
2598         return rc;
2599 }
2600
2601 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2602                                   struct ibmvnic_sub_crq_queue *scrq,
2603                                   bool do_h_free)
2604 {
2605         struct device *dev = &adapter->vdev->dev;
2606         long rc;
2607
2608         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2609
2610         if (do_h_free) {
2611                 /* Close the sub-crqs */
2612                 do {
2613                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2614                                                 adapter->vdev->unit_address,
2615                                                 scrq->crq_num);
2616                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2617
2618                 if (rc) {
2619                         netdev_err(adapter->netdev,
2620                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2621                                    scrq->crq_num, rc);
2622                 }
2623         }
2624
2625         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2626                          DMA_BIDIRECTIONAL);
2627         free_pages((unsigned long)scrq->msgs, 2);
2628         kfree(scrq);
2629 }
2630
2631 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2632                                                         *adapter)
2633 {
2634         struct device *dev = &adapter->vdev->dev;
2635         struct ibmvnic_sub_crq_queue *scrq;
2636         int rc;
2637
2638         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2639         if (!scrq)
2640                 return NULL;
2641
2642         scrq->msgs =
2643                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2644         if (!scrq->msgs) {
2645                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2646                 goto zero_page_failed;
2647         }
2648
2649         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2650                                          DMA_BIDIRECTIONAL);
2651         if (dma_mapping_error(dev, scrq->msg_token)) {
2652                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2653                 goto map_failed;
2654         }
2655
2656         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2657                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2658
2659         if (rc == H_RESOURCE)
2660                 rc = ibmvnic_reset_crq(adapter);
2661
2662         if (rc == H_CLOSED) {
2663                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2664         } else if (rc) {
2665                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2666                 goto reg_failed;
2667         }
2668
2669         scrq->adapter = adapter;
2670         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2671         spin_lock_init(&scrq->lock);
2672
2673         netdev_dbg(adapter->netdev,
2674                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2675                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2676
2677         return scrq;
2678
2679 reg_failed:
2680         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2681                          DMA_BIDIRECTIONAL);
2682 map_failed:
2683         free_pages((unsigned long)scrq->msgs, 2);
2684 zero_page_failed:
2685         kfree(scrq);
2686
2687         return NULL;
2688 }
2689
2690 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2691 {
2692         int i;
2693
2694         if (adapter->tx_scrq) {
2695                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2696                         if (!adapter->tx_scrq[i])
2697                                 continue;
2698
2699                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2700                                    i);
2701                         if (adapter->tx_scrq[i]->irq) {
2702                                 free_irq(adapter->tx_scrq[i]->irq,
2703                                          adapter->tx_scrq[i]);
2704                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2705                                 adapter->tx_scrq[i]->irq = 0;
2706                         }
2707
2708                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2709                                               do_h_free);
2710                 }
2711
2712                 kfree(adapter->tx_scrq);
2713                 adapter->tx_scrq = NULL;
2714                 adapter->num_active_tx_scrqs = 0;
2715         }
2716
2717         if (adapter->rx_scrq) {
2718                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2719                         if (!adapter->rx_scrq[i])
2720                                 continue;
2721
2722                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2723                                    i);
2724                         if (adapter->rx_scrq[i]->irq) {
2725                                 free_irq(adapter->rx_scrq[i]->irq,
2726                                          adapter->rx_scrq[i]);
2727                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2728                                 adapter->rx_scrq[i]->irq = 0;
2729                         }
2730
2731                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2732                                               do_h_free);
2733                 }
2734
2735                 kfree(adapter->rx_scrq);
2736                 adapter->rx_scrq = NULL;
2737                 adapter->num_active_rx_scrqs = 0;
2738         }
2739 }
2740
2741 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2742                             struct ibmvnic_sub_crq_queue *scrq)
2743 {
2744         struct device *dev = &adapter->vdev->dev;
2745         unsigned long rc;
2746
2747         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2748                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2749         if (rc)
2750                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2751                         scrq->hw_irq, rc);
2752         return rc;
2753 }
2754
2755 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2756                            struct ibmvnic_sub_crq_queue *scrq)
2757 {
2758         struct device *dev = &adapter->vdev->dev;
2759         unsigned long rc;
2760
2761         if (scrq->hw_irq > 0x100000000ULL) {
2762                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2763                 return 1;
2764         }
2765
2766         if (adapter->resetting &&
2767             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2768                 u64 val = (0xff000000) | scrq->hw_irq;
2769
2770                 rc = plpar_hcall_norets(H_EOI, val);
2771                 if (rc)
2772                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2773                                 val, rc);
2774         }
2775
2776         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2777                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2778         if (rc)
2779                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2780                         scrq->hw_irq, rc);
2781         return rc;
2782 }
2783
2784 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2785                                struct ibmvnic_sub_crq_queue *scrq)
2786 {
2787         struct device *dev = &adapter->vdev->dev;
2788         struct ibmvnic_tx_pool *tx_pool;
2789         struct ibmvnic_tx_buff *txbuff;
2790         union sub_crq *next;
2791         int index;
2792         int i, j;
2793
2794 restart_loop:
2795         while (pending_scrq(adapter, scrq)) {
2796                 unsigned int pool = scrq->pool_index;
2797                 int num_entries = 0;
2798
2799                 next = ibmvnic_next_scrq(adapter, scrq);
2800                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2801                         if (next->tx_comp.rcs[i]) {
2802                                 dev_err(dev, "tx error %x\n",
2803                                         next->tx_comp.rcs[i]);
2804                                 continue;
2805                         }
2806                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2807                         if (index & IBMVNIC_TSO_POOL_MASK) {
2808                                 tx_pool = &adapter->tso_pool[pool];
2809                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2810                         } else {
2811                                 tx_pool = &adapter->tx_pool[pool];
2812                         }
2813
2814                         txbuff = &tx_pool->tx_buff[index];
2815
2816                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2817                                 if (!txbuff->data_dma[j])
2818                                         continue;
2819
2820                                 txbuff->data_dma[j] = 0;
2821                         }
2822
2823                         if (txbuff->last_frag) {
2824                                 dev_kfree_skb_any(txbuff->skb);
2825                                 txbuff->skb = NULL;
2826                         }
2827
2828                         num_entries += txbuff->num_entries;
2829
2830                         tx_pool->free_map[tx_pool->producer_index] = index;
2831                         tx_pool->producer_index =
2832                                 (tx_pool->producer_index + 1) %
2833                                         tx_pool->num_buffers;
2834                 }
2835                 /* remove tx_comp scrq*/
2836                 next->tx_comp.first = 0;
2837
2838                 if (atomic_sub_return(num_entries, &scrq->used) <=
2839                     (adapter->req_tx_entries_per_subcrq / 2) &&
2840                     __netif_subqueue_stopped(adapter->netdev,
2841                                              scrq->pool_index)) {
2842                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2843                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2844                                    scrq->pool_index);
2845                 }
2846         }
2847
2848         enable_scrq_irq(adapter, scrq);
2849
2850         if (pending_scrq(adapter, scrq)) {
2851                 disable_scrq_irq(adapter, scrq);
2852                 goto restart_loop;
2853         }
2854
2855         return 0;
2856 }
2857
2858 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2859 {
2860         struct ibmvnic_sub_crq_queue *scrq = instance;
2861         struct ibmvnic_adapter *adapter = scrq->adapter;
2862
2863         disable_scrq_irq(adapter, scrq);
2864         ibmvnic_complete_tx(adapter, scrq);
2865
2866         return IRQ_HANDLED;
2867 }
2868
2869 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2870 {
2871         struct ibmvnic_sub_crq_queue *scrq = instance;
2872         struct ibmvnic_adapter *adapter = scrq->adapter;
2873
2874         /* When booting a kdump kernel we can hit pending interrupts
2875          * prior to completing driver initialization.
2876          */
2877         if (unlikely(adapter->state != VNIC_OPEN))
2878                 return IRQ_NONE;
2879
2880         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2881
2882         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2883                 disable_scrq_irq(adapter, scrq);
2884                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2885         }
2886
2887         return IRQ_HANDLED;
2888 }
2889
2890 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2891 {
2892         struct device *dev = &adapter->vdev->dev;
2893         struct ibmvnic_sub_crq_queue *scrq;
2894         int i = 0, j = 0;
2895         int rc = 0;
2896
2897         for (i = 0; i < adapter->req_tx_queues; i++) {
2898                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2899                            i);
2900                 scrq = adapter->tx_scrq[i];
2901                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2902
2903                 if (!scrq->irq) {
2904                         rc = -EINVAL;
2905                         dev_err(dev, "Error mapping irq\n");
2906                         goto req_tx_irq_failed;
2907                 }
2908
2909                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
2910                          adapter->vdev->unit_address, i);
2911                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2912                                  0, scrq->name, scrq);
2913
2914                 if (rc) {
2915                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2916                                 scrq->irq, rc);
2917                         irq_dispose_mapping(scrq->irq);
2918                         goto req_tx_irq_failed;
2919                 }
2920         }
2921
2922         for (i = 0; i < adapter->req_rx_queues; i++) {
2923                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2924                            i);
2925                 scrq = adapter->rx_scrq[i];
2926                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2927                 if (!scrq->irq) {
2928                         rc = -EINVAL;
2929                         dev_err(dev, "Error mapping irq\n");
2930                         goto req_rx_irq_failed;
2931                 }
2932                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
2933                          adapter->vdev->unit_address, i);
2934                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2935                                  0, scrq->name, scrq);
2936                 if (rc) {
2937                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2938                                 scrq->irq, rc);
2939                         irq_dispose_mapping(scrq->irq);
2940                         goto req_rx_irq_failed;
2941                 }
2942         }
2943         return rc;
2944
2945 req_rx_irq_failed:
2946         for (j = 0; j < i; j++) {
2947                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2948                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2949         }
2950         i = adapter->req_tx_queues;
2951 req_tx_irq_failed:
2952         for (j = 0; j < i; j++) {
2953                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2954                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2955         }
2956         release_sub_crqs(adapter, 1);
2957         return rc;
2958 }
2959
2960 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2961 {
2962         struct device *dev = &adapter->vdev->dev;
2963         struct ibmvnic_sub_crq_queue **allqueues;
2964         int registered_queues = 0;
2965         int total_queues;
2966         int more = 0;
2967         int i;
2968
2969         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2970
2971         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2972         if (!allqueues)
2973                 return -1;
2974
2975         for (i = 0; i < total_queues; i++) {
2976                 allqueues[i] = init_sub_crq_queue(adapter);
2977                 if (!allqueues[i]) {
2978                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2979                         break;
2980                 }
2981                 registered_queues++;
2982         }
2983
2984         /* Make sure we were able to register the minimum number of queues */
2985         if (registered_queues <
2986             adapter->min_tx_queues + adapter->min_rx_queues) {
2987                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2988                 goto tx_failed;
2989         }
2990
2991         /* Distribute the failed allocated queues*/
2992         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2993                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2994                 switch (i % 3) {
2995                 case 0:
2996                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2997                                 adapter->req_rx_queues--;
2998                         else
2999                                 more++;
3000                         break;
3001                 case 1:
3002                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3003                                 adapter->req_tx_queues--;
3004                         else
3005                                 more++;
3006                         break;
3007                 }
3008         }
3009
3010         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3011                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3012         if (!adapter->tx_scrq)
3013                 goto tx_failed;
3014
3015         for (i = 0; i < adapter->req_tx_queues; i++) {
3016                 adapter->tx_scrq[i] = allqueues[i];
3017                 adapter->tx_scrq[i]->pool_index = i;
3018                 adapter->num_active_tx_scrqs++;
3019         }
3020
3021         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3022                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3023         if (!adapter->rx_scrq)
3024                 goto rx_failed;
3025
3026         for (i = 0; i < adapter->req_rx_queues; i++) {
3027                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3028                 adapter->rx_scrq[i]->scrq_num = i;
3029                 adapter->num_active_rx_scrqs++;
3030         }
3031
3032         kfree(allqueues);
3033         return 0;
3034
3035 rx_failed:
3036         kfree(adapter->tx_scrq);
3037         adapter->tx_scrq = NULL;
3038 tx_failed:
3039         for (i = 0; i < registered_queues; i++)
3040                 release_sub_crq_queue(adapter, allqueues[i], 1);
3041         kfree(allqueues);
3042         return -1;
3043 }
3044
3045 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3046 {
3047         struct device *dev = &adapter->vdev->dev;
3048         union ibmvnic_crq crq;
3049         int max_entries;
3050
3051         if (!retry) {
3052                 /* Sub-CRQ entries are 32 byte long */
3053                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3054
3055                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3056                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3057                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3058                         return;
3059                 }
3060
3061                 if (adapter->desired.mtu)
3062                         adapter->req_mtu = adapter->desired.mtu;
3063                 else
3064                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3065
3066                 if (!adapter->desired.tx_entries)
3067                         adapter->desired.tx_entries =
3068                                         adapter->max_tx_entries_per_subcrq;
3069                 if (!adapter->desired.rx_entries)
3070                         adapter->desired.rx_entries =
3071                                         adapter->max_rx_add_entries_per_subcrq;
3072
3073                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3074                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3075
3076                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3077                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3078                         adapter->desired.tx_entries = max_entries;
3079                 }
3080
3081                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3082                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3083                         adapter->desired.rx_entries = max_entries;
3084                 }
3085
3086                 if (adapter->desired.tx_entries)
3087                         adapter->req_tx_entries_per_subcrq =
3088                                         adapter->desired.tx_entries;
3089                 else
3090                         adapter->req_tx_entries_per_subcrq =
3091                                         adapter->max_tx_entries_per_subcrq;
3092
3093                 if (adapter->desired.rx_entries)
3094                         adapter->req_rx_add_entries_per_subcrq =
3095                                         adapter->desired.rx_entries;
3096                 else
3097                         adapter->req_rx_add_entries_per_subcrq =
3098                                         adapter->max_rx_add_entries_per_subcrq;
3099
3100                 if (adapter->desired.tx_queues)
3101                         adapter->req_tx_queues =
3102                                         adapter->desired.tx_queues;
3103                 else
3104                         adapter->req_tx_queues =
3105                                         adapter->opt_tx_comp_sub_queues;
3106
3107                 if (adapter->desired.rx_queues)
3108                         adapter->req_rx_queues =
3109                                         adapter->desired.rx_queues;
3110                 else
3111                         adapter->req_rx_queues =
3112                                         adapter->opt_rx_comp_queues;
3113
3114                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3115         }
3116
3117         memset(&crq, 0, sizeof(crq));
3118         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3119         crq.request_capability.cmd = REQUEST_CAPABILITY;
3120
3121         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3122         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3123         atomic_inc(&adapter->running_cap_crqs);
3124         ibmvnic_send_crq(adapter, &crq);
3125
3126         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3127         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3128         atomic_inc(&adapter->running_cap_crqs);
3129         ibmvnic_send_crq(adapter, &crq);
3130
3131         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3132         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3133         atomic_inc(&adapter->running_cap_crqs);
3134         ibmvnic_send_crq(adapter, &crq);
3135
3136         crq.request_capability.capability =
3137             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3138         crq.request_capability.number =
3139             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3140         atomic_inc(&adapter->running_cap_crqs);
3141         ibmvnic_send_crq(adapter, &crq);
3142
3143         crq.request_capability.capability =
3144             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3145         crq.request_capability.number =
3146             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3147         atomic_inc(&adapter->running_cap_crqs);
3148         ibmvnic_send_crq(adapter, &crq);
3149
3150         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3151         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3152         atomic_inc(&adapter->running_cap_crqs);
3153         ibmvnic_send_crq(adapter, &crq);
3154
3155         if (adapter->netdev->flags & IFF_PROMISC) {
3156                 if (adapter->promisc_supported) {
3157                         crq.request_capability.capability =
3158                             cpu_to_be16(PROMISC_REQUESTED);
3159                         crq.request_capability.number = cpu_to_be64(1);
3160                         atomic_inc(&adapter->running_cap_crqs);
3161                         ibmvnic_send_crq(adapter, &crq);
3162                 }
3163         } else {
3164                 crq.request_capability.capability =
3165                     cpu_to_be16(PROMISC_REQUESTED);
3166                 crq.request_capability.number = cpu_to_be64(0);
3167                 atomic_inc(&adapter->running_cap_crqs);
3168                 ibmvnic_send_crq(adapter, &crq);
3169         }
3170 }
3171
3172 static int pending_scrq(struct ibmvnic_adapter *adapter,
3173                         struct ibmvnic_sub_crq_queue *scrq)
3174 {
3175         union sub_crq *entry = &scrq->msgs[scrq->cur];
3176
3177         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3178                 return 1;
3179         else
3180                 return 0;
3181 }
3182
3183 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3184                                         struct ibmvnic_sub_crq_queue *scrq)
3185 {
3186         union sub_crq *entry;
3187         unsigned long flags;
3188
3189         spin_lock_irqsave(&scrq->lock, flags);
3190         entry = &scrq->msgs[scrq->cur];
3191         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3192                 if (++scrq->cur == scrq->size)
3193                         scrq->cur = 0;
3194         } else {
3195                 entry = NULL;
3196         }
3197         spin_unlock_irqrestore(&scrq->lock, flags);
3198
3199         return entry;
3200 }
3201
3202 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3203 {
3204         struct ibmvnic_crq_queue *queue = &adapter->crq;
3205         union ibmvnic_crq *crq;
3206
3207         crq = &queue->msgs[queue->cur];
3208         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3209                 if (++queue->cur == queue->size)
3210                         queue->cur = 0;
3211         } else {
3212                 crq = NULL;
3213         }
3214
3215         return crq;
3216 }
3217
3218 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3219 {
3220         switch (rc) {
3221         case H_PARAMETER:
3222                 dev_warn_ratelimited(dev,
3223                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3224                                      func, rc);
3225                 break;
3226         case H_CLOSED:
3227                 dev_warn_ratelimited(dev,
3228                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3229                                      func, rc);
3230                 break;
3231         default:
3232                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3233                 break;
3234         }
3235 }
3236
3237 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3238                        union sub_crq *sub_crq)
3239 {
3240         unsigned int ua = adapter->vdev->unit_address;
3241         struct device *dev = &adapter->vdev->dev;
3242         u64 *u64_crq = (u64 *)sub_crq;
3243         int rc;
3244
3245         netdev_dbg(adapter->netdev,
3246                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3247                    (unsigned long int)cpu_to_be64(remote_handle),
3248                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3249                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3250                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3251                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3252
3253         /* Make sure the hypervisor sees the complete request */
3254         mb();
3255
3256         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3257                                 cpu_to_be64(remote_handle),
3258                                 cpu_to_be64(u64_crq[0]),
3259                                 cpu_to_be64(u64_crq[1]),
3260                                 cpu_to_be64(u64_crq[2]),
3261                                 cpu_to_be64(u64_crq[3]));
3262
3263         if (rc)
3264                 print_subcrq_error(dev, rc, __func__);
3265
3266         return rc;
3267 }
3268
3269 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3270                                 u64 remote_handle, u64 ioba, u64 num_entries)
3271 {
3272         unsigned int ua = adapter->vdev->unit_address;
3273         struct device *dev = &adapter->vdev->dev;
3274         int rc;
3275
3276         /* Make sure the hypervisor sees the complete request */
3277         mb();
3278         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3279                                 cpu_to_be64(remote_handle),
3280                                 ioba, num_entries);
3281
3282         if (rc)
3283                 print_subcrq_error(dev, rc, __func__);
3284
3285         return rc;
3286 }
3287
3288 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3289                             union ibmvnic_crq *crq)
3290 {
3291         unsigned int ua = adapter->vdev->unit_address;
3292         struct device *dev = &adapter->vdev->dev;
3293         u64 *u64_crq = (u64 *)crq;
3294         int rc;
3295
3296         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3297                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3298                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3299
3300         if (!adapter->crq.active &&
3301             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3302                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3303                 return -EINVAL;
3304         }
3305
3306         /* Make sure the hypervisor sees the complete request */
3307         mb();
3308
3309         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3310                                 cpu_to_be64(u64_crq[0]),
3311                                 cpu_to_be64(u64_crq[1]));
3312
3313         if (rc) {
3314                 if (rc == H_CLOSED) {
3315                         dev_warn(dev, "CRQ Queue closed\n");
3316                         if (adapter->resetting)
3317                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3318                 }
3319
3320                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3321         }
3322
3323         return rc;
3324 }
3325
3326 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3327 {
3328         union ibmvnic_crq crq;
3329
3330         memset(&crq, 0, sizeof(crq));
3331         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3332         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3333         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3334
3335         return ibmvnic_send_crq(adapter, &crq);
3336 }
3337
3338 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3339 {
3340         union ibmvnic_crq crq;
3341
3342         memset(&crq, 0, sizeof(crq));
3343         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3344         crq.version_exchange.cmd = VERSION_EXCHANGE;
3345         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3346
3347         return ibmvnic_send_crq(adapter, &crq);
3348 }
3349
3350 struct vnic_login_client_data {
3351         u8      type;
3352         __be16  len;
3353         char    name[];
3354 } __packed;
3355
3356 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3357 {
3358         int len;
3359
3360         /* Calculate the amount of buffer space needed for the
3361          * vnic client data in the login buffer. There are four entries,
3362          * OS name, LPAR name, device name, and a null last entry.
3363          */
3364         len = 4 * sizeof(struct vnic_login_client_data);
3365         len += 6; /* "Linux" plus NULL */
3366         len += strlen(utsname()->nodename) + 1;
3367         len += strlen(adapter->netdev->name) + 1;
3368
3369         return len;
3370 }
3371
3372 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3373                                  struct vnic_login_client_data *vlcd)
3374 {
3375         const char *os_name = "Linux";
3376         int len;
3377
3378         /* Type 1 - LPAR OS */
3379         vlcd->type = 1;
3380         len = strlen(os_name) + 1;
3381         vlcd->len = cpu_to_be16(len);
3382         strncpy(vlcd->name, os_name, len);
3383         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3384
3385         /* Type 2 - LPAR name */
3386         vlcd->type = 2;
3387         len = strlen(utsname()->nodename) + 1;
3388         vlcd->len = cpu_to_be16(len);
3389         strncpy(vlcd->name, utsname()->nodename, len);
3390         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3391
3392         /* Type 3 - device name */
3393         vlcd->type = 3;
3394         len = strlen(adapter->netdev->name) + 1;
3395         vlcd->len = cpu_to_be16(len);
3396         strncpy(vlcd->name, adapter->netdev->name, len);
3397 }
3398
3399 static int send_login(struct ibmvnic_adapter *adapter)
3400 {
3401         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3402         struct ibmvnic_login_buffer *login_buffer;
3403         struct device *dev = &adapter->vdev->dev;
3404         dma_addr_t rsp_buffer_token;
3405         dma_addr_t buffer_token;
3406         size_t rsp_buffer_size;
3407         union ibmvnic_crq crq;
3408         size_t buffer_size;
3409         __be64 *tx_list_p;
3410         __be64 *rx_list_p;
3411         int client_data_len;
3412         struct vnic_login_client_data *vlcd;
3413         int i;
3414
3415         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3416                 netdev_err(adapter->netdev,
3417                            "RX or TX queues are not allocated, device login failed\n");
3418                 return -1;
3419         }
3420
3421         release_login_rsp_buffer(adapter);
3422         client_data_len = vnic_client_data_len(adapter);
3423
3424         buffer_size =
3425             sizeof(struct ibmvnic_login_buffer) +
3426             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3427             client_data_len;
3428
3429         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3430         if (!login_buffer)
3431                 goto buf_alloc_failed;
3432
3433         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3434                                       DMA_TO_DEVICE);
3435         if (dma_mapping_error(dev, buffer_token)) {
3436                 dev_err(dev, "Couldn't map login buffer\n");
3437                 goto buf_map_failed;
3438         }
3439
3440         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3441                           sizeof(u64) * adapter->req_tx_queues +
3442                           sizeof(u64) * adapter->req_rx_queues +
3443                           sizeof(u64) * adapter->req_rx_queues +
3444                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3445
3446         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3447         if (!login_rsp_buffer)
3448                 goto buf_rsp_alloc_failed;
3449
3450         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3451                                           rsp_buffer_size, DMA_FROM_DEVICE);
3452         if (dma_mapping_error(dev, rsp_buffer_token)) {
3453                 dev_err(dev, "Couldn't map login rsp buffer\n");
3454                 goto buf_rsp_map_failed;
3455         }
3456
3457         adapter->login_buf = login_buffer;
3458         adapter->login_buf_token = buffer_token;
3459         adapter->login_buf_sz = buffer_size;
3460         adapter->login_rsp_buf = login_rsp_buffer;
3461         adapter->login_rsp_buf_token = rsp_buffer_token;
3462         adapter->login_rsp_buf_sz = rsp_buffer_size;
3463
3464         login_buffer->len = cpu_to_be32(buffer_size);
3465         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3466         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3467         login_buffer->off_txcomp_subcrqs =
3468             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3469         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3470         login_buffer->off_rxcomp_subcrqs =
3471             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3472                         sizeof(u64) * adapter->req_tx_queues);
3473         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3474         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3475
3476         tx_list_p = (__be64 *)((char *)login_buffer +
3477                                       sizeof(struct ibmvnic_login_buffer));
3478         rx_list_p = (__be64 *)((char *)login_buffer +
3479                                       sizeof(struct ibmvnic_login_buffer) +
3480                                       sizeof(u64) * adapter->req_tx_queues);
3481
3482         for (i = 0; i < adapter->req_tx_queues; i++) {
3483                 if (adapter->tx_scrq[i]) {
3484                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3485                                                    crq_num);
3486                 }
3487         }
3488
3489         for (i = 0; i < adapter->req_rx_queues; i++) {
3490                 if (adapter->rx_scrq[i]) {
3491                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3492                                                    crq_num);
3493                 }
3494         }
3495
3496         /* Insert vNIC login client data */
3497         vlcd = (struct vnic_login_client_data *)
3498                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3499         login_buffer->client_data_offset =
3500                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3501         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3502
3503         vnic_add_client_data(adapter, vlcd);
3504
3505         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3506         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3507                 netdev_dbg(adapter->netdev, "%016lx\n",
3508                            ((unsigned long int *)(adapter->login_buf))[i]);
3509         }
3510
3511         memset(&crq, 0, sizeof(crq));
3512         crq.login.first = IBMVNIC_CRQ_CMD;
3513         crq.login.cmd = LOGIN;
3514         crq.login.ioba = cpu_to_be32(buffer_token);
3515         crq.login.len = cpu_to_be32(buffer_size);
3516         ibmvnic_send_crq(adapter, &crq);
3517
3518         return 0;
3519
3520 buf_rsp_map_failed:
3521         kfree(login_rsp_buffer);
3522 buf_rsp_alloc_failed:
3523         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3524 buf_map_failed:
3525         kfree(login_buffer);
3526 buf_alloc_failed:
3527         return -1;
3528 }
3529
3530 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3531                             u32 len, u8 map_id)
3532 {
3533         union ibmvnic_crq crq;
3534
3535         memset(&crq, 0, sizeof(crq));
3536         crq.request_map.first = IBMVNIC_CRQ_CMD;
3537         crq.request_map.cmd = REQUEST_MAP;
3538         crq.request_map.map_id = map_id;
3539         crq.request_map.ioba = cpu_to_be32(addr);
3540         crq.request_map.len = cpu_to_be32(len);
3541         return ibmvnic_send_crq(adapter, &crq);
3542 }
3543
3544 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3545 {
3546         union ibmvnic_crq crq;
3547
3548         memset(&crq, 0, sizeof(crq));
3549         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3550         crq.request_unmap.cmd = REQUEST_UNMAP;
3551         crq.request_unmap.map_id = map_id;
3552         return ibmvnic_send_crq(adapter, &crq);
3553 }
3554
3555 static void send_map_query(struct ibmvnic_adapter *adapter)
3556 {
3557         union ibmvnic_crq crq;
3558
3559         memset(&crq, 0, sizeof(crq));
3560         crq.query_map.first = IBMVNIC_CRQ_CMD;
3561         crq.query_map.cmd = QUERY_MAP;
3562         ibmvnic_send_crq(adapter, &crq);
3563 }
3564
3565 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3566 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3567 {
3568         union ibmvnic_crq crq;
3569
3570         atomic_set(&adapter->running_cap_crqs, 0);
3571         memset(&crq, 0, sizeof(crq));
3572         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3573         crq.query_capability.cmd = QUERY_CAPABILITY;
3574
3575         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3576         atomic_inc(&adapter->running_cap_crqs);
3577         ibmvnic_send_crq(adapter, &crq);
3578
3579         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3580         atomic_inc(&adapter->running_cap_crqs);
3581         ibmvnic_send_crq(adapter, &crq);
3582
3583         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3584         atomic_inc(&adapter->running_cap_crqs);
3585         ibmvnic_send_crq(adapter, &crq);
3586
3587         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3588         atomic_inc(&adapter->running_cap_crqs);
3589         ibmvnic_send_crq(adapter, &crq);
3590
3591         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3592         atomic_inc(&adapter->running_cap_crqs);
3593         ibmvnic_send_crq(adapter, &crq);
3594
3595         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3596         atomic_inc(&adapter->running_cap_crqs);
3597         ibmvnic_send_crq(adapter, &crq);
3598
3599         crq.query_capability.capability =
3600             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3601         atomic_inc(&adapter->running_cap_crqs);
3602         ibmvnic_send_crq(adapter, &crq);
3603
3604         crq.query_capability.capability =
3605             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3606         atomic_inc(&adapter->running_cap_crqs);
3607         ibmvnic_send_crq(adapter, &crq);
3608
3609         crq.query_capability.capability =
3610             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3611         atomic_inc(&adapter->running_cap_crqs);
3612         ibmvnic_send_crq(adapter, &crq);
3613
3614         crq.query_capability.capability =
3615             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3616         atomic_inc(&adapter->running_cap_crqs);
3617         ibmvnic_send_crq(adapter, &crq);
3618
3619         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3620         atomic_inc(&adapter->running_cap_crqs);
3621         ibmvnic_send_crq(adapter, &crq);
3622
3623         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3624         atomic_inc(&adapter->running_cap_crqs);
3625         ibmvnic_send_crq(adapter, &crq);
3626
3627         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3628         atomic_inc(&adapter->running_cap_crqs);
3629         ibmvnic_send_crq(adapter, &crq);
3630
3631         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3632         atomic_inc(&adapter->running_cap_crqs);
3633         ibmvnic_send_crq(adapter, &crq);
3634
3635         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3636         atomic_inc(&adapter->running_cap_crqs);
3637         ibmvnic_send_crq(adapter, &crq);
3638
3639         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3640         atomic_inc(&adapter->running_cap_crqs);
3641         ibmvnic_send_crq(adapter, &crq);
3642
3643         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3644         atomic_inc(&adapter->running_cap_crqs);
3645         ibmvnic_send_crq(adapter, &crq);
3646
3647         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3648         atomic_inc(&adapter->running_cap_crqs);
3649         ibmvnic_send_crq(adapter, &crq);
3650
3651         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3652         atomic_inc(&adapter->running_cap_crqs);
3653         ibmvnic_send_crq(adapter, &crq);
3654
3655         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3656         atomic_inc(&adapter->running_cap_crqs);
3657         ibmvnic_send_crq(adapter, &crq);
3658
3659         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3660         atomic_inc(&adapter->running_cap_crqs);
3661         ibmvnic_send_crq(adapter, &crq);
3662
3663         crq.query_capability.capability =
3664                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3665         atomic_inc(&adapter->running_cap_crqs);
3666         ibmvnic_send_crq(adapter, &crq);
3667
3668         crq.query_capability.capability =
3669                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3670         atomic_inc(&adapter->running_cap_crqs);
3671         ibmvnic_send_crq(adapter, &crq);
3672
3673         crq.query_capability.capability =
3674                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3675         atomic_inc(&adapter->running_cap_crqs);
3676         ibmvnic_send_crq(adapter, &crq);
3677
3678         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3679         atomic_inc(&adapter->running_cap_crqs);
3680         ibmvnic_send_crq(adapter, &crq);
3681 }
3682
3683 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3684                                 struct ibmvnic_adapter *adapter)
3685 {
3686         struct device *dev = &adapter->vdev->dev;
3687
3688         if (crq->get_vpd_size_rsp.rc.code) {
3689                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3690                         crq->get_vpd_size_rsp.rc.code);
3691                 complete(&adapter->fw_done);
3692                 return;
3693         }
3694
3695         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3696         complete(&adapter->fw_done);
3697 }
3698
3699 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3700                            struct ibmvnic_adapter *adapter)
3701 {
3702         struct device *dev = &adapter->vdev->dev;
3703         unsigned char *substr = NULL;
3704         u8 fw_level_len = 0;
3705
3706         memset(adapter->fw_version, 0, 32);
3707
3708         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3709                          DMA_FROM_DEVICE);
3710
3711         if (crq->get_vpd_rsp.rc.code) {
3712                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3713                         crq->get_vpd_rsp.rc.code);
3714                 goto complete;
3715         }
3716
3717         /* get the position of the firmware version info
3718          * located after the ASCII 'RM' substring in the buffer
3719          */
3720         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3721         if (!substr) {
3722                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3723                 goto complete;
3724         }
3725
3726         /* get length of firmware level ASCII substring */
3727         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3728                 fw_level_len = *(substr + 2);
3729         } else {
3730                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3731                 goto complete;
3732         }
3733
3734         /* copy firmware version string from vpd into adapter */
3735         if ((substr + 3 + fw_level_len) <
3736             (adapter->vpd->buff + adapter->vpd->len)) {
3737                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3738         } else {
3739                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3740         }
3741
3742 complete:
3743         if (adapter->fw_version[0] == '\0')
3744                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3745         complete(&adapter->fw_done);
3746 }
3747
3748 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3749 {
3750         struct device *dev = &adapter->vdev->dev;
3751         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3752         netdev_features_t old_hw_features = 0;
3753         union ibmvnic_crq crq;
3754         int i;
3755
3756         dma_unmap_single(dev, adapter->ip_offload_tok,
3757                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3758
3759         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3760         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3761                 netdev_dbg(adapter->netdev, "%016lx\n",
3762                            ((unsigned long int *)(buf))[i]);
3763
3764         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3765         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3766         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3767                    buf->tcp_ipv4_chksum);
3768         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3769                    buf->tcp_ipv6_chksum);
3770         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3771                    buf->udp_ipv4_chksum);
3772         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3773                    buf->udp_ipv6_chksum);
3774         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3775                    buf->large_tx_ipv4);
3776         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3777                    buf->large_tx_ipv6);
3778         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3779                    buf->large_rx_ipv4);
3780         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3781                    buf->large_rx_ipv6);
3782         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3783                    buf->max_ipv4_header_size);
3784         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3785                    buf->max_ipv6_header_size);
3786         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3787                    buf->max_tcp_header_size);
3788         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3789                    buf->max_udp_header_size);
3790         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3791                    buf->max_large_tx_size);
3792         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3793                    buf->max_large_rx_size);
3794         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3795                    buf->ipv6_extension_header);
3796         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3797                    buf->tcp_pseudosum_req);
3798         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3799                    buf->num_ipv6_ext_headers);
3800         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3801                    buf->off_ipv6_ext_headers);
3802
3803         adapter->ip_offload_ctrl_tok =
3804             dma_map_single(dev, &adapter->ip_offload_ctrl,
3805                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3806
3807         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3808                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3809                 return;
3810         }
3811
3812         adapter->ip_offload_ctrl.len =
3813             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3814         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3815         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3816         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3817         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3818         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3819         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3820         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3821         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3822         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3823
3824         /* large_rx disabled for now, additional features needed */
3825         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3826         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3827
3828         if (adapter->state != VNIC_PROBING) {
3829                 old_hw_features = adapter->netdev->hw_features;
3830                 adapter->netdev->hw_features = 0;
3831         }
3832
3833         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
3834
3835         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3836                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
3837
3838         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3839                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
3840
3841         if ((adapter->netdev->features &
3842             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3843                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
3844
3845         if (buf->large_tx_ipv4)
3846                 adapter->netdev->hw_features |= NETIF_F_TSO;
3847         if (buf->large_tx_ipv6)
3848                 adapter->netdev->hw_features |= NETIF_F_TSO6;
3849
3850         if (adapter->state == VNIC_PROBING) {
3851                 adapter->netdev->features |= adapter->netdev->hw_features;
3852         } else if (old_hw_features != adapter->netdev->hw_features) {
3853                 netdev_features_t tmp = 0;
3854
3855                 /* disable features no longer supported */
3856                 adapter->netdev->features &= adapter->netdev->hw_features;
3857                 /* turn on features now supported if previously enabled */
3858                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
3859                         adapter->netdev->hw_features;
3860                 adapter->netdev->features |=
3861                                 tmp & adapter->netdev->wanted_features;
3862         }
3863
3864         memset(&crq, 0, sizeof(crq));
3865         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3866         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3867         crq.control_ip_offload.len =
3868             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3869         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3870         ibmvnic_send_crq(adapter, &crq);
3871 }
3872
3873 static const char *ibmvnic_fw_err_cause(u16 cause)
3874 {
3875         switch (cause) {
3876         case ADAPTER_PROBLEM:
3877                 return "adapter problem";
3878         case BUS_PROBLEM:
3879                 return "bus problem";
3880         case FW_PROBLEM:
3881                 return "firmware problem";
3882         case DD_PROBLEM:
3883                 return "device driver problem";
3884         case EEH_RECOVERY:
3885                 return "EEH recovery";
3886         case FW_UPDATED:
3887                 return "firmware updated";
3888         case LOW_MEMORY:
3889                 return "low Memory";
3890         default:
3891                 return "unknown";
3892         }
3893 }
3894
3895 static void handle_error_indication(union ibmvnic_crq *crq,
3896                                     struct ibmvnic_adapter *adapter)
3897 {
3898         struct device *dev = &adapter->vdev->dev;
3899         u16 cause;
3900
3901         cause = be16_to_cpu(crq->error_indication.error_cause);
3902
3903         dev_warn_ratelimited(dev,
3904                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3905                              crq->error_indication.flags
3906                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3907                              ibmvnic_fw_err_cause(cause));
3908
3909         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3910                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3911         else
3912                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3913 }
3914
3915 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3916                                  struct ibmvnic_adapter *adapter)
3917 {
3918         struct net_device *netdev = adapter->netdev;
3919         struct device *dev = &adapter->vdev->dev;
3920         long rc;
3921
3922         rc = crq->change_mac_addr_rsp.rc.code;
3923         if (rc) {
3924                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3925                 goto out;
3926         }
3927         ether_addr_copy(netdev->dev_addr,
3928                         &crq->change_mac_addr_rsp.mac_addr[0]);
3929 out:
3930         complete(&adapter->fw_done);
3931         return rc;
3932 }
3933
3934 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3935                                    struct ibmvnic_adapter *adapter)
3936 {
3937         struct device *dev = &adapter->vdev->dev;
3938         u64 *req_value;
3939         char *name;
3940
3941         atomic_dec(&adapter->running_cap_crqs);
3942         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3943         case REQ_TX_QUEUES:
3944                 req_value = &adapter->req_tx_queues;
3945                 name = "tx";
3946                 break;
3947         case REQ_RX_QUEUES:
3948                 req_value = &adapter->req_rx_queues;
3949                 name = "rx";
3950                 break;
3951         case REQ_RX_ADD_QUEUES:
3952                 req_value = &adapter->req_rx_add_queues;
3953                 name = "rx_add";
3954                 break;
3955         case REQ_TX_ENTRIES_PER_SUBCRQ:
3956                 req_value = &adapter->req_tx_entries_per_subcrq;
3957                 name = "tx_entries_per_subcrq";
3958                 break;
3959         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3960                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3961                 name = "rx_add_entries_per_subcrq";
3962                 break;
3963         case REQ_MTU:
3964                 req_value = &adapter->req_mtu;
3965                 name = "mtu";
3966                 break;
3967         case PROMISC_REQUESTED:
3968                 req_value = &adapter->promisc;
3969                 name = "promisc";
3970                 break;
3971         default:
3972                 dev_err(dev, "Got invalid cap request rsp %d\n",
3973                         crq->request_capability.capability);
3974                 return;
3975         }
3976
3977         switch (crq->request_capability_rsp.rc.code) {
3978         case SUCCESS:
3979                 break;
3980         case PARTIALSUCCESS:
3981                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3982                          *req_value,
3983                          (long int)be64_to_cpu(crq->request_capability_rsp.
3984                                                number), name);
3985
3986                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3987                     REQ_MTU) {
3988                         pr_err("mtu of %llu is not supported. Reverting.\n",
3989                                *req_value);
3990                         *req_value = adapter->fallback.mtu;
3991                 } else {
3992                         *req_value =
3993                                 be64_to_cpu(crq->request_capability_rsp.number);
3994                 }
3995
3996                 ibmvnic_send_req_caps(adapter, 1);
3997                 return;
3998         default:
3999                 dev_err(dev, "Error %d in request cap rsp\n",
4000                         crq->request_capability_rsp.rc.code);
4001                 return;
4002         }
4003
4004         /* Done receiving requested capabilities, query IP offload support */
4005         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4006                 union ibmvnic_crq newcrq;
4007                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4008                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4009                     &adapter->ip_offload_buf;
4010
4011                 adapter->wait_capability = false;
4012                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4013                                                          buf_sz,
4014                                                          DMA_FROM_DEVICE);
4015
4016                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4017                         if (!firmware_has_feature(FW_FEATURE_CMO))
4018                                 dev_err(dev, "Couldn't map offload buffer\n");
4019                         return;
4020                 }
4021
4022                 memset(&newcrq, 0, sizeof(newcrq));
4023                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4024                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4025                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4026                 newcrq.query_ip_offload.ioba =
4027                     cpu_to_be32(adapter->ip_offload_tok);
4028
4029                 ibmvnic_send_crq(adapter, &newcrq);
4030         }
4031 }
4032
4033 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4034                             struct ibmvnic_adapter *adapter)
4035 {
4036         struct device *dev = &adapter->vdev->dev;
4037         struct net_device *netdev = adapter->netdev;
4038         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4039         struct ibmvnic_login_buffer *login = adapter->login_buf;
4040         int i;
4041
4042         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4043                          DMA_TO_DEVICE);
4044         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4045                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4046
4047         /* If the number of queues requested can't be allocated by the
4048          * server, the login response will return with code 1. We will need
4049          * to resend the login buffer with fewer queues requested.
4050          */
4051         if (login_rsp_crq->generic.rc.code) {
4052                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4053                 complete(&adapter->init_done);
4054                 return 0;
4055         }
4056
4057         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4058
4059         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4060         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4061                 netdev_dbg(adapter->netdev, "%016lx\n",
4062                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4063         }
4064
4065         /* Sanity checks */
4066         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4067             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4068              adapter->req_rx_add_queues !=
4069              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4070                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4071                 ibmvnic_remove(adapter->vdev);
4072                 return -EIO;
4073         }
4074         release_login_buffer(adapter);
4075         complete(&adapter->init_done);
4076
4077         return 0;
4078 }
4079
4080 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4081                                      struct ibmvnic_adapter *adapter)
4082 {
4083         struct device *dev = &adapter->vdev->dev;
4084         long rc;
4085
4086         rc = crq->request_unmap_rsp.rc.code;
4087         if (rc)
4088                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4089 }
4090
4091 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4092                                  struct ibmvnic_adapter *adapter)
4093 {
4094         struct net_device *netdev = adapter->netdev;
4095         struct device *dev = &adapter->vdev->dev;
4096         long rc;
4097
4098         rc = crq->query_map_rsp.rc.code;
4099         if (rc) {
4100                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4101                 return;
4102         }
4103         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4104                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4105                    crq->query_map_rsp.free_pages);
4106 }
4107
4108 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4109                                  struct ibmvnic_adapter *adapter)
4110 {
4111         struct net_device *netdev = adapter->netdev;
4112         struct device *dev = &adapter->vdev->dev;
4113         long rc;
4114
4115         atomic_dec(&adapter->running_cap_crqs);
4116         netdev_dbg(netdev, "Outstanding queries: %d\n",
4117                    atomic_read(&adapter->running_cap_crqs));
4118         rc = crq->query_capability.rc.code;
4119         if (rc) {
4120                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4121                 goto out;
4122         }
4123
4124         switch (be16_to_cpu(crq->query_capability.capability)) {
4125         case MIN_TX_QUEUES:
4126                 adapter->min_tx_queues =
4127                     be64_to_cpu(crq->query_capability.number);
4128                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4129                            adapter->min_tx_queues);
4130                 break;
4131         case MIN_RX_QUEUES:
4132                 adapter->min_rx_queues =
4133                     be64_to_cpu(crq->query_capability.number);
4134                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4135                            adapter->min_rx_queues);
4136                 break;
4137         case MIN_RX_ADD_QUEUES:
4138                 adapter->min_rx_add_queues =
4139                     be64_to_cpu(crq->query_capability.number);
4140                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4141                            adapter->min_rx_add_queues);
4142                 break;
4143         case MAX_TX_QUEUES:
4144                 adapter->max_tx_queues =
4145                     be64_to_cpu(crq->query_capability.number);
4146                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4147                            adapter->max_tx_queues);
4148                 break;
4149         case MAX_RX_QUEUES:
4150                 adapter->max_rx_queues =
4151                     be64_to_cpu(crq->query_capability.number);
4152                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4153                            adapter->max_rx_queues);
4154                 break;
4155         case MAX_RX_ADD_QUEUES:
4156                 adapter->max_rx_add_queues =
4157                     be64_to_cpu(crq->query_capability.number);
4158                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4159                            adapter->max_rx_add_queues);
4160                 break;
4161         case MIN_TX_ENTRIES_PER_SUBCRQ:
4162                 adapter->min_tx_entries_per_subcrq =
4163                     be64_to_cpu(crq->query_capability.number);
4164                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4165                            adapter->min_tx_entries_per_subcrq);
4166                 break;
4167         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4168                 adapter->min_rx_add_entries_per_subcrq =
4169                     be64_to_cpu(crq->query_capability.number);
4170                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4171                            adapter->min_rx_add_entries_per_subcrq);
4172                 break;
4173         case MAX_TX_ENTRIES_PER_SUBCRQ:
4174                 adapter->max_tx_entries_per_subcrq =
4175                     be64_to_cpu(crq->query_capability.number);
4176                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4177                            adapter->max_tx_entries_per_subcrq);
4178                 break;
4179         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4180                 adapter->max_rx_add_entries_per_subcrq =
4181                     be64_to_cpu(crq->query_capability.number);
4182                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4183                            adapter->max_rx_add_entries_per_subcrq);
4184                 break;
4185         case TCP_IP_OFFLOAD:
4186                 adapter->tcp_ip_offload =
4187                     be64_to_cpu(crq->query_capability.number);
4188                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4189                            adapter->tcp_ip_offload);
4190                 break;
4191         case PROMISC_SUPPORTED:
4192                 adapter->promisc_supported =
4193                     be64_to_cpu(crq->query_capability.number);
4194                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4195                            adapter->promisc_supported);
4196                 break;
4197         case MIN_MTU:
4198                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4199                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4200                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4201                 break;
4202         case MAX_MTU:
4203                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4204                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4205                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4206                 break;
4207         case MAX_MULTICAST_FILTERS:
4208                 adapter->max_multicast_filters =
4209                     be64_to_cpu(crq->query_capability.number);
4210                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4211                            adapter->max_multicast_filters);
4212                 break;
4213         case VLAN_HEADER_INSERTION:
4214                 adapter->vlan_header_insertion =
4215                     be64_to_cpu(crq->query_capability.number);
4216                 if (adapter->vlan_header_insertion)
4217                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4218                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4219                            adapter->vlan_header_insertion);
4220                 break;
4221         case RX_VLAN_HEADER_INSERTION:
4222                 adapter->rx_vlan_header_insertion =
4223                     be64_to_cpu(crq->query_capability.number);
4224                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4225                            adapter->rx_vlan_header_insertion);
4226                 break;
4227         case MAX_TX_SG_ENTRIES:
4228                 adapter->max_tx_sg_entries =
4229                     be64_to_cpu(crq->query_capability.number);
4230                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4231                            adapter->max_tx_sg_entries);
4232                 break;
4233         case RX_SG_SUPPORTED:
4234                 adapter->rx_sg_supported =
4235                     be64_to_cpu(crq->query_capability.number);
4236                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4237                            adapter->rx_sg_supported);
4238                 break;
4239         case OPT_TX_COMP_SUB_QUEUES:
4240                 adapter->opt_tx_comp_sub_queues =
4241                     be64_to_cpu(crq->query_capability.number);
4242                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4243                            adapter->opt_tx_comp_sub_queues);
4244                 break;
4245         case OPT_RX_COMP_QUEUES:
4246                 adapter->opt_rx_comp_queues =
4247                     be64_to_cpu(crq->query_capability.number);
4248                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4249                            adapter->opt_rx_comp_queues);
4250                 break;
4251         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4252                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4253                     be64_to_cpu(crq->query_capability.number);
4254                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4255                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4256                 break;
4257         case OPT_TX_ENTRIES_PER_SUBCRQ:
4258                 adapter->opt_tx_entries_per_subcrq =
4259                     be64_to_cpu(crq->query_capability.number);
4260                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4261                            adapter->opt_tx_entries_per_subcrq);
4262                 break;
4263         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4264                 adapter->opt_rxba_entries_per_subcrq =
4265                     be64_to_cpu(crq->query_capability.number);
4266                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4267                            adapter->opt_rxba_entries_per_subcrq);
4268                 break;
4269         case TX_RX_DESC_REQ:
4270                 adapter->tx_rx_desc_req = crq->query_capability.number;
4271                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4272                            adapter->tx_rx_desc_req);
4273                 break;
4274
4275         default:
4276                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4277                            crq->query_capability.capability);
4278         }
4279
4280 out:
4281         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4282                 adapter->wait_capability = false;
4283                 ibmvnic_send_req_caps(adapter, 0);
4284         }
4285 }
4286
4287 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4288 {
4289         union ibmvnic_crq crq;
4290         int rc;
4291
4292         memset(&crq, 0, sizeof(crq));
4293         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4294         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4295         init_completion(&adapter->fw_done);
4296         rc = ibmvnic_send_crq(adapter, &crq);
4297         if (rc)
4298                 return rc;
4299         wait_for_completion(&adapter->fw_done);
4300         return adapter->fw_done_rc ? -EIO : 0;
4301 }
4302
4303 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4304                                        struct ibmvnic_adapter *adapter)
4305 {
4306         struct net_device *netdev = adapter->netdev;
4307         int rc;
4308
4309         rc = crq->query_phys_parms_rsp.rc.code;
4310         if (rc) {
4311                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4312                 return rc;
4313         }
4314         switch (cpu_to_be32(crq->query_phys_parms_rsp.speed)) {
4315         case IBMVNIC_10MBPS:
4316                 adapter->speed = SPEED_10;
4317                 break;
4318         case IBMVNIC_100MBPS:
4319                 adapter->speed = SPEED_100;
4320                 break;
4321         case IBMVNIC_1GBPS:
4322                 adapter->speed = SPEED_1000;
4323                 break;
4324         case IBMVNIC_10GBP:
4325                 adapter->speed = SPEED_10000;
4326                 break;
4327         case IBMVNIC_25GBPS:
4328                 adapter->speed = SPEED_25000;
4329                 break;
4330         case IBMVNIC_40GBPS:
4331                 adapter->speed = SPEED_40000;
4332                 break;
4333         case IBMVNIC_50GBPS:
4334                 adapter->speed = SPEED_50000;
4335                 break;
4336         case IBMVNIC_100GBPS:
4337                 adapter->speed = SPEED_100000;
4338                 break;
4339         default:
4340                 netdev_warn(netdev, "Unknown speed 0x%08x\n",
4341                             cpu_to_be32(crq->query_phys_parms_rsp.speed));
4342                 adapter->speed = SPEED_UNKNOWN;
4343         }
4344         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4345                 adapter->duplex = DUPLEX_FULL;
4346         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4347                 adapter->duplex = DUPLEX_HALF;
4348         else
4349                 adapter->duplex = DUPLEX_UNKNOWN;
4350
4351         return rc;
4352 }
4353
4354 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4355                                struct ibmvnic_adapter *adapter)
4356 {
4357         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4358         struct net_device *netdev = adapter->netdev;
4359         struct device *dev = &adapter->vdev->dev;
4360         u64 *u64_crq = (u64 *)crq;
4361         long rc;
4362
4363         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4364                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4365                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4366         switch (gen_crq->first) {
4367         case IBMVNIC_CRQ_INIT_RSP:
4368                 switch (gen_crq->cmd) {
4369                 case IBMVNIC_CRQ_INIT:
4370                         dev_info(dev, "Partner initialized\n");
4371                         adapter->from_passive_init = true;
4372                         adapter->failover_pending = false;
4373                         if (!completion_done(&adapter->init_done)) {
4374                                 complete(&adapter->init_done);
4375                                 adapter->init_done_rc = -EIO;
4376                         }
4377                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4378                         break;
4379                 case IBMVNIC_CRQ_INIT_COMPLETE:
4380                         dev_info(dev, "Partner initialization complete\n");
4381                         adapter->crq.active = true;
4382                         send_version_xchg(adapter);
4383                         break;
4384                 default:
4385                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4386                 }
4387                 return;
4388         case IBMVNIC_CRQ_XPORT_EVENT:
4389                 netif_carrier_off(netdev);
4390                 adapter->crq.active = false;
4391                 if (adapter->resetting)
4392                         adapter->force_reset_recovery = true;
4393                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4394                         dev_info(dev, "Migrated, re-enabling adapter\n");
4395                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4396                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4397                         dev_info(dev, "Backing device failover detected\n");
4398                         adapter->failover_pending = true;
4399                 } else {
4400                         /* The adapter lost the connection */
4401                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4402                                 gen_crq->cmd);
4403                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4404                 }
4405                 return;
4406         case IBMVNIC_CRQ_CMD_RSP:
4407                 break;
4408         default:
4409                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4410                         gen_crq->first);
4411                 return;
4412         }
4413
4414         switch (gen_crq->cmd) {
4415         case VERSION_EXCHANGE_RSP:
4416                 rc = crq->version_exchange_rsp.rc.code;
4417                 if (rc) {
4418                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4419                         break;
4420                 }
4421                 dev_info(dev, "Partner protocol version is %d\n",
4422                          crq->version_exchange_rsp.version);
4423                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4424                     ibmvnic_version)
4425                         ibmvnic_version =
4426                             be16_to_cpu(crq->version_exchange_rsp.version);
4427                 send_cap_queries(adapter);
4428                 break;
4429         case QUERY_CAPABILITY_RSP:
4430                 handle_query_cap_rsp(crq, adapter);
4431                 break;
4432         case QUERY_MAP_RSP:
4433                 handle_query_map_rsp(crq, adapter);
4434                 break;
4435         case REQUEST_MAP_RSP:
4436                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4437                 complete(&adapter->fw_done);
4438                 break;
4439         case REQUEST_UNMAP_RSP:
4440                 handle_request_unmap_rsp(crq, adapter);
4441                 break;
4442         case REQUEST_CAPABILITY_RSP:
4443                 handle_request_cap_rsp(crq, adapter);
4444                 break;
4445         case LOGIN_RSP:
4446                 netdev_dbg(netdev, "Got Login Response\n");
4447                 handle_login_rsp(crq, adapter);
4448                 break;
4449         case LOGICAL_LINK_STATE_RSP:
4450                 netdev_dbg(netdev,
4451                            "Got Logical Link State Response, state: %d rc: %d\n",
4452                            crq->logical_link_state_rsp.link_state,
4453                            crq->logical_link_state_rsp.rc.code);
4454                 adapter->logical_link_state =
4455                     crq->logical_link_state_rsp.link_state;
4456                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4457                 complete(&adapter->init_done);
4458                 break;
4459         case LINK_STATE_INDICATION:
4460                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4461                 adapter->phys_link_state =
4462                     crq->link_state_indication.phys_link_state;
4463                 adapter->logical_link_state =
4464                     crq->link_state_indication.logical_link_state;
4465                 if (adapter->phys_link_state && adapter->logical_link_state)
4466                         netif_carrier_on(netdev);
4467                 else
4468                         netif_carrier_off(netdev);
4469                 break;
4470         case CHANGE_MAC_ADDR_RSP:
4471                 netdev_dbg(netdev, "Got MAC address change Response\n");
4472                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4473                 break;
4474         case ERROR_INDICATION:
4475                 netdev_dbg(netdev, "Got Error Indication\n");
4476                 handle_error_indication(crq, adapter);
4477                 break;
4478         case REQUEST_STATISTICS_RSP:
4479                 netdev_dbg(netdev, "Got Statistics Response\n");
4480                 complete(&adapter->stats_done);
4481                 break;
4482         case QUERY_IP_OFFLOAD_RSP:
4483                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4484                 handle_query_ip_offload_rsp(adapter);
4485                 break;
4486         case MULTICAST_CTRL_RSP:
4487                 netdev_dbg(netdev, "Got multicast control Response\n");
4488                 break;
4489         case CONTROL_IP_OFFLOAD_RSP:
4490                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4491                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4492                                  sizeof(adapter->ip_offload_ctrl),
4493                                  DMA_TO_DEVICE);
4494                 complete(&adapter->init_done);
4495                 break;
4496         case COLLECT_FW_TRACE_RSP:
4497                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4498                 complete(&adapter->fw_done);
4499                 break;
4500         case GET_VPD_SIZE_RSP:
4501                 handle_vpd_size_rsp(crq, adapter);
4502                 break;
4503         case GET_VPD_RSP:
4504                 handle_vpd_rsp(crq, adapter);
4505                 break;
4506         case QUERY_PHYS_PARMS_RSP:
4507                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4508                 complete(&adapter->fw_done);
4509                 break;
4510         default:
4511                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4512                            gen_crq->cmd);
4513         }
4514 }
4515
4516 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4517 {
4518         struct ibmvnic_adapter *adapter = instance;
4519
4520         tasklet_schedule(&adapter->tasklet);
4521         return IRQ_HANDLED;
4522 }
4523
4524 static void ibmvnic_tasklet(void *data)
4525 {
4526         struct ibmvnic_adapter *adapter = data;
4527         struct ibmvnic_crq_queue *queue = &adapter->crq;
4528         union ibmvnic_crq *crq;
4529         unsigned long flags;
4530         bool done = false;
4531
4532         spin_lock_irqsave(&queue->lock, flags);
4533         while (!done) {
4534                 /* Pull all the valid messages off the CRQ */
4535                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4536                         ibmvnic_handle_crq(crq, adapter);
4537                         crq->generic.first = 0;
4538                 }
4539
4540                 /* remain in tasklet until all
4541                  * capabilities responses are received
4542                  */
4543                 if (!adapter->wait_capability)
4544                         done = true;
4545         }
4546         /* if capabilities CRQ's were sent in this tasklet, the following
4547          * tasklet must wait until all responses are received
4548          */
4549         if (atomic_read(&adapter->running_cap_crqs) != 0)
4550                 adapter->wait_capability = true;
4551         spin_unlock_irqrestore(&queue->lock, flags);
4552 }
4553
4554 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4555 {
4556         struct vio_dev *vdev = adapter->vdev;
4557         int rc;
4558
4559         do {
4560                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4561         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4562
4563         if (rc)
4564                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4565
4566         return rc;
4567 }
4568
4569 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4570 {
4571         struct ibmvnic_crq_queue *crq = &adapter->crq;
4572         struct device *dev = &adapter->vdev->dev;
4573         struct vio_dev *vdev = adapter->vdev;
4574         int rc;
4575
4576         /* Close the CRQ */
4577         do {
4578                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4579         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4580
4581         /* Clean out the queue */
4582         memset(crq->msgs, 0, PAGE_SIZE);
4583         crq->cur = 0;
4584         crq->active = false;
4585
4586         /* And re-open it again */
4587         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4588                                 crq->msg_token, PAGE_SIZE);
4589
4590         if (rc == H_CLOSED)
4591                 /* Adapter is good, but other end is not ready */
4592                 dev_warn(dev, "Partner adapter not ready\n");
4593         else if (rc != 0)
4594                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4595
4596         return rc;
4597 }
4598
4599 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4600 {
4601         struct ibmvnic_crq_queue *crq = &adapter->crq;
4602         struct vio_dev *vdev = adapter->vdev;
4603         long rc;
4604
4605         if (!crq->msgs)
4606                 return;
4607
4608         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4609         free_irq(vdev->irq, adapter);
4610         tasklet_kill(&adapter->tasklet);
4611         do {
4612                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4613         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4614
4615         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4616                          DMA_BIDIRECTIONAL);
4617         free_page((unsigned long)crq->msgs);
4618         crq->msgs = NULL;
4619         crq->active = false;
4620 }
4621
4622 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4623 {
4624         struct ibmvnic_crq_queue *crq = &adapter->crq;
4625         struct device *dev = &adapter->vdev->dev;
4626         struct vio_dev *vdev = adapter->vdev;
4627         int rc, retrc = -ENOMEM;
4628
4629         if (crq->msgs)
4630                 return 0;
4631
4632         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4633         /* Should we allocate more than one page? */
4634
4635         if (!crq->msgs)
4636                 return -ENOMEM;
4637
4638         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4639         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4640                                         DMA_BIDIRECTIONAL);
4641         if (dma_mapping_error(dev, crq->msg_token))
4642                 goto map_failed;
4643
4644         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4645                                 crq->msg_token, PAGE_SIZE);
4646
4647         if (rc == H_RESOURCE)
4648                 /* maybe kexecing and resource is busy. try a reset */
4649                 rc = ibmvnic_reset_crq(adapter);
4650         retrc = rc;
4651
4652         if (rc == H_CLOSED) {
4653                 dev_warn(dev, "Partner adapter not ready\n");
4654         } else if (rc) {
4655                 dev_warn(dev, "Error %d opening adapter\n", rc);
4656                 goto reg_crq_failed;
4657         }
4658
4659         retrc = 0;
4660
4661         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4662                      (unsigned long)adapter);
4663
4664         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4665         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4666                  adapter->vdev->unit_address);
4667         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4668         if (rc) {
4669                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4670                         vdev->irq, rc);
4671                 goto req_irq_failed;
4672         }
4673
4674         rc = vio_enable_interrupts(vdev);
4675         if (rc) {
4676                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4677                 goto req_irq_failed;
4678         }
4679
4680         crq->cur = 0;
4681         spin_lock_init(&crq->lock);
4682
4683         return retrc;
4684
4685 req_irq_failed:
4686         tasklet_kill(&adapter->tasklet);
4687         do {
4688                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4689         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4690 reg_crq_failed:
4691         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4692 map_failed:
4693         free_page((unsigned long)crq->msgs);
4694         crq->msgs = NULL;
4695         return retrc;
4696 }
4697
4698 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4699 {
4700         struct device *dev = &adapter->vdev->dev;
4701         unsigned long timeout = msecs_to_jiffies(30000);
4702         u64 old_num_rx_queues, old_num_tx_queues;
4703         int rc;
4704
4705         adapter->from_passive_init = false;
4706
4707         old_num_rx_queues = adapter->req_rx_queues;
4708         old_num_tx_queues = adapter->req_tx_queues;
4709
4710         reinit_completion(&adapter->init_done);
4711         adapter->init_done_rc = 0;
4712         ibmvnic_send_crq_init(adapter);
4713         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4714                 dev_err(dev, "Initialization sequence timed out\n");
4715                 return -1;
4716         }
4717
4718         if (adapter->init_done_rc) {
4719                 release_crq_queue(adapter);
4720                 return adapter->init_done_rc;
4721         }
4722
4723         if (adapter->from_passive_init) {
4724                 adapter->state = VNIC_OPEN;
4725                 adapter->from_passive_init = false;
4726                 return -1;
4727         }
4728
4729         if (adapter->resetting && !adapter->wait_for_reset &&
4730             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4731                 if (adapter->req_rx_queues != old_num_rx_queues ||
4732                     adapter->req_tx_queues != old_num_tx_queues) {
4733                         release_sub_crqs(adapter, 0);
4734                         rc = init_sub_crqs(adapter);
4735                 } else {
4736                         rc = reset_sub_crq_queues(adapter);
4737                 }
4738         } else {
4739                 rc = init_sub_crqs(adapter);
4740         }
4741
4742         if (rc) {
4743                 dev_err(dev, "Initialization of sub crqs failed\n");
4744                 release_crq_queue(adapter);
4745                 return rc;
4746         }
4747
4748         rc = init_sub_crq_irqs(adapter);
4749         if (rc) {
4750                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4751                 release_crq_queue(adapter);
4752         }
4753
4754         return rc;
4755 }
4756
4757 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4758 {
4759         struct device *dev = &adapter->vdev->dev;
4760         unsigned long timeout = msecs_to_jiffies(30000);
4761         int rc;
4762
4763         adapter->from_passive_init = false;
4764
4765         adapter->init_done_rc = 0;
4766         ibmvnic_send_crq_init(adapter);
4767         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4768                 dev_err(dev, "Initialization sequence timed out\n");
4769                 return -1;
4770         }
4771
4772         if (adapter->init_done_rc) {
4773                 release_crq_queue(adapter);
4774                 return adapter->init_done_rc;
4775         }
4776
4777         if (adapter->from_passive_init) {
4778                 adapter->state = VNIC_OPEN;
4779                 adapter->from_passive_init = false;
4780                 return -1;
4781         }
4782
4783         rc = init_sub_crqs(adapter);
4784         if (rc) {
4785                 dev_err(dev, "Initialization of sub crqs failed\n");
4786                 release_crq_queue(adapter);
4787                 return rc;
4788         }
4789
4790         rc = init_sub_crq_irqs(adapter);
4791         if (rc) {
4792                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4793                 release_crq_queue(adapter);
4794         }
4795
4796         return rc;
4797 }
4798
4799 static struct device_attribute dev_attr_failover;
4800
4801 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4802 {
4803         struct ibmvnic_adapter *adapter;
4804         struct net_device *netdev;
4805         unsigned char *mac_addr_p;
4806         int rc;
4807
4808         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4809                 dev->unit_address);
4810
4811         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4812                                                         VETH_MAC_ADDR, NULL);
4813         if (!mac_addr_p) {
4814                 dev_err(&dev->dev,
4815                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4816                         __FILE__, __LINE__);
4817                 return 0;
4818         }
4819
4820         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4821                                    IBMVNIC_MAX_QUEUES);
4822         if (!netdev)
4823                 return -ENOMEM;
4824
4825         adapter = netdev_priv(netdev);
4826         adapter->state = VNIC_PROBING;
4827         dev_set_drvdata(&dev->dev, netdev);
4828         adapter->vdev = dev;
4829         adapter->netdev = netdev;
4830
4831         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4832         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4833         netdev->irq = dev->irq;
4834         netdev->netdev_ops = &ibmvnic_netdev_ops;
4835         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4836         SET_NETDEV_DEV(netdev, &dev->dev);
4837
4838         spin_lock_init(&adapter->stats_lock);
4839
4840         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4841         INIT_LIST_HEAD(&adapter->rwi_list);
4842         spin_lock_init(&adapter->rwi_lock);
4843         init_completion(&adapter->init_done);
4844         adapter->resetting = false;
4845
4846         do {
4847                 rc = init_crq_queue(adapter);
4848                 if (rc) {
4849                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4850                                 rc);
4851                         goto ibmvnic_init_fail;
4852                 }
4853
4854                 rc = ibmvnic_init(adapter);
4855                 if (rc && rc != EAGAIN)
4856                         goto ibmvnic_init_fail;
4857         } while (rc == EAGAIN);
4858
4859         rc = init_stats_buffers(adapter);
4860         if (rc)
4861                 goto ibmvnic_init_fail;
4862
4863         rc = init_stats_token(adapter);
4864         if (rc)
4865                 goto ibmvnic_stats_fail;
4866
4867         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4868         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4869         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4870
4871         rc = device_create_file(&dev->dev, &dev_attr_failover);
4872         if (rc)
4873                 goto ibmvnic_dev_file_err;
4874
4875         netif_carrier_off(netdev);
4876         rc = register_netdev(netdev);
4877         if (rc) {
4878                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4879                 goto ibmvnic_register_fail;
4880         }
4881         dev_info(&dev->dev, "ibmvnic registered\n");
4882
4883         adapter->state = VNIC_PROBED;
4884
4885         adapter->wait_for_reset = false;
4886
4887         return 0;
4888
4889 ibmvnic_register_fail:
4890         device_remove_file(&dev->dev, &dev_attr_failover);
4891
4892 ibmvnic_dev_file_err:
4893         release_stats_token(adapter);
4894
4895 ibmvnic_stats_fail:
4896         release_stats_buffers(adapter);
4897
4898 ibmvnic_init_fail:
4899         release_sub_crqs(adapter, 1);
4900         release_crq_queue(adapter);
4901         free_netdev(netdev);
4902
4903         return rc;
4904 }
4905
4906 static int ibmvnic_remove(struct vio_dev *dev)
4907 {
4908         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4909         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4910
4911         adapter->state = VNIC_REMOVING;
4912         rtnl_lock();
4913         unregister_netdevice(netdev);
4914
4915         release_resources(adapter);
4916         release_sub_crqs(adapter, 1);
4917         release_crq_queue(adapter);
4918
4919         release_stats_token(adapter);
4920         release_stats_buffers(adapter);
4921
4922         adapter->state = VNIC_REMOVED;
4923
4924         rtnl_unlock();
4925         device_remove_file(&dev->dev, &dev_attr_failover);
4926         free_netdev(netdev);
4927         dev_set_drvdata(&dev->dev, NULL);
4928
4929         return 0;
4930 }
4931
4932 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4933                               const char *buf, size_t count)
4934 {
4935         struct net_device *netdev = dev_get_drvdata(dev);
4936         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4937         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4938         __be64 session_token;
4939         long rc;
4940
4941         if (!sysfs_streq(buf, "1"))
4942                 return -EINVAL;
4943
4944         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4945                          H_GET_SESSION_TOKEN, 0, 0, 0);
4946         if (rc) {
4947                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4948                            rc);
4949                 return -EINVAL;
4950         }
4951
4952         session_token = (__be64)retbuf[0];
4953         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4954                    be64_to_cpu(session_token));
4955         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4956                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4957         if (rc) {
4958                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4959                            rc);
4960                 return -EINVAL;
4961         }
4962
4963         return count;
4964 }
4965
4966 static DEVICE_ATTR_WO(failover);
4967
4968 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4969 {
4970         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4971         struct ibmvnic_adapter *adapter;
4972         struct iommu_table *tbl;
4973         unsigned long ret = 0;
4974         int i;
4975
4976         tbl = get_iommu_table_base(&vdev->dev);
4977
4978         /* netdev inits at probe time along with the structures we need below*/
4979         if (!netdev)
4980                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4981
4982         adapter = netdev_priv(netdev);
4983
4984         ret += PAGE_SIZE; /* the crq message queue */
4985         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4986
4987         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4988                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4989
4990         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4991              i++)
4992                 ret += adapter->rx_pool[i].size *
4993                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4994
4995         return ret;
4996 }
4997
4998 static int ibmvnic_resume(struct device *dev)
4999 {
5000         struct net_device *netdev = dev_get_drvdata(dev);
5001         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5002
5003         if (adapter->state != VNIC_OPEN)
5004                 return 0;
5005
5006         tasklet_schedule(&adapter->tasklet);
5007
5008         return 0;
5009 }
5010
5011 static const struct vio_device_id ibmvnic_device_table[] = {
5012         {"network", "IBM,vnic"},
5013         {"", "" }
5014 };
5015 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5016
5017 static const struct dev_pm_ops ibmvnic_pm_ops = {
5018         .resume = ibmvnic_resume
5019 };
5020
5021 static struct vio_driver ibmvnic_driver = {
5022         .id_table       = ibmvnic_device_table,
5023         .probe          = ibmvnic_probe,
5024         .remove         = ibmvnic_remove,
5025         .get_desired_dma = ibmvnic_get_desired_dma,
5026         .name           = ibmvnic_driver_name,
5027         .pm             = &ibmvnic_pm_ops,
5028 };
5029
5030 /* module functions */
5031 static int __init ibmvnic_module_init(void)
5032 {
5033         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5034                 IBMVNIC_DRIVER_VERSION);
5035
5036         return vio_register_driver(&ibmvnic_driver);
5037 }
5038
5039 static void __exit ibmvnic_module_exit(void)
5040 {
5041         vio_unregister_driver(&ibmvnic_driver);
5042 }
5043
5044 module_init(ibmvnic_module_init);
5045 module_exit(ibmvnic_module_exit);