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                         alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 432                                              rx_pool->size *
 433                                              rx_pool->buff_size);
 434                 } else {
 435                         rc = reset_long_term_buff(adapter,
 436                                                   &rx_pool->long_term_buff);
 437                 }
 438
 439                 if (rc)
 440                         return rc;
 441
 442                 for (j = 0; j < rx_pool->size; j++)
 443                         rx_pool->free_map[j] = j;
 444
 445                 memset(rx_pool->rx_buff, 0,
 446                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 447
 448                 atomic_set(&rx_pool->available, 0);
 449                 rx_pool->next_alloc = 0;
 450                 rx_pool->next_free = 0;
 451                 rx_pool->active = 1;
 452         }
 453
 454         return 0;
 455 }
 456
 457 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 458 {
 459         struct ibmvnic_rx_pool *rx_pool;
 460         int i, j;
 461
 462         if (!adapter->rx_pool)
 463                 return;
 464
 465         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 466                 rx_pool = &adapter->rx_pool[i];
 467
 468                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 469
 470                 kfree(rx_pool->free_map);
 471                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 472
 473                 if (!rx_pool->rx_buff)
 474                         continue;
 475
 476                 for (j = 0; j < rx_pool->size; j++) {
 477                         if (rx_pool->rx_buff[j].skb) {
 478                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 479                                 rx_pool->rx_buff[j].skb = NULL;
 480                         }
 481                 }
 482
 483                 kfree(rx_pool->rx_buff);
 484         }
 485
 486         kfree(adapter->rx_pool);
 487         adapter->rx_pool = NULL;
 488         adapter->num_active_rx_pools = 0;
 489 }
 490
 491 static int init_rx_pools(struct net_device *netdev)
 492 {
 493         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 494         struct device *dev = &adapter->vdev->dev;
 495         struct ibmvnic_rx_pool *rx_pool;
 496         int rxadd_subcrqs;
 497         u64 *size_array;
 498         int i, j;
 499
 500         rxadd_subcrqs =
 501                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 502         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 503                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 504
 505         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 506                                    sizeof(struct ibmvnic_rx_pool),
 507                                    GFP_KERNEL);
 508         if (!adapter->rx_pool) {
 509                 dev_err(dev, "Failed to allocate rx pools\n");
 510                 return -1;
 511         }
 512
 513         adapter->num_active_rx_pools = rxadd_subcrqs;
 514
 515         for (i = 0; i < rxadd_subcrqs; i++) {
 516                 rx_pool = &adapter->rx_pool[i];
 517
 518                 netdev_dbg(adapter->netdev,
 519                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 520                            i, adapter->req_rx_add_entries_per_subcrq,
 521                            be64_to_cpu(size_array[i]));
 522
 523                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 524                 rx_pool->index = i;
 525                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 526                 rx_pool->active = 1;
 527
 528                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 529                                             GFP_KERNEL);
 530                 if (!rx_pool->free_map) {
 531                         release_rx_pools(adapter);
 532                         return -1;
 533                 }
 534
 535                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 536                                            sizeof(struct ibmvnic_rx_buff),
 537                                            GFP_KERNEL);
 538                 if (!rx_pool->rx_buff) {
 539                         dev_err(dev, "Couldn't alloc rx buffers\n");
 540                         release_rx_pools(adapter);
 541                         return -1;
 542                 }
 543
 544                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 545                                          rx_pool->size * rx_pool->buff_size)) {
 546                         release_rx_pools(adapter);
 547                         return -1;
 548                 }
 549
 550                 for (j = 0; j < rx_pool->size; ++j)
 551                         rx_pool->free_map[j] = j;
 552
 553                 atomic_set(&rx_pool->available, 0);
 554                 rx_pool->next_alloc = 0;
 555                 rx_pool->next_free = 0;
 556         }
 557
 558         return 0;
 559 }
 560
 561 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 562                              struct ibmvnic_tx_pool *tx_pool)
 563 {
 564         int rc, i;
 565
 566         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 567         if (rc)
 568                 return rc;
 569
 570         memset(tx_pool->tx_buff, 0,
 571                tx_pool->num_buffers *
 572                sizeof(struct ibmvnic_tx_buff));
 573
 574         for (i = 0; i < tx_pool->num_buffers; i++)
 575                 tx_pool->free_map[i] = i;
 576
 577         tx_pool->consumer_index = 0;
 578         tx_pool->producer_index = 0;
 579
 580         return 0;
 581 }
 582
 583 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 584 {
 585         int tx_scrqs;
 586         int i, rc;
 587
 588         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 589         for (i = 0; i < tx_scrqs; i++) {
 590                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 591                 if (rc)
 592                         return rc;
 593                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 594                 if (rc)
 595                         return rc;
 596         }
 597
 598         return 0;
 599 }
 600
 601 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 602 {
 603         if (!adapter->vpd)
 604                 return;
 605
 606         kfree(adapter->vpd->buff);
 607         kfree(adapter->vpd);
 608
 609         adapter->vpd = NULL;
 610 }
 611
 612 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 613                                 struct ibmvnic_tx_pool *tx_pool)
 614 {
 615         kfree(tx_pool->tx_buff);
 616         kfree(tx_pool->free_map);
 617         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 618 }
 619
 620 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 621 {
 622         int i;
 623
 624         if (!adapter->tx_pool)
 625                 return;
 626
 627         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 628                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 629                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 630         }
 631
 632         kfree(adapter->tx_pool);
 633         adapter->tx_pool = NULL;
 634         kfree(adapter->tso_pool);
 635         adapter->tso_pool = NULL;
 636         adapter->num_active_tx_pools = 0;
 637 }
 638
 639 static int init_one_tx_pool(struct net_device *netdev,
 640                             struct ibmvnic_tx_pool *tx_pool,
 641                             int num_entries, int buf_size)
 642 {
 643         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 644         int i;
 645
 646         tx_pool->tx_buff = kcalloc(num_entries,
 647                                    sizeof(struct ibmvnic_tx_buff),
 648                                    GFP_KERNEL);
 649         if (!tx_pool->tx_buff)
 650                 return -1;
 651
 652         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 653                                  num_entries * buf_size))
 654                 return -1;
 655
 656         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 657         if (!tx_pool->free_map)
 658                 return -1;
 659
 660         for (i = 0; i < num_entries; i++)
 661                 tx_pool->free_map[i] = i;
 662
 663         tx_pool->consumer_index = 0;
 664         tx_pool->producer_index = 0;
 665         tx_pool->num_buffers = num_entries;
 666         tx_pool->buf_size = buf_size;
 667
 668         return 0;
 669 }
 670
 671 static int init_tx_pools(struct net_device *netdev)
 672 {
 673         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 674         int tx_subcrqs;
 675         int i, rc;
 676
 677         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 678         adapter->tx_pool = kcalloc(tx_subcrqs,
 679                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 680         if (!adapter->tx_pool)
 681                 return -1;
 682
 683         adapter->tso_pool = kcalloc(tx_subcrqs,
 684                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 685         if (!adapter->tso_pool)
 686                 return -1;
 687
 688         adapter->num_active_tx_pools = tx_subcrqs;
 689
 690         for (i = 0; i < tx_subcrqs; i++) {
 691                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 692                                       adapter->req_tx_entries_per_subcrq,
 693                                       adapter->req_mtu + VLAN_HLEN);
 694                 if (rc) {
 695                         release_tx_pools(adapter);
 696                         return rc;
 697                 }
 698
 699                 init_one_tx_pool(netdev, &adapter->tso_pool[i],
 700                                  IBMVNIC_TSO_BUFS,
 701                                  IBMVNIC_TSO_BUF_SZ);
 702                 if (rc) {
 703                         release_tx_pools(adapter);
 704                         return rc;
 705                 }
 706         }
 707
 708         return 0;
 709 }
 710
 711 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 712 {
 713         int i;
 714
 715         if (adapter->napi_enabled)
 716                 return;
 717
 718         for (i = 0; i < adapter->req_rx_queues; i++)
 719                 napi_enable(&adapter->napi[i]);
 720
 721         adapter->napi_enabled = true;
 722 }
 723
 724 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 725 {
 726         int i;
 727
 728         if (!adapter->napi_enabled)
 729                 return;
 730
 731         for (i = 0; i < adapter->req_rx_queues; i++) {
 732                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 733                 napi_disable(&adapter->napi[i]);
 734         }
 735
 736         adapter->napi_enabled = false;
 737 }
 738
 739 static int init_napi(struct ibmvnic_adapter *adapter)
 740 {
 741         int i;
 742
 743         adapter->napi = kcalloc(adapter->req_rx_queues,
 744                                 sizeof(struct napi_struct), GFP_KERNEL);
 745         if (!adapter->napi)
 746                 return -ENOMEM;
 747
 748         for (i = 0; i < adapter->req_rx_queues; i++) {
 749                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 750                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 751                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 752         }
 753
 754         adapter->num_active_rx_napi = adapter->req_rx_queues;
 755         return 0;
 756 }
 757
 758 static void release_napi(struct ibmvnic_adapter *adapter)
 759 {
 760         int i;
 761
 762         if (!adapter->napi)
 763                 return;
 764
 765         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 766                 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
 767                 netif_napi_del(&adapter->napi[i]);
 768         }
 769
 770         kfree(adapter->napi);
 771         adapter->napi = NULL;
 772         adapter->num_active_rx_napi = 0;
 773         adapter->napi_enabled = false;
 774 }
 775
 776 static int ibmvnic_login(struct net_device *netdev)
 777 {
 778         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 779         unsigned long timeout = msecs_to_jiffies(30000);
 780         int retry_count = 0;
 781         bool retry;
 782         int rc;
 783
 784         do {
 785                 retry = false;
 786                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 787                         netdev_warn(netdev, "Login attempts exceeded\n");
 788                         return -1;
 789                 }
 790
 791                 adapter->init_done_rc = 0;
 792                 reinit_completion(&adapter->init_done);
 793                 rc = send_login(adapter);
 794                 if (rc) {
 795                         netdev_warn(netdev, "Unable to login\n");
 796                         return rc;
 797                 }
 798
 799                 if (!wait_for_completion_timeout(&adapter->init_done,
 800                                                  timeout)) {
 801                         netdev_warn(netdev, "Login timed out\n");
 802                         return -1;
 803                 }
 804
 805                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 806                         retry_count++;
 807                         release_sub_crqs(adapter, 1);
 808
 809                         retry = true;
 810                         netdev_dbg(netdev,
 811                                    "Received partial success, retrying...\n");
 812                         adapter->init_done_rc = 0;
 813                         reinit_completion(&adapter->init_done);
 814                         send_cap_queries(adapter);
 815                         if (!wait_for_completion_timeout(&adapter->init_done,
 816                                                          timeout)) {
 817                                 netdev_warn(netdev,
 818                                             "Capabilities query timed out\n");
 819                                 return -1;
 820                         }
 821
 822                         rc = init_sub_crqs(adapter);
 823                         if (rc) {
 824                                 netdev_warn(netdev,
 825                                             "SCRQ initialization failed\n");
 826                                 return -1;
 827                         }
 828
 829                         rc = init_sub_crq_irqs(adapter);
 830                         if (rc) {
 831                                 netdev_warn(netdev,
 832                                             "SCRQ irq initialization failed\n");
 833                                 return -1;
 834                         }
 835                 } else if (adapter->init_done_rc) {
 836                         netdev_warn(netdev, "Adapter login failed\n");
 837                         return -1;
 838                 }
 839         } while (retry);
 840
 841         __ibmvnic_set_mac(netdev, adapter->mac_addr);
 842
 843         return 0;
 844 }
 845
 846 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 847 {
 848         kfree(adapter->login_buf);
 849         adapter->login_buf = NULL;
 850 }
 851
 852 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 853 {
 854         kfree(adapter->login_rsp_buf);
 855         adapter->login_rsp_buf = NULL;
 856 }
 857
 858 static void release_resources(struct ibmvnic_adapter *adapter)
 859 {
 860         release_vpd_data(adapter);
 861
 862         release_tx_pools(adapter);
 863         release_rx_pools(adapter);
 864
 865         release_napi(adapter);
 866         release_login_rsp_buffer(adapter);
 867 }
 868
 869 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 870 {
 871         struct net_device *netdev = adapter->netdev;
 872         unsigned long timeout = msecs_to_jiffies(30000);
 873         union ibmvnic_crq crq;
 874         bool resend;
 875         int rc;
 876
 877         netdev_dbg(netdev, "setting link state %d\n", link_state);
 878
 879         memset(&crq, 0, sizeof(crq));
 880         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 881         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 882         crq.logical_link_state.link_state = link_state;
 883
 884         do {
 885                 resend = false;
 886
 887                 reinit_completion(&adapter->init_done);
 888                 rc = ibmvnic_send_crq(adapter, &crq);
 889                 if (rc) {
 890                         netdev_err(netdev, "Failed to set link state\n");
 891                         return rc;
 892                 }
 893
 894                 if (!wait_for_completion_timeout(&adapter->init_done,
 895                                                  timeout)) {
 896                         netdev_err(netdev, "timeout setting link state\n");
 897                         return -1;
 898                 }
 899
 900                 if (adapter->init_done_rc == 1) {
 901                         /* Partuial success, delay and re-send */
 902                         mdelay(1000);
 903                         resend = true;
 904                 } else if (adapter->init_done_rc) {
 905                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 906                                     adapter->init_done_rc);
 907                         return adapter->init_done_rc;
 908                 }
 909         } while (resend);
 910
 911         return 0;
 912 }
 913
 914 static int set_real_num_queues(struct net_device *netdev)
 915 {
 916         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 917         int rc;
 918
 919         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 920                    adapter->req_tx_queues, adapter->req_rx_queues);
 921
 922         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 923         if (rc) {
 924                 netdev_err(netdev, "failed to set the number of tx queues\n");
 925                 return rc;
 926         }
 927
 928         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 929         if (rc)
 930                 netdev_err(netdev, "failed to set the number of rx queues\n");
 931
 932         return rc;
 933 }
 934
 935 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 936 {
 937         struct device *dev = &adapter->vdev->dev;
 938         union ibmvnic_crq crq;
 939         int len = 0;
 940         int rc;
 941
 942         if (adapter->vpd->buff)
 943                 len = adapter->vpd->len;
 944
 945         init_completion(&adapter->fw_done);
 946         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 947         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 948         rc = ibmvnic_send_crq(adapter, &crq);
 949         if (rc)
 950                 return rc;
 951         wait_for_completion(&adapter->fw_done);
 952
 953         if (!adapter->vpd->len)
 954                 return -ENODATA;
 955
 956         if (!adapter->vpd->buff)
 957                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 958         else if (adapter->vpd->len != len)
 959                 adapter->vpd->buff =
 960                         krealloc(adapter->vpd->buff,
 961                                  adapter->vpd->len, GFP_KERNEL);
 962
 963         if (!adapter->vpd->buff) {
 964                 dev_err(dev, "Could allocate VPD buffer\n");
 965                 return -ENOMEM;
 966         }
 967
 968         adapter->vpd->dma_addr =
 969                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 970                                DMA_FROM_DEVICE);
 971         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 972                 dev_err(dev, "Could not map VPD buffer\n");
 973                 kfree(adapter->vpd->buff);
 974                 adapter->vpd->buff = NULL;
 975                 return -ENOMEM;
 976         }
 977
 978         reinit_completion(&adapter->fw_done);
 979         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 980         crq.get_vpd.cmd = GET_VPD;
 981         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 982         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 983         rc = ibmvnic_send_crq(adapter, &crq);
 984         if (rc) {
 985                 kfree(adapter->vpd->buff);
 986                 adapter->vpd->buff = NULL;
 987                 return rc;
 988         }
 989         wait_for_completion(&adapter->fw_done);
 990
 991         return 0;
 992 }
 993
 994 static int init_resources(struct ibmvnic_adapter *adapter)
 995 {
 996         struct net_device *netdev = adapter->netdev;
 997         int rc;
 998
 999         rc = set_real_num_queues(netdev);
1000         if (rc)
1001                 return rc;
1002
1003         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1004         if (!adapter->vpd)
1005                 return -ENOMEM;
1006
1007         /* Vital Product Data (VPD) */
1008         rc = ibmvnic_get_vpd(adapter);
1009         if (rc) {
1010                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1011                 return rc;
1012         }
1013
1014         adapter->map_id = 1;
1015
1016         rc = init_napi(adapter);
1017         if (rc)
1018                 return rc;
1019
1020         send_map_query(adapter);
1021
1022         rc = init_rx_pools(netdev);
1023         if (rc)
1024                 return rc;
1025
1026         rc = init_tx_pools(netdev);
1027         return rc;
1028 }
1029
1030 static int __ibmvnic_open(struct net_device *netdev)
1031 {
1032         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1033         enum vnic_state prev_state = adapter->state;
1034         int i, rc;
1035
1036         adapter->state = VNIC_OPENING;
1037         replenish_pools(adapter);
1038         ibmvnic_napi_enable(adapter);
1039
1040         /* We're ready to receive frames, enable the sub-crq interrupts and
1041          * set the logical link state to up
1042          */
1043         for (i = 0; i < adapter->req_rx_queues; i++) {
1044                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1045                 if (prev_state == VNIC_CLOSED)
1046                         enable_irq(adapter->rx_scrq[i]->irq);
1047                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1048         }
1049
1050         for (i = 0; i < adapter->req_tx_queues; i++) {
1051                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1052                 if (prev_state == VNIC_CLOSED)
1053                         enable_irq(adapter->tx_scrq[i]->irq);
1054                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1055         }
1056
1057         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1058         if (rc) {
1059                 for (i = 0; i < adapter->req_rx_queues; i++)
1060                         napi_disable(&adapter->napi[i]);
1061                 release_resources(adapter);
1062                 return rc;
1063         }
1064
1065         netif_tx_start_all_queues(netdev);
1066
1067         if (prev_state == VNIC_CLOSED) {
1068                 for (i = 0; i < adapter->req_rx_queues; i++)
1069                         napi_schedule(&adapter->napi[i]);
1070         }
1071
1072         adapter->state = VNIC_OPEN;
1073         return rc;
1074 }
1075
1076 static int ibmvnic_open(struct net_device *netdev)
1077 {
1078         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1079         int rc;
1080
1081         /* If device failover is pending, just set device state and return.
1082          * Device operation will be handled by reset routine.
1083          */
1084         if (adapter->failover_pending) {
1085                 adapter->state = VNIC_OPEN;
1086                 return 0;
1087         }
1088
1089         if (adapter->state != VNIC_CLOSED) {
1090                 rc = ibmvnic_login(netdev);
1091                 if (rc)
1092                         return rc;
1093
1094                 rc = init_resources(adapter);
1095                 if (rc) {
1096                         netdev_err(netdev, "failed to initialize resources\n");
1097                         release_resources(adapter);
1098                         return rc;
1099                 }
1100         }
1101
1102         rc = __ibmvnic_open(netdev);
1103
1104         return rc;
1105 }
1106
1107 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1108 {
1109         struct ibmvnic_rx_pool *rx_pool;
1110         struct ibmvnic_rx_buff *rx_buff;
1111         u64 rx_entries;
1112         int rx_scrqs;
1113         int i, j;
1114
1115         if (!adapter->rx_pool)
1116                 return;
1117
1118         rx_scrqs = adapter->num_active_rx_pools;
1119         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1120
1121         /* Free any remaining skbs in the rx buffer pools */
1122         for (i = 0; i < rx_scrqs; i++) {
1123                 rx_pool = &adapter->rx_pool[i];
1124                 if (!rx_pool || !rx_pool->rx_buff)
1125                         continue;
1126
1127                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1128                 for (j = 0; j < rx_entries; j++) {
1129                         rx_buff = &rx_pool->rx_buff[j];
1130                         if (rx_buff && rx_buff->skb) {
1131                                 dev_kfree_skb_any(rx_buff->skb);
1132                                 rx_buff->skb = NULL;
1133                         }
1134                 }
1135         }
1136 }
1137
1138 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1139                               struct ibmvnic_tx_pool *tx_pool)
1140 {
1141         struct ibmvnic_tx_buff *tx_buff;
1142         u64 tx_entries;
1143         int i;
1144
1145         if (!tx_pool || !tx_pool->tx_buff)
1146                 return;
1147
1148         tx_entries = tx_pool->num_buffers;
1149
1150         for (i = 0; i < tx_entries; i++) {
1151                 tx_buff = &tx_pool->tx_buff[i];
1152                 if (tx_buff && tx_buff->skb) {
1153                         dev_kfree_skb_any(tx_buff->skb);
1154                         tx_buff->skb = NULL;
1155                 }
1156         }
1157 }
1158
1159 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1160 {
1161         int tx_scrqs;
1162         int i;
1163
1164         if (!adapter->tx_pool || !adapter->tso_pool)
1165                 return;
1166
1167         tx_scrqs = adapter->num_active_tx_pools;
1168
1169         /* Free any remaining skbs in the tx buffer pools */
1170         for (i = 0; i < tx_scrqs; i++) {
1171                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1172                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1173                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1174         }
1175 }
1176
1177 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1178 {
1179         struct net_device *netdev = adapter->netdev;
1180         int i;
1181
1182         if (adapter->tx_scrq) {
1183                 for (i = 0; i < adapter->req_tx_queues; i++)
1184                         if (adapter->tx_scrq[i]->irq) {
1185                                 netdev_dbg(netdev,
1186                                            "Disabling tx_scrq[%d] irq\n", i);
1187                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1188                                 disable_irq(adapter->tx_scrq[i]->irq);
1189                         }
1190         }
1191
1192         if (adapter->rx_scrq) {
1193                 for (i = 0; i < adapter->req_rx_queues; i++) {
1194                         if (adapter->rx_scrq[i]->irq) {
1195                                 netdev_dbg(netdev,
1196                                            "Disabling rx_scrq[%d] irq\n", i);
1197                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1198                                 disable_irq(adapter->rx_scrq[i]->irq);
1199                         }
1200                 }
1201         }
1202 }
1203
1204 static void ibmvnic_cleanup(struct net_device *netdev)
1205 {
1206         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1207
1208         /* ensure that transmissions are stopped if called by do_reset */
1209         if (adapter->resetting)
1210                 netif_tx_disable(netdev);
1211         else
1212                 netif_tx_stop_all_queues(netdev);
1213
1214         ibmvnic_napi_disable(adapter);
1215         ibmvnic_disable_irqs(adapter);
1216
1217         clean_rx_pools(adapter);
1218         clean_tx_pools(adapter);
1219 }
1220
1221 static int __ibmvnic_close(struct net_device *netdev)
1222 {
1223         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1224         int rc = 0;
1225
1226         adapter->state = VNIC_CLOSING;
1227         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1228         if (rc)
1229                 return rc;
1230         adapter->state = VNIC_CLOSED;
1231         return 0;
1232 }
1233
1234 static int ibmvnic_close(struct net_device *netdev)
1235 {
1236         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1237         int rc;
1238
1239         /* If device failover is pending, just set device state and return.
1240          * Device operation will be handled by reset routine.
1241          */
1242         if (adapter->failover_pending) {
1243                 adapter->state = VNIC_CLOSED;
1244                 return 0;
1245         }
1246
1247         rc = __ibmvnic_close(netdev);
1248         ibmvnic_cleanup(netdev);
1249
1250         return rc;
1251 }
1252
1253 /**
1254  * build_hdr_data - creates L2/L3/L4 header data buffer
1255  * @hdr_field - bitfield determining needed headers
1256  * @skb - socket buffer
1257  * @hdr_len - array of header lengths
1258  * @tot_len - total length of data
1259  *
1260  * Reads hdr_field to determine which headers are needed by firmware.
1261  * Builds a buffer containing these headers.  Saves individual header
1262  * lengths and total buffer length to be used to build descriptors.
1263  */
1264 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1265                           int *hdr_len, u8 *hdr_data)
1266 {
1267         int len = 0;
1268         u8 *hdr;
1269
1270         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1271                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1272         else
1273                 hdr_len[0] = sizeof(struct ethhdr);
1274
1275         if (skb->protocol == htons(ETH_P_IP)) {
1276                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1277                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1278                         hdr_len[2] = tcp_hdrlen(skb);
1279                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1280                         hdr_len[2] = sizeof(struct udphdr);
1281         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1282                 hdr_len[1] = sizeof(struct ipv6hdr);
1283                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1284                         hdr_len[2] = tcp_hdrlen(skb);
1285                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1286                         hdr_len[2] = sizeof(struct udphdr);
1287         } else if (skb->protocol == htons(ETH_P_ARP)) {
1288                 hdr_len[1] = arp_hdr_len(skb->dev);
1289                 hdr_len[2] = 0;
1290         }
1291
1292         memset(hdr_data, 0, 120);
1293         if ((hdr_field >> 6) & 1) {
1294                 hdr = skb_mac_header(skb);
1295                 memcpy(hdr_data, hdr, hdr_len[0]);
1296                 len += hdr_len[0];
1297         }
1298
1299         if ((hdr_field >> 5) & 1) {
1300                 hdr = skb_network_header(skb);
1301                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1302                 len += hdr_len[1];
1303         }
1304
1305         if ((hdr_field >> 4) & 1) {
1306                 hdr = skb_transport_header(skb);
1307                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1308                 len += hdr_len[2];
1309         }
1310         return len;
1311 }
1312
1313 /**
1314  * create_hdr_descs - create header and header extension descriptors
1315  * @hdr_field - bitfield determining needed headers
1316  * @data - buffer containing header data
1317  * @len - length of data buffer
1318  * @hdr_len - array of individual header lengths
1319  * @scrq_arr - descriptor array
1320  *
1321  * Creates header and, if needed, header extension descriptors and
1322  * places them in a descriptor array, scrq_arr
1323  */
1324
1325 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1326                             union sub_crq *scrq_arr)
1327 {
1328         union sub_crq hdr_desc;
1329         int tmp_len = len;
1330         int num_descs = 0;
1331         u8 *data, *cur;
1332         int tmp;
1333
1334         while (tmp_len > 0) {
1335                 cur = hdr_data + len - tmp_len;
1336
1337                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1338                 if (cur != hdr_data) {
1339                         data = hdr_desc.hdr_ext.data;
1340                         tmp = tmp_len > 29 ? 29 : tmp_len;
1341                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1342                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1343                         hdr_desc.hdr_ext.len = tmp;
1344                 } else {
1345                         data = hdr_desc.hdr.data;
1346                         tmp = tmp_len > 24 ? 24 : tmp_len;
1347                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1348                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1349                         hdr_desc.hdr.len = tmp;
1350                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1351                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1352                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1353                         hdr_desc.hdr.flag = hdr_field << 1;
1354                 }
1355                 memcpy(data, cur, tmp);
1356                 tmp_len -= tmp;
1357                 *scrq_arr = hdr_desc;
1358                 scrq_arr++;
1359                 num_descs++;
1360         }
1361
1362         return num_descs;
1363 }
1364
1365 /**
1366  * build_hdr_descs_arr - build a header descriptor array
1367  * @skb - socket buffer
1368  * @num_entries - number of descriptors to be sent
1369  * @subcrq - first TX descriptor
1370  * @hdr_field - bit field determining which headers will be sent
1371  *
1372  * This function will build a TX descriptor array with applicable
1373  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1374  */
1375
1376 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1377                                 int *num_entries, u8 hdr_field)
1378 {
1379         int hdr_len[3] = {0, 0, 0};
1380         int tot_len;
1381         u8 *hdr_data = txbuff->hdr_data;
1382
1383         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1384                                  txbuff->hdr_data);
1385         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1386                          txbuff->indir_arr + 1);
1387 }
1388
1389 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1390                                     struct net_device *netdev)
1391 {
1392         /* For some backing devices, mishandling of small packets
1393          * can result in a loss of connection or TX stall. Device
1394          * architects recommend that no packet should be smaller
1395          * than the minimum MTU value provided to the driver, so
1396          * pad any packets to that length
1397          */
1398         if (skb->len < netdev->min_mtu)
1399                 return skb_put_padto(skb, netdev->min_mtu);
1400
1401         return 0;
1402 }
1403
1404 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1405 {
1406         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1407         int queue_num = skb_get_queue_mapping(skb);
1408         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1409         struct device *dev = &adapter->vdev->dev;
1410         struct ibmvnic_tx_buff *tx_buff = NULL;
1411         struct ibmvnic_sub_crq_queue *tx_scrq;
1412         struct ibmvnic_tx_pool *tx_pool;
1413         unsigned int tx_send_failed = 0;
1414         unsigned int tx_map_failed = 0;
1415         unsigned int tx_dropped = 0;
1416         unsigned int tx_packets = 0;
1417         unsigned int tx_bytes = 0;
1418         dma_addr_t data_dma_addr;
1419         struct netdev_queue *txq;
1420         unsigned long lpar_rc;
1421         union sub_crq tx_crq;
1422         unsigned int offset;
1423         int num_entries = 1;
1424         unsigned char *dst;
1425         u64 *handle_array;
1426         int index = 0;
1427         u8 proto = 0;
1428         netdev_tx_t ret = NETDEV_TX_OK;
1429
1430         if (adapter->resetting) {
1431                 if (!netif_subqueue_stopped(netdev, skb))
1432                         netif_stop_subqueue(netdev, queue_num);
1433                 dev_kfree_skb_any(skb);
1434
1435                 tx_send_failed++;
1436                 tx_dropped++;
1437                 ret = NETDEV_TX_OK;
1438                 goto out;
1439         }
1440
1441         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1442                 tx_dropped++;
1443                 tx_send_failed++;
1444                 ret = NETDEV_TX_OK;
1445                 goto out;
1446         }
1447         if (skb_is_gso(skb))
1448                 tx_pool = &adapter->tso_pool[queue_num];
1449         else
1450                 tx_pool = &adapter->tx_pool[queue_num];
1451
1452         tx_scrq = adapter->tx_scrq[queue_num];
1453         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1454         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1455                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1456
1457         index = tx_pool->free_map[tx_pool->consumer_index];
1458
1459         if (index == IBMVNIC_INVALID_MAP) {
1460                 dev_kfree_skb_any(skb);
1461                 tx_send_failed++;
1462                 tx_dropped++;
1463                 ret = NETDEV_TX_OK;
1464                 goto out;
1465         }
1466
1467         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1468
1469         offset = index * tx_pool->buf_size;
1470         dst = tx_pool->long_term_buff.buff + offset;
1471         memset(dst, 0, tx_pool->buf_size);
1472         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1473
1474         if (skb_shinfo(skb)->nr_frags) {
1475                 int cur, i;
1476
1477                 /* Copy the head */
1478                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1479                 cur = skb_headlen(skb);
1480
1481                 /* Copy the frags */
1482                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1483                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1484
1485                         memcpy(dst + cur,
1486                                page_address(skb_frag_page(frag)) +
1487                                frag->page_offset, skb_frag_size(frag));
1488                         cur += skb_frag_size(frag);
1489                 }
1490         } else {
1491                 skb_copy_from_linear_data(skb, dst, skb->len);
1492         }
1493
1494         tx_pool->consumer_index =
1495             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1496
1497         tx_buff = &tx_pool->tx_buff[index];
1498         tx_buff->skb = skb;
1499         tx_buff->data_dma[0] = data_dma_addr;
1500         tx_buff->data_len[0] = skb->len;
1501         tx_buff->index = index;
1502         tx_buff->pool_index = queue_num;
1503         tx_buff->last_frag = true;
1504
1505         memset(&tx_crq, 0, sizeof(tx_crq));
1506         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1507         tx_crq.v1.type = IBMVNIC_TX_DESC;
1508         tx_crq.v1.n_crq_elem = 1;
1509         tx_crq.v1.n_sge = 1;
1510         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1511
1512         if (skb_is_gso(skb))
1513                 tx_crq.v1.correlator =
1514                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1515         else
1516                 tx_crq.v1.correlator = cpu_to_be32(index);
1517         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1518         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1519         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1520
1521         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1522                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1523                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1524         }
1525
1526         if (skb->protocol == htons(ETH_P_IP)) {
1527                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1528                 proto = ip_hdr(skb)->protocol;
1529         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1530                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1531                 proto = ipv6_hdr(skb)->nexthdr;
1532         }
1533
1534         if (proto == IPPROTO_TCP)
1535                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1536         else if (proto == IPPROTO_UDP)
1537                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1538
1539         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1540                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1541                 hdrs += 2;
1542         }
1543         if (skb_is_gso(skb)) {
1544                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1545                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1546                 hdrs += 2;
1547         }
1548         /* determine if l2/3/4 headers are sent to firmware */
1549         if ((*hdrs >> 7) & 1) {
1550                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1551                 tx_crq.v1.n_crq_elem = num_entries;
1552                 tx_buff->num_entries = num_entries;
1553                 tx_buff->indir_arr[0] = tx_crq;
1554                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1555                                                     sizeof(tx_buff->indir_arr),
1556                                                     DMA_TO_DEVICE);
1557                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1558                         dev_kfree_skb_any(skb);
1559                         tx_buff->skb = NULL;
1560                         if (!firmware_has_feature(FW_FEATURE_CMO))
1561                                 dev_err(dev, "tx: unable to map descriptor array\n");
1562                         tx_map_failed++;
1563                         tx_dropped++;
1564                         ret = NETDEV_TX_OK;
1565                         goto tx_err_out;
1566                 }
1567                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1568                                                (u64)tx_buff->indir_dma,
1569                                                (u64)num_entries);
1570         } else {
1571                 tx_buff->num_entries = num_entries;
1572                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1573                                       &tx_crq);
1574         }
1575         if (lpar_rc != H_SUCCESS) {
1576                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1577                         dev_err_ratelimited(dev, "tx: send failed\n");
1578                 dev_kfree_skb_any(skb);
1579                 tx_buff->skb = NULL;
1580
1581                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1582                         /* Disable TX and report carrier off if queue is closed
1583                          * or pending failover.
1584                          * Firmware guarantees that a signal will be sent to the
1585                          * driver, triggering a reset or some other action.
1586                          */
1587                         netif_tx_stop_all_queues(netdev);
1588                         netif_carrier_off(netdev);
1589                 }
1590
1591                 tx_send_failed++;
1592                 tx_dropped++;
1593                 ret = NETDEV_TX_OK;
1594                 goto tx_err_out;
1595         }
1596
1597         if (atomic_add_return(num_entries, &tx_scrq->used)
1598                                         >= adapter->req_tx_entries_per_subcrq) {
1599                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1600                 netif_stop_subqueue(netdev, queue_num);
1601         }
1602
1603         tx_packets++;
1604         tx_bytes += skb->len;
1605         txq->trans_start = jiffies;
1606         ret = NETDEV_TX_OK;
1607         goto out;
1608
1609 tx_err_out:
1610         /* roll back consumer index and map array*/
1611         if (tx_pool->consumer_index == 0)
1612                 tx_pool->consumer_index =
1613                         tx_pool->num_buffers - 1;
1614         else
1615                 tx_pool->consumer_index--;
1616         tx_pool->free_map[tx_pool->consumer_index] = index;
1617 out:
1618         netdev->stats.tx_dropped += tx_dropped;
1619         netdev->stats.tx_bytes += tx_bytes;
1620         netdev->stats.tx_packets += tx_packets;
1621         adapter->tx_send_failed += tx_send_failed;
1622         adapter->tx_map_failed += tx_map_failed;
1623         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1624         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1625         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1626
1627         return ret;
1628 }
1629
1630 static void ibmvnic_set_multi(struct net_device *netdev)
1631 {
1632         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1633         struct netdev_hw_addr *ha;
1634         union ibmvnic_crq crq;
1635
1636         memset(&crq, 0, sizeof(crq));
1637         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1638         crq.request_capability.cmd = REQUEST_CAPABILITY;
1639
1640         if (netdev->flags & IFF_PROMISC) {
1641                 if (!adapter->promisc_supported)
1642                         return;
1643         } else {
1644                 if (netdev->flags & IFF_ALLMULTI) {
1645                         /* Accept all multicast */
1646                         memset(&crq, 0, sizeof(crq));
1647                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1648                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1649                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1650                         ibmvnic_send_crq(adapter, &crq);
1651                 } else if (netdev_mc_empty(netdev)) {
1652                         /* Reject all multicast */
1653                         memset(&crq, 0, sizeof(crq));
1654                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1655                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1656                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1657                         ibmvnic_send_crq(adapter, &crq);
1658                 } else {
1659                         /* Accept one or more multicast(s) */
1660                         netdev_for_each_mc_addr(ha, netdev) {
1661                                 memset(&crq, 0, sizeof(crq));
1662                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1663                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1664                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1665                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1666                                                 ha->addr);
1667                                 ibmvnic_send_crq(adapter, &crq);
1668                         }
1669                 }
1670         }
1671 }
1672
1673 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1674 {
1675         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1676         union ibmvnic_crq crq;
1677         int rc;
1678
1679         if (!is_valid_ether_addr(dev_addr)) {
1680                 rc = -EADDRNOTAVAIL;
1681                 goto err;
1682         }
1683
1684         memset(&crq, 0, sizeof(crq));
1685         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1686         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1687         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1688
1689         init_completion(&adapter->fw_done);
1690         rc = ibmvnic_send_crq(adapter, &crq);
1691         if (rc) {
1692                 rc = -EIO;
1693                 goto err;
1694         }
1695
1696         wait_for_completion(&adapter->fw_done);
1697         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1698         if (adapter->fw_done_rc) {
1699                 rc = -EIO;
1700                 goto err;
1701         }
1702
1703         return 0;
1704 err:
1705         ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1706         return rc;
1707 }
1708
1709 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1710 {
1711         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1712         struct sockaddr *addr = p;
1713         int rc;
1714
1715         rc = 0;
1716         ether_addr_copy(adapter->mac_addr, addr->sa_data);
1717         if (adapter->state != VNIC_PROBED)
1718                 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1719
1720         return rc;
1721 }
1722
1723 /**
1724  * do_reset returns zero if we are able to keep processing reset events, or
1725  * non-zero if we hit a fatal error and must halt.
1726  */
1727 static int do_reset(struct ibmvnic_adapter *adapter,
1728                     struct ibmvnic_rwi *rwi, u32 reset_state)
1729 {
1730         u64 old_num_rx_queues, old_num_tx_queues;
1731         u64 old_num_rx_slots, old_num_tx_slots;
1732         struct net_device *netdev = adapter->netdev;
1733         int i, rc;
1734
1735         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1736                    rwi->reset_reason);
1737
1738         netif_carrier_off(netdev);
1739         adapter->reset_reason = rwi->reset_reason;
1740
1741         old_num_rx_queues = adapter->req_rx_queues;
1742         old_num_tx_queues = adapter->req_tx_queues;
1743         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1744         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1745
1746         ibmvnic_cleanup(netdev);
1747
1748         if (reset_state == VNIC_OPEN &&
1749             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1750             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1751                 rc = __ibmvnic_close(netdev);
1752                 if (rc)
1753                         return rc;
1754         }
1755
1756         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1757             adapter->wait_for_reset) {
1758                 release_resources(adapter);
1759                 release_sub_crqs(adapter, 1);
1760                 release_crq_queue(adapter);
1761         }
1762
1763         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1764                 /* remove the closed state so when we call open it appears
1765                  * we are coming from the probed state.
1766                  */
1767                 adapter->state = VNIC_PROBED;
1768
1769                 if (adapter->wait_for_reset) {
1770                         rc = init_crq_queue(adapter);
1771                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1772                         rc = ibmvnic_reenable_crq_queue(adapter);
1773                         release_sub_crqs(adapter, 1);
1774                 } else {
1775                         rc = ibmvnic_reset_crq(adapter);
1776                         if (!rc)
1777                                 rc = vio_enable_interrupts(adapter->vdev);
1778                 }
1779
1780                 if (rc) {
1781                         netdev_err(adapter->netdev,
1782                                    "Couldn't initialize crq. rc=%d\n", rc);
1783                         return rc;
1784                 }
1785
1786                 rc = ibmvnic_reset_init(adapter);
1787                 if (rc)
1788                         return IBMVNIC_INIT_FAILED;
1789
1790                 /* If the adapter was in PROBE state prior to the reset,
1791                  * exit here.
1792                  */
1793                 if (reset_state == VNIC_PROBED)
1794                         return 0;
1795
1796                 rc = ibmvnic_login(netdev);
1797                 if (rc) {
1798                         adapter->state = reset_state;
1799                         return rc;
1800                 }
1801
1802                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1803                     adapter->wait_for_reset) {
1804                         rc = init_resources(adapter);
1805                         if (rc)
1806                                 return rc;
1807                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1808                            adapter->req_tx_queues != old_num_tx_queues ||
1809                            adapter->req_rx_add_entries_per_subcrq !=
1810                                                         old_num_rx_slots ||
1811                            adapter->req_tx_entries_per_subcrq !=
1812                                                         old_num_tx_slots) {
1813                         release_rx_pools(adapter);
1814                         release_tx_pools(adapter);
1815                         release_napi(adapter);
1816                         release_vpd_data(adapter);
1817
1818                         rc = init_resources(adapter);
1819                         if (rc)
1820                                 return rc;
1821
1822                 } else {
1823                         rc = reset_tx_pools(adapter);
1824                         if (rc)
1825                                 return rc;
1826
1827                         rc = reset_rx_pools(adapter);
1828                         if (rc)
1829                                 return rc;
1830                 }
1831                 ibmvnic_disable_irqs(adapter);
1832         }
1833         adapter->state = VNIC_CLOSED;
1834
1835         if (reset_state == VNIC_CLOSED)
1836                 return 0;
1837
1838         rc = __ibmvnic_open(netdev);
1839         if (rc) {
1840                 if (list_empty(&adapter->rwi_list))
1841                         adapter->state = VNIC_CLOSED;
1842                 else
1843                         adapter->state = reset_state;
1844
1845                 return 0;
1846         }
1847
1848         /* kick napi */
1849         for (i = 0; i < adapter->req_rx_queues; i++)
1850                 napi_schedule(&adapter->napi[i]);
1851
1852         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1853             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1854                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1855
1856         return 0;
1857 }
1858
1859 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1860                          struct ibmvnic_rwi *rwi, u32 reset_state)
1861 {
1862         struct net_device *netdev = adapter->netdev;
1863         int rc;
1864
1865         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1866                    rwi->reset_reason);
1867
1868         netif_carrier_off(netdev);
1869         adapter->reset_reason = rwi->reset_reason;
1870
1871         ibmvnic_cleanup(netdev);
1872         release_resources(adapter);
1873         release_sub_crqs(adapter, 0);
1874         release_crq_queue(adapter);
1875
1876         /* remove the closed state so when we call open it appears
1877          * we are coming from the probed state.
1878          */
1879         adapter->state = VNIC_PROBED;
1880
1881         reinit_completion(&adapter->init_done);
1882         rc = init_crq_queue(adapter);
1883         if (rc) {
1884                 netdev_err(adapter->netdev,
1885                            "Couldn't initialize crq. rc=%d\n", rc);
1886                 return rc;
1887         }
1888
1889         rc = ibmvnic_init(adapter);
1890         if (rc)
1891                 return rc;
1892
1893         /* If the adapter was in PROBE state prior to the reset,
1894          * exit here.
1895          */
1896         if (reset_state == VNIC_PROBED)
1897                 return 0;
1898
1899         rc = ibmvnic_login(netdev);
1900         if (rc) {
1901                 adapter->state = VNIC_PROBED;
1902                 return 0;
1903         }
1904
1905         rc = init_resources(adapter);
1906         if (rc)
1907                 return rc;
1908
1909         ibmvnic_disable_irqs(adapter);
1910         adapter->state = VNIC_CLOSED;
1911
1912         if (reset_state == VNIC_CLOSED)
1913                 return 0;
1914
1915         rc = __ibmvnic_open(netdev);
1916         if (rc) {
1917                 if (list_empty(&adapter->rwi_list))
1918                         adapter->state = VNIC_CLOSED;
1919                 else
1920                         adapter->state = reset_state;
1921
1922                 return 0;
1923         }
1924
1925         return 0;
1926 }
1927
1928 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1929 {
1930         struct ibmvnic_rwi *rwi;
1931         unsigned long flags;
1932
1933         spin_lock_irqsave(&adapter->rwi_lock, flags);
1934
1935         if (!list_empty(&adapter->rwi_list)) {
1936                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1937                                        list);
1938                 list_del(&rwi->list);
1939         } else {
1940                 rwi = NULL;
1941         }
1942
1943         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1944         return rwi;
1945 }
1946
1947 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1948 {
1949         struct ibmvnic_rwi *rwi;
1950
1951         rwi = get_next_rwi(adapter);
1952         while (rwi) {
1953                 kfree(rwi);
1954                 rwi = get_next_rwi(adapter);
1955         }
1956 }
1957
1958 static void __ibmvnic_reset(struct work_struct *work)
1959 {
1960         struct ibmvnic_rwi *rwi;
1961         struct ibmvnic_adapter *adapter;
1962         bool we_lock_rtnl = false;
1963         u32 reset_state;
1964         int rc = 0;
1965
1966         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1967
1968         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1969          * unless wait_for_reset is set, in which case the rtnl lock
1970          * has already been taken before initializing the reset
1971          */
1972         if (!adapter->wait_for_reset) {
1973                 rtnl_lock();
1974                 we_lock_rtnl = true;
1975         }
1976         reset_state = adapter->state;
1977
1978         rwi = get_next_rwi(adapter);
1979         while (rwi) {
1980                 if (adapter->force_reset_recovery) {
1981                         adapter->force_reset_recovery = false;
1982                         rc = do_hard_reset(adapter, rwi, reset_state);
1983                 } else {
1984                         rc = do_reset(adapter, rwi, reset_state);
1985                 }
1986                 kfree(rwi);
1987                 if (rc && rc != IBMVNIC_INIT_FAILED &&
1988                     !adapter->force_reset_recovery)
1989                         break;
1990
1991                 rwi = get_next_rwi(adapter);
1992         }
1993
1994         if (adapter->wait_for_reset) {
1995                 adapter->wait_for_reset = false;
1996                 adapter->reset_done_rc = rc;
1997                 complete(&adapter->reset_done);
1998         }
1999
2000         if (rc) {
2001                 netdev_dbg(adapter->netdev, "Reset failed\n");
2002                 free_all_rwi(adapter);
2003         }
2004
2005         adapter->resetting = false;
2006         if (we_lock_rtnl)
2007                 rtnl_unlock();
2008 }
2009
2010 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2011                          enum ibmvnic_reset_reason reason)
2012 {
2013         struct list_head *entry, *tmp_entry;
2014         struct ibmvnic_rwi *rwi, *tmp;
2015         struct net_device *netdev = adapter->netdev;
2016         unsigned long flags;
2017         int ret;
2018
2019         if (adapter->state == VNIC_REMOVING ||
2020             adapter->state == VNIC_REMOVED ||
2021             adapter->failover_pending) {
2022                 ret = EBUSY;
2023                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2024                 goto err;
2025         }
2026
2027         if (adapter->state == VNIC_PROBING) {
2028                 netdev_warn(netdev, "Adapter reset during probe\n");
2029                 ret = adapter->init_done_rc = EAGAIN;
2030                 goto err;
2031         }
2032
2033         spin_lock_irqsave(&adapter->rwi_lock, flags);
2034
2035         list_for_each(entry, &adapter->rwi_list) {
2036                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2037                 if (tmp->reset_reason == reason) {
2038                         netdev_dbg(netdev, "Skipping matching reset\n");
2039                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2040                         ret = EBUSY;
2041                         goto err;
2042                 }
2043         }
2044
2045         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2046         if (!rwi) {
2047                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2048                 ibmvnic_close(netdev);
2049                 ret = ENOMEM;
2050                 goto err;
2051         }
2052         /* if we just received a transport event,
2053          * flush reset queue and process this reset
2054          */
2055         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2056                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2057                         list_del(entry);
2058         }
2059         rwi->reset_reason = reason;
2060         list_add_tail(&rwi->list, &adapter->rwi_list);
2061         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2062         adapter->resetting = true;
2063         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2064         schedule_work(&adapter->ibmvnic_reset);
2065
2066         return 0;
2067 err:
2068         if (adapter->wait_for_reset)
2069                 adapter->wait_for_reset = false;
2070         return -ret;
2071 }
2072
2073 static void ibmvnic_tx_timeout(struct net_device *dev)
2074 {
2075         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2076
2077         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2078 }
2079
2080 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2081                                   struct ibmvnic_rx_buff *rx_buff)
2082 {
2083         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2084
2085         rx_buff->skb = NULL;
2086
2087         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2088         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2089
2090         atomic_dec(&pool->available);
2091 }
2092
2093 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2094 {
2095         struct net_device *netdev = napi->dev;
2096         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2097         int scrq_num = (int)(napi - adapter->napi);
2098         int frames_processed = 0;
2099
2100 restart_poll:
2101         while (frames_processed < budget) {
2102                 struct sk_buff *skb;
2103                 struct ibmvnic_rx_buff *rx_buff;
2104                 union sub_crq *next;
2105                 u32 length;
2106                 u16 offset;
2107                 u8 flags = 0;
2108
2109                 if (unlikely(adapter->resetting &&
2110                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2111                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2112                         napi_complete_done(napi, frames_processed);
2113                         return frames_processed;
2114                 }
2115
2116                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2117                         break;
2118                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2119                 rx_buff =
2120                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2121                                                           rx_comp.correlator);
2122                 /* do error checking */
2123                 if (next->rx_comp.rc) {
2124                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2125                                    be16_to_cpu(next->rx_comp.rc));
2126                         /* free the entry */
2127                         next->rx_comp.first = 0;
2128                         dev_kfree_skb_any(rx_buff->skb);
2129                         remove_buff_from_pool(adapter, rx_buff);
2130                         continue;
2131                 } else if (!rx_buff->skb) {
2132                         /* free the entry */
2133                         next->rx_comp.first = 0;
2134                         remove_buff_from_pool(adapter, rx_buff);
2135                         continue;
2136                 }
2137
2138                 length = be32_to_cpu(next->rx_comp.len);
2139                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2140                 flags = next->rx_comp.flags;
2141                 skb = rx_buff->skb;
2142                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2143                                         length);
2144
2145                 /* VLAN Header has been stripped by the system firmware and
2146                  * needs to be inserted by the driver
2147                  */
2148                 if (adapter->rx_vlan_header_insertion &&
2149                     (flags & IBMVNIC_VLAN_STRIPPED))
2150                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2151                                                ntohs(next->rx_comp.vlan_tci));
2152
2153                 /* free the entry */
2154                 next->rx_comp.first = 0;
2155                 remove_buff_from_pool(adapter, rx_buff);
2156
2157                 skb_put(skb, length);
2158                 skb->protocol = eth_type_trans(skb, netdev);
2159                 skb_record_rx_queue(skb, scrq_num);
2160
2161                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2162                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2163                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2164                 }
2165
2166                 length = skb->len;
2167                 napi_gro_receive(napi, skb); /* send it up */
2168                 netdev->stats.rx_packets++;
2169                 netdev->stats.rx_bytes += length;
2170                 adapter->rx_stats_buffers[scrq_num].packets++;
2171                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2172                 frames_processed++;
2173         }
2174
2175         if (adapter->state != VNIC_CLOSING)
2176                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2177
2178         if (frames_processed < budget) {
2179                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2180                 napi_complete_done(napi, frames_processed);
2181                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2182                     napi_reschedule(napi)) {
2183                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2184                         goto restart_poll;
2185                 }
2186         }
2187         return frames_processed;
2188 }
2189
2190 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2191 {
2192         int rc, ret;
2193
2194         adapter->fallback.mtu = adapter->req_mtu;
2195         adapter->fallback.rx_queues = adapter->req_rx_queues;
2196         adapter->fallback.tx_queues = adapter->req_tx_queues;
2197         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2198         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2199
2200         init_completion(&adapter->reset_done);
2201         adapter->wait_for_reset = true;
2202         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2203         if (rc)
2204                 return rc;
2205         wait_for_completion(&adapter->reset_done);
2206
2207         ret = 0;
2208         if (adapter->reset_done_rc) {
2209                 ret = -EIO;
2210                 adapter->desired.mtu = adapter->fallback.mtu;
2211                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2212                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2213                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2214                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2215
2216                 init_completion(&adapter->reset_done);
2217                 adapter->wait_for_reset = true;
2218                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2219                 if (rc)
2220                         return ret;
2221                 wait_for_completion(&adapter->reset_done);
2222         }
2223         adapter->wait_for_reset = false;
2224
2225         return ret;
2226 }
2227
2228 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2229 {
2230         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2231
2232         adapter->desired.mtu = new_mtu + ETH_HLEN;
2233
2234         return wait_for_reset(adapter);
2235 }
2236
2237 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2238                                                 struct net_device *dev,
2239                                                 netdev_features_t features)
2240 {
2241         /* Some backing hardware adapters can not
2242          * handle packets with a MSS less than 224
2243          * or with only one segment.
2244          */
2245         if (skb_is_gso(skb)) {
2246                 if (skb_shinfo(skb)->gso_size < 224 ||
2247                     skb_shinfo(skb)->gso_segs == 1)
2248                         features &= ~NETIF_F_GSO_MASK;
2249         }
2250
2251         return features;
2252 }
2253
2254 static const struct net_device_ops ibmvnic_netdev_ops = {
2255         .ndo_open               = ibmvnic_open,
2256         .ndo_stop               = ibmvnic_close,
2257         .ndo_start_xmit         = ibmvnic_xmit,
2258         .ndo_set_rx_mode        = ibmvnic_set_multi,
2259         .ndo_set_mac_address    = ibmvnic_set_mac,
2260         .ndo_validate_addr      = eth_validate_addr,
2261         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2262         .ndo_change_mtu         = ibmvnic_change_mtu,
2263         .ndo_features_check     = ibmvnic_features_check,
2264 };
2265
2266 /* ethtool functions */
2267
2268 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2269                                       struct ethtool_link_ksettings *cmd)
2270 {
2271         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2272         int rc;
2273
2274         rc = send_query_phys_parms(adapter);
2275         if (rc) {
2276                 adapter->speed = SPEED_UNKNOWN;
2277                 adapter->duplex = DUPLEX_UNKNOWN;
2278         }
2279         cmd->base.speed = adapter->speed;
2280         cmd->base.duplex = adapter->duplex;
2281         cmd->base.port = PORT_FIBRE;
2282         cmd->base.phy_address = 0;
2283         cmd->base.autoneg = AUTONEG_ENABLE;
2284
2285         return 0;
2286 }
2287
2288 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2289                                 struct ethtool_drvinfo *info)
2290 {
2291         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2292
2293         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2294         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2295         strlcpy(info->fw_version, adapter->fw_version,
2296                 sizeof(info->fw_version));
2297 }
2298
2299 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2300 {
2301         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2302
2303         return adapter->msg_enable;
2304 }
2305
2306 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2307 {
2308         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2309
2310         adapter->msg_enable = data;
2311 }
2312
2313 static u32 ibmvnic_get_link(struct net_device *netdev)
2314 {
2315         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2316
2317         /* Don't need to send a query because we request a logical link up at
2318          * init and then we wait for link state indications
2319          */
2320         return adapter->logical_link_state;
2321 }
2322
2323 static void ibmvnic_get_ringparam(struct net_device *netdev,
2324                                   struct ethtool_ringparam *ring)
2325 {
2326         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2327
2328         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2329                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2330                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2331         } else {
2332                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2333                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2334         }
2335         ring->rx_mini_max_pending = 0;
2336         ring->rx_jumbo_max_pending = 0;
2337         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2338         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2339         ring->rx_mini_pending = 0;
2340         ring->rx_jumbo_pending = 0;
2341 }
2342
2343 static int ibmvnic_set_ringparam(struct net_device *netdev,
2344                                  struct ethtool_ringparam *ring)
2345 {
2346         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2347         int ret;
2348
2349         ret = 0;
2350         adapter->desired.rx_entries = ring->rx_pending;
2351         adapter->desired.tx_entries = ring->tx_pending;
2352
2353         ret = wait_for_reset(adapter);
2354
2355         if (!ret &&
2356             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2357              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2358                 netdev_info(netdev,
2359                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2360                             ring->rx_pending, ring->tx_pending,
2361                             adapter->req_rx_add_entries_per_subcrq,
2362                             adapter->req_tx_entries_per_subcrq);
2363         return ret;
2364 }
2365
2366 static void ibmvnic_get_channels(struct net_device *netdev,
2367                                  struct ethtool_channels *channels)
2368 {
2369         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2370
2371         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2372                 channels->max_rx = adapter->max_rx_queues;
2373                 channels->max_tx = adapter->max_tx_queues;
2374         } else {
2375                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2376                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2377         }
2378
2379         channels->max_other = 0;
2380         channels->max_combined = 0;
2381         channels->rx_count = adapter->req_rx_queues;
2382         channels->tx_count = adapter->req_tx_queues;
2383         channels->other_count = 0;
2384         channels->combined_count = 0;
2385 }
2386
2387 static int ibmvnic_set_channels(struct net_device *netdev,
2388                                 struct ethtool_channels *channels)
2389 {
2390         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2391         int ret;
2392
2393         ret = 0;
2394         adapter->desired.rx_queues = channels->rx_count;
2395         adapter->desired.tx_queues = channels->tx_count;
2396
2397         ret = wait_for_reset(adapter);
2398
2399         if (!ret &&
2400             (adapter->req_rx_queues != channels->rx_count ||
2401              adapter->req_tx_queues != channels->tx_count))
2402                 netdev_info(netdev,
2403                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2404                             channels->rx_count, channels->tx_count,
2405                             adapter->req_rx_queues, adapter->req_tx_queues);
2406         return ret;
2407
2408 }
2409
2410 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2411 {
2412         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2413         int i;
2414
2415         switch (stringset) {
2416         case ETH_SS_STATS:
2417                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2418                                 i++, data += ETH_GSTRING_LEN)
2419                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2420
2421                 for (i = 0; i < adapter->req_tx_queues; i++) {
2422                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2423                         data += ETH_GSTRING_LEN;
2424
2425                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2426                         data += ETH_GSTRING_LEN;
2427
2428                         snprintf(data, ETH_GSTRING_LEN,
2429                                  "tx%d_dropped_packets", i);
2430                         data += ETH_GSTRING_LEN;
2431                 }
2432
2433                 for (i = 0; i < adapter->req_rx_queues; i++) {
2434                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2435                         data += ETH_GSTRING_LEN;
2436
2437                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2438                         data += ETH_GSTRING_LEN;
2439
2440                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2441                         data += ETH_GSTRING_LEN;
2442                 }
2443                 break;
2444
2445         case ETH_SS_PRIV_FLAGS:
2446                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2447                         strcpy(data + i * ETH_GSTRING_LEN,
2448                                ibmvnic_priv_flags[i]);
2449                 break;
2450         default:
2451                 return;
2452         }
2453 }
2454
2455 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2456 {
2457         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2458
2459         switch (sset) {
2460         case ETH_SS_STATS:
2461                 return ARRAY_SIZE(ibmvnic_stats) +
2462                        adapter->req_tx_queues * NUM_TX_STATS +
2463                        adapter->req_rx_queues * NUM_RX_STATS;
2464         case ETH_SS_PRIV_FLAGS:
2465                 return ARRAY_SIZE(ibmvnic_priv_flags);
2466         default:
2467                 return -EOPNOTSUPP;
2468         }
2469 }
2470
2471 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2472                                       struct ethtool_stats *stats, u64 *data)
2473 {
2474         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2475         union ibmvnic_crq crq;
2476         int i, j;
2477         int rc;
2478
2479         memset(&crq, 0, sizeof(crq));
2480         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2481         crq.request_statistics.cmd = REQUEST_STATISTICS;
2482         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2483         crq.request_statistics.len =
2484             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2485
2486         /* Wait for data to be written */
2487         init_completion(&adapter->stats_done);
2488         rc = ibmvnic_send_crq(adapter, &crq);
2489         if (rc)
2490                 return;
2491         wait_for_completion(&adapter->stats_done);
2492
2493         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2494                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2495                                                 ibmvnic_stats[i].offset));
2496
2497         for (j = 0; j < adapter->req_tx_queues; j++) {
2498                 data[i] = adapter->tx_stats_buffers[j].packets;
2499                 i++;
2500                 data[i] = adapter->tx_stats_buffers[j].bytes;
2501                 i++;
2502                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2503                 i++;
2504         }
2505
2506         for (j = 0; j < adapter->req_rx_queues; j++) {
2507                 data[i] = adapter->rx_stats_buffers[j].packets;
2508                 i++;
2509                 data[i] = adapter->rx_stats_buffers[j].bytes;
2510                 i++;
2511                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2512                 i++;
2513         }
2514 }
2515
2516 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2517 {
2518         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2519
2520         return adapter->priv_flags;
2521 }
2522
2523 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2524 {
2525         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2526         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2527
2528         if (which_maxes)
2529                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2530         else
2531                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2532
2533         return 0;
2534 }
2535 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2536         .get_drvinfo            = ibmvnic_get_drvinfo,
2537         .get_msglevel           = ibmvnic_get_msglevel,
2538         .set_msglevel           = ibmvnic_set_msglevel,
2539         .get_link               = ibmvnic_get_link,
2540         .get_ringparam          = ibmvnic_get_ringparam,
2541         .set_ringparam          = ibmvnic_set_ringparam,
2542         .get_channels           = ibmvnic_get_channels,
2543         .set_channels           = ibmvnic_set_channels,
2544         .get_strings            = ibmvnic_get_strings,
2545         .get_sset_count         = ibmvnic_get_sset_count,
2546         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2547         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2548         .get_priv_flags         = ibmvnic_get_priv_flags,
2549         .set_priv_flags         = ibmvnic_set_priv_flags,
2550 };
2551
2552 /* Routines for managing CRQs/sCRQs  */
2553
2554 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2555                                    struct ibmvnic_sub_crq_queue *scrq)
2556 {
2557         int rc;
2558
2559         if (scrq->irq) {
2560                 free_irq(scrq->irq, scrq);
2561                 irq_dispose_mapping(scrq->irq);
2562                 scrq->irq = 0;
2563         }
2564
2565         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2566         atomic_set(&scrq->used, 0);
2567         scrq->cur = 0;
2568
2569         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2570                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2571         return rc;
2572 }
2573
2574 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2575 {
2576         int i, rc;
2577
2578         for (i = 0; i < adapter->req_tx_queues; i++) {
2579                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2580                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2581                 if (rc)
2582                         return rc;
2583         }
2584
2585         for (i = 0; i < adapter->req_rx_queues; i++) {
2586                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2587                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2588                 if (rc)
2589                         return rc;
2590         }
2591
2592         return rc;
2593 }
2594
2595 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2596                                   struct ibmvnic_sub_crq_queue *scrq,
2597                                   bool do_h_free)
2598 {
2599         struct device *dev = &adapter->vdev->dev;
2600         long rc;
2601
2602         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2603
2604         if (do_h_free) {
2605                 /* Close the sub-crqs */
2606                 do {
2607                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2608                                                 adapter->vdev->unit_address,
2609                                                 scrq->crq_num);
2610                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2611
2612                 if (rc) {
2613                         netdev_err(adapter->netdev,
2614                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2615                                    scrq->crq_num, rc);
2616                 }
2617         }
2618
2619         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2620                          DMA_BIDIRECTIONAL);
2621         free_pages((unsigned long)scrq->msgs, 2);
2622         kfree(scrq);
2623 }
2624
2625 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2626                                                         *adapter)
2627 {
2628         struct device *dev = &adapter->vdev->dev;
2629         struct ibmvnic_sub_crq_queue *scrq;
2630         int rc;
2631
2632         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2633         if (!scrq)
2634                 return NULL;
2635
2636         scrq->msgs =
2637                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2638         if (!scrq->msgs) {
2639                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2640                 goto zero_page_failed;
2641         }
2642
2643         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2644                                          DMA_BIDIRECTIONAL);
2645         if (dma_mapping_error(dev, scrq->msg_token)) {
2646                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2647                 goto map_failed;
2648         }
2649
2650         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2651                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2652
2653         if (rc == H_RESOURCE)
2654                 rc = ibmvnic_reset_crq(adapter);
2655
2656         if (rc == H_CLOSED) {
2657                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2658         } else if (rc) {
2659                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2660                 goto reg_failed;
2661         }
2662
2663         scrq->adapter = adapter;
2664         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2665         spin_lock_init(&scrq->lock);
2666
2667         netdev_dbg(adapter->netdev,
2668                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2669                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2670
2671         return scrq;
2672
2673 reg_failed:
2674         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2675                          DMA_BIDIRECTIONAL);
2676 map_failed:
2677         free_pages((unsigned long)scrq->msgs, 2);
2678 zero_page_failed:
2679         kfree(scrq);
2680
2681         return NULL;
2682 }
2683
2684 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2685 {
2686         int i;
2687
2688         if (adapter->tx_scrq) {
2689                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2690                         if (!adapter->tx_scrq[i])
2691                                 continue;
2692
2693                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2694                                    i);
2695                         if (adapter->tx_scrq[i]->irq) {
2696                                 free_irq(adapter->tx_scrq[i]->irq,
2697                                          adapter->tx_scrq[i]);
2698                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2699                                 adapter->tx_scrq[i]->irq = 0;
2700                         }
2701
2702                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2703                                               do_h_free);
2704                 }
2705
2706                 kfree(adapter->tx_scrq);
2707                 adapter->tx_scrq = NULL;
2708                 adapter->num_active_tx_scrqs = 0;
2709         }
2710
2711         if (adapter->rx_scrq) {
2712                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2713                         if (!adapter->rx_scrq[i])
2714                                 continue;
2715
2716                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2717                                    i);
2718                         if (adapter->rx_scrq[i]->irq) {
2719                                 free_irq(adapter->rx_scrq[i]->irq,
2720                                          adapter->rx_scrq[i]);
2721                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2722                                 adapter->rx_scrq[i]->irq = 0;
2723                         }
2724
2725                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2726                                               do_h_free);
2727                 }
2728
2729                 kfree(adapter->rx_scrq);
2730                 adapter->rx_scrq = NULL;
2731                 adapter->num_active_rx_scrqs = 0;
2732         }
2733 }
2734
2735 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2736                             struct ibmvnic_sub_crq_queue *scrq)
2737 {
2738         struct device *dev = &adapter->vdev->dev;
2739         unsigned long rc;
2740
2741         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2742                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2743         if (rc)
2744                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2745                         scrq->hw_irq, rc);
2746         return rc;
2747 }
2748
2749 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2750                            struct ibmvnic_sub_crq_queue *scrq)
2751 {
2752         struct device *dev = &adapter->vdev->dev;
2753         unsigned long rc;
2754
2755         if (scrq->hw_irq > 0x100000000ULL) {
2756                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2757                 return 1;
2758         }
2759
2760         if (adapter->resetting &&
2761             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2762                 u64 val = (0xff000000) | scrq->hw_irq;
2763
2764                 rc = plpar_hcall_norets(H_EOI, val);
2765                 if (rc)
2766                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2767                                 val, rc);
2768         }
2769
2770         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2771                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2772         if (rc)
2773                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2774                         scrq->hw_irq, rc);
2775         return rc;
2776 }
2777
2778 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2779                                struct ibmvnic_sub_crq_queue *scrq)
2780 {
2781         struct device *dev = &adapter->vdev->dev;
2782         struct ibmvnic_tx_pool *tx_pool;
2783         struct ibmvnic_tx_buff *txbuff;
2784         union sub_crq *next;
2785         int index;
2786         int i, j;
2787         u8 *first;
2788
2789 restart_loop:
2790         while (pending_scrq(adapter, scrq)) {
2791                 unsigned int pool = scrq->pool_index;
2792                 int num_entries = 0;
2793
2794                 next = ibmvnic_next_scrq(adapter, scrq);
2795                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2796                         if (next->tx_comp.rcs[i]) {
2797                                 dev_err(dev, "tx error %x\n",
2798                                         next->tx_comp.rcs[i]);
2799                                 continue;
2800                         }
2801                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2802                         if (index & IBMVNIC_TSO_POOL_MASK) {
2803                                 tx_pool = &adapter->tso_pool[pool];
2804                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2805                         } else {
2806                                 tx_pool = &adapter->tx_pool[pool];
2807                         }
2808
2809                         txbuff = &tx_pool->tx_buff[index];
2810
2811                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2812                                 if (!txbuff->data_dma[j])
2813                                         continue;
2814
2815                                 txbuff->data_dma[j] = 0;
2816                         }
2817                         /* if sub_crq was sent indirectly */
2818                         first = &txbuff->indir_arr[0].generic.first;
2819                         if (*first == IBMVNIC_CRQ_CMD) {
2820                                 dma_unmap_single(dev, txbuff->indir_dma,
2821                                                  sizeof(txbuff->indir_arr),
2822                                                  DMA_TO_DEVICE);
2823                                 *first = 0;
2824                         }
2825
2826                         if (txbuff->last_frag) {
2827                                 dev_kfree_skb_any(txbuff->skb);
2828                                 txbuff->skb = NULL;
2829                         }
2830
2831                         num_entries += txbuff->num_entries;
2832
2833                         tx_pool->free_map[tx_pool->producer_index] = index;
2834                         tx_pool->producer_index =
2835                                 (tx_pool->producer_index + 1) %
2836                                         tx_pool->num_buffers;
2837                 }
2838                 /* remove tx_comp scrq*/
2839                 next->tx_comp.first = 0;
2840
2841                 if (atomic_sub_return(num_entries, &scrq->used) <=
2842                     (adapter->req_tx_entries_per_subcrq / 2) &&
2843                     __netif_subqueue_stopped(adapter->netdev,
2844                                              scrq->pool_index)) {
2845                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2846                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2847                                    scrq->pool_index);
2848                 }
2849         }
2850
2851         enable_scrq_irq(adapter, scrq);
2852
2853         if (pending_scrq(adapter, scrq)) {
2854                 disable_scrq_irq(adapter, scrq);
2855                 goto restart_loop;
2856         }
2857
2858         return 0;
2859 }
2860
2861 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2862 {
2863         struct ibmvnic_sub_crq_queue *scrq = instance;
2864         struct ibmvnic_adapter *adapter = scrq->adapter;
2865
2866         disable_scrq_irq(adapter, scrq);
2867         ibmvnic_complete_tx(adapter, scrq);
2868
2869         return IRQ_HANDLED;
2870 }
2871
2872 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2873 {
2874         struct ibmvnic_sub_crq_queue *scrq = instance;
2875         struct ibmvnic_adapter *adapter = scrq->adapter;
2876
2877         /* When booting a kdump kernel we can hit pending interrupts
2878          * prior to completing driver initialization.
2879          */
2880         if (unlikely(adapter->state != VNIC_OPEN))
2881                 return IRQ_NONE;
2882
2883         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2884
2885         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2886                 disable_scrq_irq(adapter, scrq);
2887                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2888         }
2889
2890         return IRQ_HANDLED;
2891 }
2892
2893 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2894 {
2895         struct device *dev = &adapter->vdev->dev;
2896         struct ibmvnic_sub_crq_queue *scrq;
2897         int i = 0, j = 0;
2898         int rc = 0;
2899
2900         for (i = 0; i < adapter->req_tx_queues; i++) {
2901                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2902                            i);
2903                 scrq = adapter->tx_scrq[i];
2904                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2905
2906                 if (!scrq->irq) {
2907                         rc = -EINVAL;
2908                         dev_err(dev, "Error mapping irq\n");
2909                         goto req_tx_irq_failed;
2910                 }
2911
2912                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
2913                          adapter->vdev->unit_address, i);
2914                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2915                                  0, scrq->name, scrq);
2916
2917                 if (rc) {
2918                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2919                                 scrq->irq, rc);
2920                         irq_dispose_mapping(scrq->irq);
2921                         goto req_tx_irq_failed;
2922                 }
2923         }
2924
2925         for (i = 0; i < adapter->req_rx_queues; i++) {
2926                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2927                            i);
2928                 scrq = adapter->rx_scrq[i];
2929                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2930                 if (!scrq->irq) {
2931                         rc = -EINVAL;
2932                         dev_err(dev, "Error mapping irq\n");
2933                         goto req_rx_irq_failed;
2934                 }
2935                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
2936                          adapter->vdev->unit_address, i);
2937                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2938                                  0, scrq->name, scrq);
2939                 if (rc) {
2940                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2941                                 scrq->irq, rc);
2942                         irq_dispose_mapping(scrq->irq);
2943                         goto req_rx_irq_failed;
2944                 }
2945         }
2946         return rc;
2947
2948 req_rx_irq_failed:
2949         for (j = 0; j < i; j++) {
2950                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2951                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2952         }
2953         i = adapter->req_tx_queues;
2954 req_tx_irq_failed:
2955         for (j = 0; j < i; j++) {
2956                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2957                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2958         }
2959         release_sub_crqs(adapter, 1);
2960         return rc;
2961 }
2962
2963 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2964 {
2965         struct device *dev = &adapter->vdev->dev;
2966         struct ibmvnic_sub_crq_queue **allqueues;
2967         int registered_queues = 0;
2968         int total_queues;
2969         int more = 0;
2970         int i;
2971
2972         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2973
2974         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2975         if (!allqueues)
2976                 return -1;
2977
2978         for (i = 0; i < total_queues; i++) {
2979                 allqueues[i] = init_sub_crq_queue(adapter);
2980                 if (!allqueues[i]) {
2981                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2982                         break;
2983                 }
2984                 registered_queues++;
2985         }
2986
2987         /* Make sure we were able to register the minimum number of queues */
2988         if (registered_queues <
2989             adapter->min_tx_queues + adapter->min_rx_queues) {
2990                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2991                 goto tx_failed;
2992         }
2993
2994         /* Distribute the failed allocated queues*/
2995         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2996                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2997                 switch (i % 3) {
2998                 case 0:
2999                         if (adapter->req_rx_queues > adapter->min_rx_queues)
3000                                 adapter->req_rx_queues--;
3001                         else
3002                                 more++;
3003                         break;
3004                 case 1:
3005                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3006                                 adapter->req_tx_queues--;
3007                         else
3008                                 more++;
3009                         break;
3010                 }
3011         }
3012
3013         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3014                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3015         if (!adapter->tx_scrq)
3016                 goto tx_failed;
3017
3018         for (i = 0; i < adapter->req_tx_queues; i++) {
3019                 adapter->tx_scrq[i] = allqueues[i];
3020                 adapter->tx_scrq[i]->pool_index = i;
3021                 adapter->num_active_tx_scrqs++;
3022         }
3023
3024         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3025                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3026         if (!adapter->rx_scrq)
3027                 goto rx_failed;
3028
3029         for (i = 0; i < adapter->req_rx_queues; i++) {
3030                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3031                 adapter->rx_scrq[i]->scrq_num = i;
3032                 adapter->num_active_rx_scrqs++;
3033         }
3034
3035         kfree(allqueues);
3036         return 0;
3037
3038 rx_failed:
3039         kfree(adapter->tx_scrq);
3040         adapter->tx_scrq = NULL;
3041 tx_failed:
3042         for (i = 0; i < registered_queues; i++)
3043                 release_sub_crq_queue(adapter, allqueues[i], 1);
3044         kfree(allqueues);
3045         return -1;
3046 }
3047
3048 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3049 {
3050         struct device *dev = &adapter->vdev->dev;
3051         union ibmvnic_crq crq;
3052         int max_entries;
3053
3054         if (!retry) {
3055                 /* Sub-CRQ entries are 32 byte long */
3056                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3057
3058                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3059                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3060                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3061                         return;
3062                 }
3063
3064                 if (adapter->desired.mtu)
3065                         adapter->req_mtu = adapter->desired.mtu;
3066                 else
3067                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3068
3069                 if (!adapter->desired.tx_entries)
3070                         adapter->desired.tx_entries =
3071                                         adapter->max_tx_entries_per_subcrq;
3072                 if (!adapter->desired.rx_entries)
3073                         adapter->desired.rx_entries =
3074                                         adapter->max_rx_add_entries_per_subcrq;
3075
3076                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3077                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3078
3079                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3080                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3081                         adapter->desired.tx_entries = max_entries;
3082                 }
3083
3084                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3085                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3086                         adapter->desired.rx_entries = max_entries;
3087                 }
3088
3089                 if (adapter->desired.tx_entries)
3090                         adapter->req_tx_entries_per_subcrq =
3091                                         adapter->desired.tx_entries;
3092                 else
3093                         adapter->req_tx_entries_per_subcrq =
3094                                         adapter->max_tx_entries_per_subcrq;
3095
3096                 if (adapter->desired.rx_entries)
3097                         adapter->req_rx_add_entries_per_subcrq =
3098                                         adapter->desired.rx_entries;
3099                 else
3100                         adapter->req_rx_add_entries_per_subcrq =
3101                                         adapter->max_rx_add_entries_per_subcrq;
3102
3103                 if (adapter->desired.tx_queues)
3104                         adapter->req_tx_queues =
3105                                         adapter->desired.tx_queues;
3106                 else
3107                         adapter->req_tx_queues =
3108                                         adapter->opt_tx_comp_sub_queues;
3109
3110                 if (adapter->desired.rx_queues)
3111                         adapter->req_rx_queues =
3112                                         adapter->desired.rx_queues;
3113                 else
3114                         adapter->req_rx_queues =
3115                                         adapter->opt_rx_comp_queues;
3116
3117                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3118         }
3119
3120         memset(&crq, 0, sizeof(crq));
3121         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3122         crq.request_capability.cmd = REQUEST_CAPABILITY;
3123
3124         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3125         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3126         atomic_inc(&adapter->running_cap_crqs);
3127         ibmvnic_send_crq(adapter, &crq);
3128
3129         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3130         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3131         atomic_inc(&adapter->running_cap_crqs);
3132         ibmvnic_send_crq(adapter, &crq);
3133
3134         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3135         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3136         atomic_inc(&adapter->running_cap_crqs);
3137         ibmvnic_send_crq(adapter, &crq);
3138
3139         crq.request_capability.capability =
3140             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3141         crq.request_capability.number =
3142             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3143         atomic_inc(&adapter->running_cap_crqs);
3144         ibmvnic_send_crq(adapter, &crq);
3145
3146         crq.request_capability.capability =
3147             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3148         crq.request_capability.number =
3149             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3150         atomic_inc(&adapter->running_cap_crqs);
3151         ibmvnic_send_crq(adapter, &crq);
3152
3153         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3154         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3155         atomic_inc(&adapter->running_cap_crqs);
3156         ibmvnic_send_crq(adapter, &crq);
3157
3158         if (adapter->netdev->flags & IFF_PROMISC) {
3159                 if (adapter->promisc_supported) {
3160                         crq.request_capability.capability =
3161                             cpu_to_be16(PROMISC_REQUESTED);
3162                         crq.request_capability.number = cpu_to_be64(1);
3163                         atomic_inc(&adapter->running_cap_crqs);
3164                         ibmvnic_send_crq(adapter, &crq);
3165                 }
3166         } else {
3167                 crq.request_capability.capability =
3168                     cpu_to_be16(PROMISC_REQUESTED);
3169                 crq.request_capability.number = cpu_to_be64(0);
3170                 atomic_inc(&adapter->running_cap_crqs);
3171                 ibmvnic_send_crq(adapter, &crq);
3172         }
3173 }
3174
3175 static int pending_scrq(struct ibmvnic_adapter *adapter,
3176                         struct ibmvnic_sub_crq_queue *scrq)
3177 {
3178         union sub_crq *entry = &scrq->msgs[scrq->cur];
3179
3180         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3181                 return 1;
3182         else
3183                 return 0;
3184 }
3185
3186 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3187                                         struct ibmvnic_sub_crq_queue *scrq)
3188 {
3189         union sub_crq *entry;
3190         unsigned long flags;
3191
3192         spin_lock_irqsave(&scrq->lock, flags);
3193         entry = &scrq->msgs[scrq->cur];
3194         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3195                 if (++scrq->cur == scrq->size)
3196                         scrq->cur = 0;
3197         } else {
3198                 entry = NULL;
3199         }
3200         spin_unlock_irqrestore(&scrq->lock, flags);
3201
3202         return entry;
3203 }
3204
3205 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3206 {
3207         struct ibmvnic_crq_queue *queue = &adapter->crq;
3208         union ibmvnic_crq *crq;
3209
3210         crq = &queue->msgs[queue->cur];
3211         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3212                 if (++queue->cur == queue->size)
3213                         queue->cur = 0;
3214         } else {
3215                 crq = NULL;
3216         }
3217
3218         return crq;
3219 }
3220
3221 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3222 {
3223         switch (rc) {
3224         case H_PARAMETER:
3225                 dev_warn_ratelimited(dev,
3226                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3227                                      func, rc);
3228                 break;
3229         case H_CLOSED:
3230                 dev_warn_ratelimited(dev,
3231                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3232                                      func, rc);
3233                 break;
3234         default:
3235                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3236                 break;
3237         }
3238 }
3239
3240 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3241                        union sub_crq *sub_crq)
3242 {
3243         unsigned int ua = adapter->vdev->unit_address;
3244         struct device *dev = &adapter->vdev->dev;
3245         u64 *u64_crq = (u64 *)sub_crq;
3246         int rc;
3247
3248         netdev_dbg(adapter->netdev,
3249                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3250                    (unsigned long int)cpu_to_be64(remote_handle),
3251                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3252                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3253                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3254                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3255
3256         /* Make sure the hypervisor sees the complete request */
3257         mb();
3258
3259         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3260                                 cpu_to_be64(remote_handle),
3261                                 cpu_to_be64(u64_crq[0]),
3262                                 cpu_to_be64(u64_crq[1]),
3263                                 cpu_to_be64(u64_crq[2]),
3264                                 cpu_to_be64(u64_crq[3]));
3265
3266         if (rc)
3267                 print_subcrq_error(dev, rc, __func__);
3268
3269         return rc;
3270 }
3271
3272 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3273                                 u64 remote_handle, u64 ioba, u64 num_entries)
3274 {
3275         unsigned int ua = adapter->vdev->unit_address;
3276         struct device *dev = &adapter->vdev->dev;
3277         int rc;
3278
3279         /* Make sure the hypervisor sees the complete request */
3280         mb();
3281         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3282                                 cpu_to_be64(remote_handle),
3283                                 ioba, num_entries);
3284
3285         if (rc)
3286                 print_subcrq_error(dev, rc, __func__);
3287
3288         return rc;
3289 }
3290
3291 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3292                             union ibmvnic_crq *crq)
3293 {
3294         unsigned int ua = adapter->vdev->unit_address;
3295         struct device *dev = &adapter->vdev->dev;
3296         u64 *u64_crq = (u64 *)crq;
3297         int rc;
3298
3299         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3300                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3301                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3302
3303         if (!adapter->crq.active &&
3304             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3305                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3306                 return -EINVAL;
3307         }
3308
3309         /* Make sure the hypervisor sees the complete request */
3310         mb();
3311
3312         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3313                                 cpu_to_be64(u64_crq[0]),
3314                                 cpu_to_be64(u64_crq[1]));
3315
3316         if (rc) {
3317                 if (rc == H_CLOSED) {
3318                         dev_warn(dev, "CRQ Queue closed\n");
3319                         if (adapter->resetting)
3320                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3321                 }
3322
3323                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3324         }
3325
3326         return rc;
3327 }
3328
3329 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3330 {
3331         union ibmvnic_crq crq;
3332
3333         memset(&crq, 0, sizeof(crq));
3334         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3335         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3336         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3337
3338         return ibmvnic_send_crq(adapter, &crq);
3339 }
3340
3341 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3342 {
3343         union ibmvnic_crq crq;
3344
3345         memset(&crq, 0, sizeof(crq));
3346         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3347         crq.version_exchange.cmd = VERSION_EXCHANGE;
3348         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3349
3350         return ibmvnic_send_crq(adapter, &crq);
3351 }
3352
3353 struct vnic_login_client_data {
3354         u8      type;
3355         __be16  len;
3356         char    name[];
3357 } __packed;
3358
3359 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3360 {
3361         int len;
3362
3363         /* Calculate the amount of buffer space needed for the
3364          * vnic client data in the login buffer. There are four entries,
3365          * OS name, LPAR name, device name, and a null last entry.
3366          */
3367         len = 4 * sizeof(struct vnic_login_client_data);
3368         len += 6; /* "Linux" plus NULL */
3369         len += strlen(utsname()->nodename) + 1;
3370         len += strlen(adapter->netdev->name) + 1;
3371
3372         return len;
3373 }
3374
3375 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3376                                  struct vnic_login_client_data *vlcd)
3377 {
3378         const char *os_name = "Linux";
3379         int len;
3380
3381         /* Type 1 - LPAR OS */
3382         vlcd->type = 1;
3383         len = strlen(os_name) + 1;
3384         vlcd->len = cpu_to_be16(len);
3385         strncpy(vlcd->name, os_name, len);
3386         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3387
3388         /* Type 2 - LPAR name */
3389         vlcd->type = 2;
3390         len = strlen(utsname()->nodename) + 1;
3391         vlcd->len = cpu_to_be16(len);
3392         strncpy(vlcd->name, utsname()->nodename, len);
3393         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3394
3395         /* Type 3 - device name */
3396         vlcd->type = 3;
3397         len = strlen(adapter->netdev->name) + 1;
3398         vlcd->len = cpu_to_be16(len);
3399         strncpy(vlcd->name, adapter->netdev->name, len);
3400 }
3401
3402 static int send_login(struct ibmvnic_adapter *adapter)
3403 {
3404         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3405         struct ibmvnic_login_buffer *login_buffer;
3406         struct device *dev = &adapter->vdev->dev;
3407         dma_addr_t rsp_buffer_token;
3408         dma_addr_t buffer_token;
3409         size_t rsp_buffer_size;
3410         union ibmvnic_crq crq;
3411         size_t buffer_size;
3412         __be64 *tx_list_p;
3413         __be64 *rx_list_p;
3414         int client_data_len;
3415         struct vnic_login_client_data *vlcd;
3416         int i;
3417
3418         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3419                 netdev_err(adapter->netdev,
3420                            "RX or TX queues are not allocated, device login failed\n");
3421                 return -1;
3422         }
3423
3424         release_login_rsp_buffer(adapter);
3425         client_data_len = vnic_client_data_len(adapter);
3426
3427         buffer_size =
3428             sizeof(struct ibmvnic_login_buffer) +
3429             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3430             client_data_len;
3431
3432         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3433         if (!login_buffer)
3434                 goto buf_alloc_failed;
3435
3436         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3437                                       DMA_TO_DEVICE);
3438         if (dma_mapping_error(dev, buffer_token)) {
3439                 dev_err(dev, "Couldn't map login buffer\n");
3440                 goto buf_map_failed;
3441         }
3442
3443         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3444                           sizeof(u64) * adapter->req_tx_queues +
3445                           sizeof(u64) * adapter->req_rx_queues +
3446                           sizeof(u64) * adapter->req_rx_queues +
3447                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3448
3449         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3450         if (!login_rsp_buffer)
3451                 goto buf_rsp_alloc_failed;
3452
3453         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3454                                           rsp_buffer_size, DMA_FROM_DEVICE);
3455         if (dma_mapping_error(dev, rsp_buffer_token)) {
3456                 dev_err(dev, "Couldn't map login rsp buffer\n");
3457                 goto buf_rsp_map_failed;
3458         }
3459
3460         adapter->login_buf = login_buffer;
3461         adapter->login_buf_token = buffer_token;
3462         adapter->login_buf_sz = buffer_size;
3463         adapter->login_rsp_buf = login_rsp_buffer;
3464         adapter->login_rsp_buf_token = rsp_buffer_token;
3465         adapter->login_rsp_buf_sz = rsp_buffer_size;
3466
3467         login_buffer->len = cpu_to_be32(buffer_size);
3468         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3469         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3470         login_buffer->off_txcomp_subcrqs =
3471             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3472         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3473         login_buffer->off_rxcomp_subcrqs =
3474             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3475                         sizeof(u64) * adapter->req_tx_queues);
3476         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3477         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3478
3479         tx_list_p = (__be64 *)((char *)login_buffer +
3480                                       sizeof(struct ibmvnic_login_buffer));
3481         rx_list_p = (__be64 *)((char *)login_buffer +
3482                                       sizeof(struct ibmvnic_login_buffer) +
3483                                       sizeof(u64) * adapter->req_tx_queues);
3484
3485         for (i = 0; i < adapter->req_tx_queues; i++) {
3486                 if (adapter->tx_scrq[i]) {
3487                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3488                                                    crq_num);
3489                 }
3490         }
3491
3492         for (i = 0; i < adapter->req_rx_queues; i++) {
3493                 if (adapter->rx_scrq[i]) {
3494                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3495                                                    crq_num);
3496                 }
3497         }
3498
3499         /* Insert vNIC login client data */
3500         vlcd = (struct vnic_login_client_data *)
3501                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3502         login_buffer->client_data_offset =
3503                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3504         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3505
3506         vnic_add_client_data(adapter, vlcd);
3507
3508         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3509         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3510                 netdev_dbg(adapter->netdev, "%016lx\n",
3511                            ((unsigned long int *)(adapter->login_buf))[i]);
3512         }
3513
3514         memset(&crq, 0, sizeof(crq));
3515         crq.login.first = IBMVNIC_CRQ_CMD;
3516         crq.login.cmd = LOGIN;
3517         crq.login.ioba = cpu_to_be32(buffer_token);
3518         crq.login.len = cpu_to_be32(buffer_size);
3519         ibmvnic_send_crq(adapter, &crq);
3520
3521         return 0;
3522
3523 buf_rsp_map_failed:
3524         kfree(login_rsp_buffer);
3525 buf_rsp_alloc_failed:
3526         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3527 buf_map_failed:
3528         kfree(login_buffer);
3529 buf_alloc_failed:
3530         return -1;
3531 }
3532
3533 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3534                             u32 len, u8 map_id)
3535 {
3536         union ibmvnic_crq crq;
3537
3538         memset(&crq, 0, sizeof(crq));
3539         crq.request_map.first = IBMVNIC_CRQ_CMD;
3540         crq.request_map.cmd = REQUEST_MAP;
3541         crq.request_map.map_id = map_id;
3542         crq.request_map.ioba = cpu_to_be32(addr);
3543         crq.request_map.len = cpu_to_be32(len);
3544         return ibmvnic_send_crq(adapter, &crq);
3545 }
3546
3547 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3548 {
3549         union ibmvnic_crq crq;
3550
3551         memset(&crq, 0, sizeof(crq));
3552         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3553         crq.request_unmap.cmd = REQUEST_UNMAP;
3554         crq.request_unmap.map_id = map_id;
3555         return ibmvnic_send_crq(adapter, &crq);
3556 }
3557
3558 static void send_map_query(struct ibmvnic_adapter *adapter)
3559 {
3560         union ibmvnic_crq crq;
3561
3562         memset(&crq, 0, sizeof(crq));
3563         crq.query_map.first = IBMVNIC_CRQ_CMD;
3564         crq.query_map.cmd = QUERY_MAP;
3565         ibmvnic_send_crq(adapter, &crq);
3566 }
3567
3568 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3569 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3570 {
3571         union ibmvnic_crq crq;
3572
3573         atomic_set(&adapter->running_cap_crqs, 0);
3574         memset(&crq, 0, sizeof(crq));
3575         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3576         crq.query_capability.cmd = QUERY_CAPABILITY;
3577
3578         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3579         atomic_inc(&adapter->running_cap_crqs);
3580         ibmvnic_send_crq(adapter, &crq);
3581
3582         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3583         atomic_inc(&adapter->running_cap_crqs);
3584         ibmvnic_send_crq(adapter, &crq);
3585
3586         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3587         atomic_inc(&adapter->running_cap_crqs);
3588         ibmvnic_send_crq(adapter, &crq);
3589
3590         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3591         atomic_inc(&adapter->running_cap_crqs);
3592         ibmvnic_send_crq(adapter, &crq);
3593
3594         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3595         atomic_inc(&adapter->running_cap_crqs);
3596         ibmvnic_send_crq(adapter, &crq);
3597
3598         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3599         atomic_inc(&adapter->running_cap_crqs);
3600         ibmvnic_send_crq(adapter, &crq);
3601
3602         crq.query_capability.capability =
3603             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3604         atomic_inc(&adapter->running_cap_crqs);
3605         ibmvnic_send_crq(adapter, &crq);
3606
3607         crq.query_capability.capability =
3608             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3609         atomic_inc(&adapter->running_cap_crqs);
3610         ibmvnic_send_crq(adapter, &crq);
3611
3612         crq.query_capability.capability =
3613             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3614         atomic_inc(&adapter->running_cap_crqs);
3615         ibmvnic_send_crq(adapter, &crq);
3616
3617         crq.query_capability.capability =
3618             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3619         atomic_inc(&adapter->running_cap_crqs);
3620         ibmvnic_send_crq(adapter, &crq);
3621
3622         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3623         atomic_inc(&adapter->running_cap_crqs);
3624         ibmvnic_send_crq(adapter, &crq);
3625
3626         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3627         atomic_inc(&adapter->running_cap_crqs);
3628         ibmvnic_send_crq(adapter, &crq);
3629
3630         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3631         atomic_inc(&adapter->running_cap_crqs);
3632         ibmvnic_send_crq(adapter, &crq);
3633
3634         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3635         atomic_inc(&adapter->running_cap_crqs);
3636         ibmvnic_send_crq(adapter, &crq);
3637
3638         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3639         atomic_inc(&adapter->running_cap_crqs);
3640         ibmvnic_send_crq(adapter, &crq);
3641
3642         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3643         atomic_inc(&adapter->running_cap_crqs);
3644         ibmvnic_send_crq(adapter, &crq);
3645
3646         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3647         atomic_inc(&adapter->running_cap_crqs);
3648         ibmvnic_send_crq(adapter, &crq);
3649
3650         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3651         atomic_inc(&adapter->running_cap_crqs);
3652         ibmvnic_send_crq(adapter, &crq);
3653
3654         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3655         atomic_inc(&adapter->running_cap_crqs);
3656         ibmvnic_send_crq(adapter, &crq);
3657
3658         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3659         atomic_inc(&adapter->running_cap_crqs);
3660         ibmvnic_send_crq(adapter, &crq);
3661
3662         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3663         atomic_inc(&adapter->running_cap_crqs);
3664         ibmvnic_send_crq(adapter, &crq);
3665
3666         crq.query_capability.capability =
3667                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3668         atomic_inc(&adapter->running_cap_crqs);
3669         ibmvnic_send_crq(adapter, &crq);
3670
3671         crq.query_capability.capability =
3672                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3673         atomic_inc(&adapter->running_cap_crqs);
3674         ibmvnic_send_crq(adapter, &crq);
3675
3676         crq.query_capability.capability =
3677                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3678         atomic_inc(&adapter->running_cap_crqs);
3679         ibmvnic_send_crq(adapter, &crq);
3680
3681         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3682         atomic_inc(&adapter->running_cap_crqs);
3683         ibmvnic_send_crq(adapter, &crq);
3684 }
3685
3686 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3687                                 struct ibmvnic_adapter *adapter)
3688 {
3689         struct device *dev = &adapter->vdev->dev;
3690
3691         if (crq->get_vpd_size_rsp.rc.code) {
3692                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3693                         crq->get_vpd_size_rsp.rc.code);
3694                 complete(&adapter->fw_done);
3695                 return;
3696         }
3697
3698         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3699         complete(&adapter->fw_done);
3700 }
3701
3702 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3703                            struct ibmvnic_adapter *adapter)
3704 {
3705         struct device *dev = &adapter->vdev->dev;
3706         unsigned char *substr = NULL;
3707         u8 fw_level_len = 0;
3708
3709         memset(adapter->fw_version, 0, 32);
3710
3711         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3712                          DMA_FROM_DEVICE);
3713
3714         if (crq->get_vpd_rsp.rc.code) {
3715                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3716                         crq->get_vpd_rsp.rc.code);
3717                 goto complete;
3718         }
3719
3720         /* get the position of the firmware version info
3721          * located after the ASCII 'RM' substring in the buffer
3722          */
3723         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3724         if (!substr) {
3725                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3726                 goto complete;
3727         }
3728
3729         /* get length of firmware level ASCII substring */
3730         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3731                 fw_level_len = *(substr + 2);
3732         } else {
3733                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3734                 goto complete;
3735         }
3736
3737         /* copy firmware version string from vpd into adapter */
3738         if ((substr + 3 + fw_level_len) <
3739             (adapter->vpd->buff + adapter->vpd->len)) {
3740                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3741         } else {
3742                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3743         }
3744
3745 complete:
3746         if (adapter->fw_version[0] == '\0')
3747                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3748         complete(&adapter->fw_done);
3749 }
3750
3751 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3752 {
3753         struct device *dev = &adapter->vdev->dev;
3754         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3755         netdev_features_t old_hw_features = 0;
3756         union ibmvnic_crq crq;
3757         int i;
3758
3759         dma_unmap_single(dev, adapter->ip_offload_tok,
3760                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3761
3762         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3763         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3764                 netdev_dbg(adapter->netdev, "%016lx\n",
3765                            ((unsigned long int *)(buf))[i]);
3766
3767         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3768         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3769         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3770                    buf->tcp_ipv4_chksum);
3771         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3772                    buf->tcp_ipv6_chksum);
3773         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3774                    buf->udp_ipv4_chksum);
3775         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3776                    buf->udp_ipv6_chksum);
3777         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3778                    buf->large_tx_ipv4);
3779         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3780                    buf->large_tx_ipv6);
3781         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3782                    buf->large_rx_ipv4);
3783         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3784                    buf->large_rx_ipv6);
3785         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3786                    buf->max_ipv4_header_size);
3787         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3788                    buf->max_ipv6_header_size);
3789         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3790                    buf->max_tcp_header_size);
3791         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3792                    buf->max_udp_header_size);
3793         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3794                    buf->max_large_tx_size);
3795         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3796                    buf->max_large_rx_size);
3797         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3798                    buf->ipv6_extension_header);
3799         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3800                    buf->tcp_pseudosum_req);
3801         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3802                    buf->num_ipv6_ext_headers);
3803         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3804                    buf->off_ipv6_ext_headers);
3805
3806         adapter->ip_offload_ctrl_tok =
3807             dma_map_single(dev, &adapter->ip_offload_ctrl,
3808                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3809
3810         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3811                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3812                 return;
3813         }
3814
3815         adapter->ip_offload_ctrl.len =
3816             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3817         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3818         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3819         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3820         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3821         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3822         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3823         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3824         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3825         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3826
3827         /* large_rx disabled for now, additional features needed */
3828         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3829         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3830
3831         if (adapter->state != VNIC_PROBING) {
3832                 old_hw_features = adapter->netdev->hw_features;
3833                 adapter->netdev->hw_features = 0;
3834         }
3835
3836         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
3837
3838         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3839                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
3840
3841         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3842                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
3843
3844         if ((adapter->netdev->features &
3845             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3846                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
3847
3848         if (buf->large_tx_ipv4)
3849                 adapter->netdev->hw_features |= NETIF_F_TSO;
3850         if (buf->large_tx_ipv6)
3851                 adapter->netdev->hw_features |= NETIF_F_TSO6;
3852
3853         if (adapter->state == VNIC_PROBING) {
3854                 adapter->netdev->features |= adapter->netdev->hw_features;
3855         } else if (old_hw_features != adapter->netdev->hw_features) {
3856                 netdev_features_t tmp = 0;
3857
3858                 /* disable features no longer supported */
3859                 adapter->netdev->features &= adapter->netdev->hw_features;
3860                 /* turn on features now supported if previously enabled */
3861                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
3862                         adapter->netdev->hw_features;
3863                 adapter->netdev->features |=
3864                                 tmp & adapter->netdev->wanted_features;
3865         }
3866
3867         memset(&crq, 0, sizeof(crq));
3868         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3869         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3870         crq.control_ip_offload.len =
3871             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3872         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3873         ibmvnic_send_crq(adapter, &crq);
3874 }
3875
3876 static const char *ibmvnic_fw_err_cause(u16 cause)
3877 {
3878         switch (cause) {
3879         case ADAPTER_PROBLEM:
3880                 return "adapter problem";
3881         case BUS_PROBLEM:
3882                 return "bus problem";
3883         case FW_PROBLEM:
3884                 return "firmware problem";
3885         case DD_PROBLEM:
3886                 return "device driver problem";
3887         case EEH_RECOVERY:
3888                 return "EEH recovery";
3889         case FW_UPDATED:
3890                 return "firmware updated";
3891         case LOW_MEMORY:
3892                 return "low Memory";
3893         default:
3894                 return "unknown";
3895         }
3896 }
3897
3898 static void handle_error_indication(union ibmvnic_crq *crq,
3899                                     struct ibmvnic_adapter *adapter)
3900 {
3901         struct device *dev = &adapter->vdev->dev;
3902         u16 cause;
3903
3904         cause = be16_to_cpu(crq->error_indication.error_cause);
3905
3906         dev_warn_ratelimited(dev,
3907                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3908                              crq->error_indication.flags
3909                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3910                              ibmvnic_fw_err_cause(cause));
3911
3912         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3913                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3914         else
3915                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3916 }
3917
3918 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3919                                  struct ibmvnic_adapter *adapter)
3920 {
3921         struct net_device *netdev = adapter->netdev;
3922         struct device *dev = &adapter->vdev->dev;
3923         long rc;
3924
3925         rc = crq->change_mac_addr_rsp.rc.code;
3926         if (rc) {
3927                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3928                 goto out;
3929         }
3930         ether_addr_copy(netdev->dev_addr,
3931                         &crq->change_mac_addr_rsp.mac_addr[0]);
3932 out:
3933         complete(&adapter->fw_done);
3934         return rc;
3935 }
3936
3937 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3938                                    struct ibmvnic_adapter *adapter)
3939 {
3940         struct device *dev = &adapter->vdev->dev;
3941         u64 *req_value;
3942         char *name;
3943
3944         atomic_dec(&adapter->running_cap_crqs);
3945         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3946         case REQ_TX_QUEUES:
3947                 req_value = &adapter->req_tx_queues;
3948                 name = "tx";
3949                 break;
3950         case REQ_RX_QUEUES:
3951                 req_value = &adapter->req_rx_queues;
3952                 name = "rx";
3953                 break;
3954         case REQ_RX_ADD_QUEUES:
3955                 req_value = &adapter->req_rx_add_queues;
3956                 name = "rx_add";
3957                 break;
3958         case REQ_TX_ENTRIES_PER_SUBCRQ:
3959                 req_value = &adapter->req_tx_entries_per_subcrq;
3960                 name = "tx_entries_per_subcrq";
3961                 break;
3962         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3963                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3964                 name = "rx_add_entries_per_subcrq";
3965                 break;
3966         case REQ_MTU:
3967                 req_value = &adapter->req_mtu;
3968                 name = "mtu";
3969                 break;
3970         case PROMISC_REQUESTED:
3971                 req_value = &adapter->promisc;
3972                 name = "promisc";
3973                 break;
3974         default:
3975                 dev_err(dev, "Got invalid cap request rsp %d\n",
3976                         crq->request_capability.capability);
3977                 return;
3978         }
3979
3980         switch (crq->request_capability_rsp.rc.code) {
3981         case SUCCESS:
3982                 break;
3983         case PARTIALSUCCESS:
3984                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3985                          *req_value,
3986                          (long int)be64_to_cpu(crq->request_capability_rsp.
3987                                                number), name);
3988
3989                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3990                     REQ_MTU) {
3991                         pr_err("mtu of %llu is not supported. Reverting.\n",
3992                                *req_value);
3993                         *req_value = adapter->fallback.mtu;
3994                 } else {
3995                         *req_value =
3996                                 be64_to_cpu(crq->request_capability_rsp.number);
3997                 }
3998
3999                 ibmvnic_send_req_caps(adapter, 1);
4000                 return;
4001         default:
4002                 dev_err(dev, "Error %d in request cap rsp\n",
4003                         crq->request_capability_rsp.rc.code);
4004                 return;
4005         }
4006
4007         /* Done receiving requested capabilities, query IP offload support */
4008         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4009                 union ibmvnic_crq newcrq;
4010                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4011                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4012                     &adapter->ip_offload_buf;
4013
4014                 adapter->wait_capability = false;
4015                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4016                                                          buf_sz,
4017                                                          DMA_FROM_DEVICE);
4018
4019                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4020                         if (!firmware_has_feature(FW_FEATURE_CMO))
4021                                 dev_err(dev, "Couldn't map offload buffer\n");
4022                         return;
4023                 }
4024
4025                 memset(&newcrq, 0, sizeof(newcrq));
4026                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4027                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4028                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4029                 newcrq.query_ip_offload.ioba =
4030                     cpu_to_be32(adapter->ip_offload_tok);
4031
4032                 ibmvnic_send_crq(adapter, &newcrq);
4033         }
4034 }
4035
4036 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4037                             struct ibmvnic_adapter *adapter)
4038 {
4039         struct device *dev = &adapter->vdev->dev;
4040         struct net_device *netdev = adapter->netdev;
4041         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4042         struct ibmvnic_login_buffer *login = adapter->login_buf;
4043         int i;
4044
4045         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4046                          DMA_TO_DEVICE);
4047         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4048                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4049
4050         /* If the number of queues requested can't be allocated by the
4051          * server, the login response will return with code 1. We will need
4052          * to resend the login buffer with fewer queues requested.
4053          */
4054         if (login_rsp_crq->generic.rc.code) {
4055                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4056                 complete(&adapter->init_done);
4057                 return 0;
4058         }
4059
4060         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4061
4062         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4063         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4064                 netdev_dbg(adapter->netdev, "%016lx\n",
4065                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4066         }
4067
4068         /* Sanity checks */
4069         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4070             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4071              adapter->req_rx_add_queues !=
4072              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4073                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4074                 ibmvnic_remove(adapter->vdev);
4075                 return -EIO;
4076         }
4077         release_login_buffer(adapter);
4078         complete(&adapter->init_done);
4079
4080         return 0;
4081 }
4082
4083 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4084                                      struct ibmvnic_adapter *adapter)
4085 {
4086         struct device *dev = &adapter->vdev->dev;
4087         long rc;
4088
4089         rc = crq->request_unmap_rsp.rc.code;
4090         if (rc)
4091                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4092 }
4093
4094 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4095                                  struct ibmvnic_adapter *adapter)
4096 {
4097         struct net_device *netdev = adapter->netdev;
4098         struct device *dev = &adapter->vdev->dev;
4099         long rc;
4100
4101         rc = crq->query_map_rsp.rc.code;
4102         if (rc) {
4103                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4104                 return;
4105         }
4106         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4107                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4108                    crq->query_map_rsp.free_pages);
4109 }
4110
4111 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4112                                  struct ibmvnic_adapter *adapter)
4113 {
4114         struct net_device *netdev = adapter->netdev;
4115         struct device *dev = &adapter->vdev->dev;
4116         long rc;
4117
4118         atomic_dec(&adapter->running_cap_crqs);
4119         netdev_dbg(netdev, "Outstanding queries: %d\n",
4120                    atomic_read(&adapter->running_cap_crqs));
4121         rc = crq->query_capability.rc.code;
4122         if (rc) {
4123                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4124                 goto out;
4125         }
4126
4127         switch (be16_to_cpu(crq->query_capability.capability)) {
4128         case MIN_TX_QUEUES:
4129                 adapter->min_tx_queues =
4130                     be64_to_cpu(crq->query_capability.number);
4131                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4132                            adapter->min_tx_queues);
4133                 break;
4134         case MIN_RX_QUEUES:
4135                 adapter->min_rx_queues =
4136                     be64_to_cpu(crq->query_capability.number);
4137                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4138                            adapter->min_rx_queues);
4139                 break;
4140         case MIN_RX_ADD_QUEUES:
4141                 adapter->min_rx_add_queues =
4142                     be64_to_cpu(crq->query_capability.number);
4143                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4144                            adapter->min_rx_add_queues);
4145                 break;
4146         case MAX_TX_QUEUES:
4147                 adapter->max_tx_queues =
4148                     be64_to_cpu(crq->query_capability.number);
4149                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4150                            adapter->max_tx_queues);
4151                 break;
4152         case MAX_RX_QUEUES:
4153                 adapter->max_rx_queues =
4154                     be64_to_cpu(crq->query_capability.number);
4155                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4156                            adapter->max_rx_queues);
4157                 break;
4158         case MAX_RX_ADD_QUEUES:
4159                 adapter->max_rx_add_queues =
4160                     be64_to_cpu(crq->query_capability.number);
4161                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4162                            adapter->max_rx_add_queues);
4163                 break;
4164         case MIN_TX_ENTRIES_PER_SUBCRQ:
4165                 adapter->min_tx_entries_per_subcrq =
4166                     be64_to_cpu(crq->query_capability.number);
4167                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4168                            adapter->min_tx_entries_per_subcrq);
4169                 break;
4170         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4171                 adapter->min_rx_add_entries_per_subcrq =
4172                     be64_to_cpu(crq->query_capability.number);
4173                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4174                            adapter->min_rx_add_entries_per_subcrq);
4175                 break;
4176         case MAX_TX_ENTRIES_PER_SUBCRQ:
4177                 adapter->max_tx_entries_per_subcrq =
4178                     be64_to_cpu(crq->query_capability.number);
4179                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4180                            adapter->max_tx_entries_per_subcrq);
4181                 break;
4182         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4183                 adapter->max_rx_add_entries_per_subcrq =
4184                     be64_to_cpu(crq->query_capability.number);
4185                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4186                            adapter->max_rx_add_entries_per_subcrq);
4187                 break;
4188         case TCP_IP_OFFLOAD:
4189                 adapter->tcp_ip_offload =
4190                     be64_to_cpu(crq->query_capability.number);
4191                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4192                            adapter->tcp_ip_offload);
4193                 break;
4194         case PROMISC_SUPPORTED:
4195                 adapter->promisc_supported =
4196                     be64_to_cpu(crq->query_capability.number);
4197                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4198                            adapter->promisc_supported);
4199                 break;
4200         case MIN_MTU:
4201                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4202                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4203                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4204                 break;
4205         case MAX_MTU:
4206                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4207                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4208                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4209                 break;
4210         case MAX_MULTICAST_FILTERS:
4211                 adapter->max_multicast_filters =
4212                     be64_to_cpu(crq->query_capability.number);
4213                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4214                            adapter->max_multicast_filters);
4215                 break;
4216         case VLAN_HEADER_INSERTION:
4217                 adapter->vlan_header_insertion =
4218                     be64_to_cpu(crq->query_capability.number);
4219                 if (adapter->vlan_header_insertion)
4220                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4221                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4222                            adapter->vlan_header_insertion);
4223                 break;
4224         case RX_VLAN_HEADER_INSERTION:
4225                 adapter->rx_vlan_header_insertion =
4226                     be64_to_cpu(crq->query_capability.number);
4227                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4228                            adapter->rx_vlan_header_insertion);
4229                 break;
4230         case MAX_TX_SG_ENTRIES:
4231                 adapter->max_tx_sg_entries =
4232                     be64_to_cpu(crq->query_capability.number);
4233                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4234                            adapter->max_tx_sg_entries);
4235                 break;
4236         case RX_SG_SUPPORTED:
4237                 adapter->rx_sg_supported =
4238                     be64_to_cpu(crq->query_capability.number);
4239                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4240                            adapter->rx_sg_supported);
4241                 break;
4242         case OPT_TX_COMP_SUB_QUEUES:
4243                 adapter->opt_tx_comp_sub_queues =
4244                     be64_to_cpu(crq->query_capability.number);
4245                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4246                            adapter->opt_tx_comp_sub_queues);
4247                 break;
4248         case OPT_RX_COMP_QUEUES:
4249                 adapter->opt_rx_comp_queues =
4250                     be64_to_cpu(crq->query_capability.number);
4251                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4252                            adapter->opt_rx_comp_queues);
4253                 break;
4254         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4255                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4256                     be64_to_cpu(crq->query_capability.number);
4257                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4258                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4259                 break;
4260         case OPT_TX_ENTRIES_PER_SUBCRQ:
4261                 adapter->opt_tx_entries_per_subcrq =
4262                     be64_to_cpu(crq->query_capability.number);
4263                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4264                            adapter->opt_tx_entries_per_subcrq);
4265                 break;
4266         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4267                 adapter->opt_rxba_entries_per_subcrq =
4268                     be64_to_cpu(crq->query_capability.number);
4269                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4270                            adapter->opt_rxba_entries_per_subcrq);
4271                 break;
4272         case TX_RX_DESC_REQ:
4273                 adapter->tx_rx_desc_req = crq->query_capability.number;
4274                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4275                            adapter->tx_rx_desc_req);
4276                 break;
4277
4278         default:
4279                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4280                            crq->query_capability.capability);
4281         }
4282
4283 out:
4284         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4285                 adapter->wait_capability = false;
4286                 ibmvnic_send_req_caps(adapter, 0);
4287         }
4288 }
4289
4290 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4291 {
4292         union ibmvnic_crq crq;
4293         int rc;
4294
4295         memset(&crq, 0, sizeof(crq));
4296         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4297         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4298         init_completion(&adapter->fw_done);
4299         rc = ibmvnic_send_crq(adapter, &crq);
4300         if (rc)
4301                 return rc;
4302         wait_for_completion(&adapter->fw_done);
4303         return adapter->fw_done_rc ? -EIO : 0;
4304 }
4305
4306 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4307                                        struct ibmvnic_adapter *adapter)
4308 {
4309         struct net_device *netdev = adapter->netdev;
4310         int rc;
4311
4312         rc = crq->query_phys_parms_rsp.rc.code;
4313         if (rc) {
4314                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4315                 return rc;
4316         }
4317         switch (cpu_to_be32(crq->query_phys_parms_rsp.speed)) {
4318         case IBMVNIC_10MBPS:
4319                 adapter->speed = SPEED_10;
4320                 break;
4321         case IBMVNIC_100MBPS:
4322                 adapter->speed = SPEED_100;
4323                 break;
4324         case IBMVNIC_1GBPS:
4325                 adapter->speed = SPEED_1000;
4326                 break;
4327         case IBMVNIC_10GBP:
4328                 adapter->speed = SPEED_10000;
4329                 break;
4330         case IBMVNIC_25GBPS:
4331                 adapter->speed = SPEED_25000;
4332                 break;
4333         case IBMVNIC_40GBPS:
4334                 adapter->speed = SPEED_40000;
4335                 break;
4336         case IBMVNIC_50GBPS:
4337                 adapter->speed = SPEED_50000;
4338                 break;
4339         case IBMVNIC_100GBPS:
4340                 adapter->speed = SPEED_100000;
4341                 break;
4342         default:
4343                 netdev_warn(netdev, "Unknown speed 0x%08x\n",
4344                             cpu_to_be32(crq->query_phys_parms_rsp.speed));
4345                 adapter->speed = SPEED_UNKNOWN;
4346         }
4347         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4348                 adapter->duplex = DUPLEX_FULL;
4349         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4350                 adapter->duplex = DUPLEX_HALF;
4351         else
4352                 adapter->duplex = DUPLEX_UNKNOWN;
4353
4354         return rc;
4355 }
4356
4357 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4358                                struct ibmvnic_adapter *adapter)
4359 {
4360         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4361         struct net_device *netdev = adapter->netdev;
4362         struct device *dev = &adapter->vdev->dev;
4363         u64 *u64_crq = (u64 *)crq;
4364         long rc;
4365
4366         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4367                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4368                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4369         switch (gen_crq->first) {
4370         case IBMVNIC_CRQ_INIT_RSP:
4371                 switch (gen_crq->cmd) {
4372                 case IBMVNIC_CRQ_INIT:
4373                         dev_info(dev, "Partner initialized\n");
4374                         adapter->from_passive_init = true;
4375                         adapter->failover_pending = false;
4376                         if (!completion_done(&adapter->init_done)) {
4377                                 complete(&adapter->init_done);
4378                                 adapter->init_done_rc = -EIO;
4379                         }
4380                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4381                         break;
4382                 case IBMVNIC_CRQ_INIT_COMPLETE:
4383                         dev_info(dev, "Partner initialization complete\n");
4384                         adapter->crq.active = true;
4385                         send_version_xchg(adapter);
4386                         break;
4387                 default:
4388                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4389                 }
4390                 return;
4391         case IBMVNIC_CRQ_XPORT_EVENT:
4392                 netif_carrier_off(netdev);
4393                 adapter->crq.active = false;
4394                 if (adapter->resetting)
4395                         adapter->force_reset_recovery = true;
4396                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4397                         dev_info(dev, "Migrated, re-enabling adapter\n");
4398                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4399                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4400                         dev_info(dev, "Backing device failover detected\n");
4401                         adapter->failover_pending = true;
4402                 } else {
4403                         /* The adapter lost the connection */
4404                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4405                                 gen_crq->cmd);
4406                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4407                 }
4408                 return;
4409         case IBMVNIC_CRQ_CMD_RSP:
4410                 break;
4411         default:
4412                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4413                         gen_crq->first);
4414                 return;
4415         }
4416
4417         switch (gen_crq->cmd) {
4418         case VERSION_EXCHANGE_RSP:
4419                 rc = crq->version_exchange_rsp.rc.code;
4420                 if (rc) {
4421                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4422                         break;
4423                 }
4424                 dev_info(dev, "Partner protocol version is %d\n",
4425                          crq->version_exchange_rsp.version);
4426                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4427                     ibmvnic_version)
4428                         ibmvnic_version =
4429                             be16_to_cpu(crq->version_exchange_rsp.version);
4430                 send_cap_queries(adapter);
4431                 break;
4432         case QUERY_CAPABILITY_RSP:
4433                 handle_query_cap_rsp(crq, adapter);
4434                 break;
4435         case QUERY_MAP_RSP:
4436                 handle_query_map_rsp(crq, adapter);
4437                 break;
4438         case REQUEST_MAP_RSP:
4439                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4440                 complete(&adapter->fw_done);
4441                 break;
4442         case REQUEST_UNMAP_RSP:
4443                 handle_request_unmap_rsp(crq, adapter);
4444                 break;
4445         case REQUEST_CAPABILITY_RSP:
4446                 handle_request_cap_rsp(crq, adapter);
4447                 break;
4448         case LOGIN_RSP:
4449                 netdev_dbg(netdev, "Got Login Response\n");
4450                 handle_login_rsp(crq, adapter);
4451                 break;
4452         case LOGICAL_LINK_STATE_RSP:
4453                 netdev_dbg(netdev,
4454                            "Got Logical Link State Response, state: %d rc: %d\n",
4455                            crq->logical_link_state_rsp.link_state,
4456                            crq->logical_link_state_rsp.rc.code);
4457                 adapter->logical_link_state =
4458                     crq->logical_link_state_rsp.link_state;
4459                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4460                 complete(&adapter->init_done);
4461                 break;
4462         case LINK_STATE_INDICATION:
4463                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4464                 adapter->phys_link_state =
4465                     crq->link_state_indication.phys_link_state;
4466                 adapter->logical_link_state =
4467                     crq->link_state_indication.logical_link_state;
4468                 if (adapter->phys_link_state && adapter->logical_link_state)
4469                         netif_carrier_on(netdev);
4470                 else
4471                         netif_carrier_off(netdev);
4472                 break;
4473         case CHANGE_MAC_ADDR_RSP:
4474                 netdev_dbg(netdev, "Got MAC address change Response\n");
4475                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4476                 break;
4477         case ERROR_INDICATION:
4478                 netdev_dbg(netdev, "Got Error Indication\n");
4479                 handle_error_indication(crq, adapter);
4480                 break;
4481         case REQUEST_STATISTICS_RSP:
4482                 netdev_dbg(netdev, "Got Statistics Response\n");
4483                 complete(&adapter->stats_done);
4484                 break;
4485         case QUERY_IP_OFFLOAD_RSP:
4486                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4487                 handle_query_ip_offload_rsp(adapter);
4488                 break;
4489         case MULTICAST_CTRL_RSP:
4490                 netdev_dbg(netdev, "Got multicast control Response\n");
4491                 break;
4492         case CONTROL_IP_OFFLOAD_RSP:
4493                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4494                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4495                                  sizeof(adapter->ip_offload_ctrl),
4496                                  DMA_TO_DEVICE);
4497                 complete(&adapter->init_done);
4498                 break;
4499         case COLLECT_FW_TRACE_RSP:
4500                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4501                 complete(&adapter->fw_done);
4502                 break;
4503         case GET_VPD_SIZE_RSP:
4504                 handle_vpd_size_rsp(crq, adapter);
4505                 break;
4506         case GET_VPD_RSP:
4507                 handle_vpd_rsp(crq, adapter);
4508                 break;
4509         case QUERY_PHYS_PARMS_RSP:
4510                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4511                 complete(&adapter->fw_done);
4512                 break;
4513         default:
4514                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4515                            gen_crq->cmd);
4516         }
4517 }
4518
4519 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4520 {
4521         struct ibmvnic_adapter *adapter = instance;
4522
4523         tasklet_schedule(&adapter->tasklet);
4524         return IRQ_HANDLED;
4525 }
4526
4527 static void ibmvnic_tasklet(void *data)
4528 {
4529         struct ibmvnic_adapter *adapter = data;
4530         struct ibmvnic_crq_queue *queue = &adapter->crq;
4531         union ibmvnic_crq *crq;
4532         unsigned long flags;
4533         bool done = false;
4534
4535         spin_lock_irqsave(&queue->lock, flags);
4536         while (!done) {
4537                 /* Pull all the valid messages off the CRQ */
4538                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4539                         ibmvnic_handle_crq(crq, adapter);
4540                         crq->generic.first = 0;
4541                 }
4542
4543                 /* remain in tasklet until all
4544                  * capabilities responses are received
4545                  */
4546                 if (!adapter->wait_capability)
4547                         done = true;
4548         }
4549         /* if capabilities CRQ's were sent in this tasklet, the following
4550          * tasklet must wait until all responses are received
4551          */
4552         if (atomic_read(&adapter->running_cap_crqs) != 0)
4553                 adapter->wait_capability = true;
4554         spin_unlock_irqrestore(&queue->lock, flags);
4555 }
4556
4557 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4558 {
4559         struct vio_dev *vdev = adapter->vdev;
4560         int rc;
4561
4562         do {
4563                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4564         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4565
4566         if (rc)
4567                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4568
4569         return rc;
4570 }
4571
4572 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4573 {
4574         struct ibmvnic_crq_queue *crq = &adapter->crq;
4575         struct device *dev = &adapter->vdev->dev;
4576         struct vio_dev *vdev = adapter->vdev;
4577         int rc;
4578
4579         /* Close the CRQ */
4580         do {
4581                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4582         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4583
4584         /* Clean out the queue */
4585         memset(crq->msgs, 0, PAGE_SIZE);
4586         crq->cur = 0;
4587         crq->active = false;
4588
4589         /* And re-open it again */
4590         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4591                                 crq->msg_token, PAGE_SIZE);
4592
4593         if (rc == H_CLOSED)
4594                 /* Adapter is good, but other end is not ready */
4595                 dev_warn(dev, "Partner adapter not ready\n");
4596         else if (rc != 0)
4597                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4598
4599         return rc;
4600 }
4601
4602 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4603 {
4604         struct ibmvnic_crq_queue *crq = &adapter->crq;
4605         struct vio_dev *vdev = adapter->vdev;
4606         long rc;
4607
4608         if (!crq->msgs)
4609                 return;
4610
4611         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4612         free_irq(vdev->irq, adapter);
4613         tasklet_kill(&adapter->tasklet);
4614         do {
4615                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4616         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4617
4618         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4619                          DMA_BIDIRECTIONAL);
4620         free_page((unsigned long)crq->msgs);
4621         crq->msgs = NULL;
4622         crq->active = false;
4623 }
4624
4625 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4626 {
4627         struct ibmvnic_crq_queue *crq = &adapter->crq;
4628         struct device *dev = &adapter->vdev->dev;
4629         struct vio_dev *vdev = adapter->vdev;
4630         int rc, retrc = -ENOMEM;
4631
4632         if (crq->msgs)
4633                 return 0;
4634
4635         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4636         /* Should we allocate more than one page? */
4637
4638         if (!crq->msgs)
4639                 return -ENOMEM;
4640
4641         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4642         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4643                                         DMA_BIDIRECTIONAL);
4644         if (dma_mapping_error(dev, crq->msg_token))
4645                 goto map_failed;
4646
4647         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4648                                 crq->msg_token, PAGE_SIZE);
4649
4650         if (rc == H_RESOURCE)
4651                 /* maybe kexecing and resource is busy. try a reset */
4652                 rc = ibmvnic_reset_crq(adapter);
4653         retrc = rc;
4654
4655         if (rc == H_CLOSED) {
4656                 dev_warn(dev, "Partner adapter not ready\n");
4657         } else if (rc) {
4658                 dev_warn(dev, "Error %d opening adapter\n", rc);
4659                 goto reg_crq_failed;
4660         }
4661
4662         retrc = 0;
4663
4664         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4665                      (unsigned long)adapter);
4666
4667         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4668         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4669                  adapter->vdev->unit_address);
4670         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4671         if (rc) {
4672                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4673                         vdev->irq, rc);
4674                 goto req_irq_failed;
4675         }
4676
4677         rc = vio_enable_interrupts(vdev);
4678         if (rc) {
4679                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4680                 goto req_irq_failed;
4681         }
4682
4683         crq->cur = 0;
4684         spin_lock_init(&crq->lock);
4685
4686         return retrc;
4687
4688 req_irq_failed:
4689         tasklet_kill(&adapter->tasklet);
4690         do {
4691                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4692         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4693 reg_crq_failed:
4694         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4695 map_failed:
4696         free_page((unsigned long)crq->msgs);
4697         crq->msgs = NULL;
4698         return retrc;
4699 }
4700
4701 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4702 {
4703         struct device *dev = &adapter->vdev->dev;
4704         unsigned long timeout = msecs_to_jiffies(30000);
4705         u64 old_num_rx_queues, old_num_tx_queues;
4706         int rc;
4707
4708         adapter->from_passive_init = false;
4709
4710         old_num_rx_queues = adapter->req_rx_queues;
4711         old_num_tx_queues = adapter->req_tx_queues;
4712
4713         reinit_completion(&adapter->init_done);
4714         adapter->init_done_rc = 0;
4715         ibmvnic_send_crq_init(adapter);
4716         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4717                 dev_err(dev, "Initialization sequence timed out\n");
4718                 return -1;
4719         }
4720
4721         if (adapter->init_done_rc) {
4722                 release_crq_queue(adapter);
4723                 return adapter->init_done_rc;
4724         }
4725
4726         if (adapter->from_passive_init) {
4727                 adapter->state = VNIC_OPEN;
4728                 adapter->from_passive_init = false;
4729                 return -1;
4730         }
4731
4732         if (adapter->resetting && !adapter->wait_for_reset &&
4733             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4734                 if (adapter->req_rx_queues != old_num_rx_queues ||
4735                     adapter->req_tx_queues != old_num_tx_queues) {
4736                         release_sub_crqs(adapter, 0);
4737                         rc = init_sub_crqs(adapter);
4738                 } else {
4739                         rc = reset_sub_crq_queues(adapter);
4740                 }
4741         } else {
4742                 rc = init_sub_crqs(adapter);
4743         }
4744
4745         if (rc) {
4746                 dev_err(dev, "Initialization of sub crqs failed\n");
4747                 release_crq_queue(adapter);
4748                 return rc;
4749         }
4750
4751         rc = init_sub_crq_irqs(adapter);
4752         if (rc) {
4753                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4754                 release_crq_queue(adapter);
4755         }
4756
4757         return rc;
4758 }
4759
4760 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4761 {
4762         struct device *dev = &adapter->vdev->dev;
4763         unsigned long timeout = msecs_to_jiffies(30000);
4764         int rc;
4765
4766         adapter->from_passive_init = false;
4767
4768         adapter->init_done_rc = 0;
4769         ibmvnic_send_crq_init(adapter);
4770         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4771                 dev_err(dev, "Initialization sequence timed out\n");
4772                 return -1;
4773         }
4774
4775         if (adapter->init_done_rc) {
4776                 release_crq_queue(adapter);
4777                 return adapter->init_done_rc;
4778         }
4779
4780         if (adapter->from_passive_init) {
4781                 adapter->state = VNIC_OPEN;
4782                 adapter->from_passive_init = false;
4783                 return -1;
4784         }
4785
4786         rc = init_sub_crqs(adapter);
4787         if (rc) {
4788                 dev_err(dev, "Initialization of sub crqs failed\n");
4789                 release_crq_queue(adapter);
4790                 return rc;
4791         }
4792
4793         rc = init_sub_crq_irqs(adapter);
4794         if (rc) {
4795                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4796                 release_crq_queue(adapter);
4797         }
4798
4799         return rc;
4800 }
4801
4802 static struct device_attribute dev_attr_failover;
4803
4804 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4805 {
4806         struct ibmvnic_adapter *adapter;
4807         struct net_device *netdev;
4808         unsigned char *mac_addr_p;
4809         int rc;
4810
4811         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4812                 dev->unit_address);
4813
4814         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4815                                                         VETH_MAC_ADDR, NULL);
4816         if (!mac_addr_p) {
4817                 dev_err(&dev->dev,
4818                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4819                         __FILE__, __LINE__);
4820                 return 0;
4821         }
4822
4823         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4824                                    IBMVNIC_MAX_QUEUES);
4825         if (!netdev)
4826                 return -ENOMEM;
4827
4828         adapter = netdev_priv(netdev);
4829         adapter->state = VNIC_PROBING;
4830         dev_set_drvdata(&dev->dev, netdev);
4831         adapter->vdev = dev;
4832         adapter->netdev = netdev;
4833
4834         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4835         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4836         netdev->irq = dev->irq;
4837         netdev->netdev_ops = &ibmvnic_netdev_ops;
4838         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4839         SET_NETDEV_DEV(netdev, &dev->dev);
4840
4841         spin_lock_init(&adapter->stats_lock);
4842
4843         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4844         INIT_LIST_HEAD(&adapter->rwi_list);
4845         spin_lock_init(&adapter->rwi_lock);
4846         init_completion(&adapter->init_done);
4847         adapter->resetting = false;
4848
4849         do {
4850                 rc = init_crq_queue(adapter);
4851                 if (rc) {
4852                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4853                                 rc);
4854                         goto ibmvnic_init_fail;
4855                 }
4856
4857                 rc = ibmvnic_init(adapter);
4858                 if (rc && rc != EAGAIN)
4859                         goto ibmvnic_init_fail;
4860         } while (rc == EAGAIN);
4861
4862         rc = init_stats_buffers(adapter);
4863         if (rc)
4864                 goto ibmvnic_init_fail;
4865
4866         rc = init_stats_token(adapter);
4867         if (rc)
4868                 goto ibmvnic_stats_fail;
4869
4870         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4871         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4872         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4873
4874         rc = device_create_file(&dev->dev, &dev_attr_failover);
4875         if (rc)
4876                 goto ibmvnic_dev_file_err;
4877
4878         netif_carrier_off(netdev);
4879         rc = register_netdev(netdev);
4880         if (rc) {
4881                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4882                 goto ibmvnic_register_fail;
4883         }
4884         dev_info(&dev->dev, "ibmvnic registered\n");
4885
4886         adapter->state = VNIC_PROBED;
4887
4888         adapter->wait_for_reset = false;
4889
4890         return 0;
4891
4892 ibmvnic_register_fail:
4893         device_remove_file(&dev->dev, &dev_attr_failover);
4894
4895 ibmvnic_dev_file_err:
4896         release_stats_token(adapter);
4897
4898 ibmvnic_stats_fail:
4899         release_stats_buffers(adapter);
4900
4901 ibmvnic_init_fail:
4902         release_sub_crqs(adapter, 1);
4903         release_crq_queue(adapter);
4904         free_netdev(netdev);
4905
4906         return rc;
4907 }
4908
4909 static int ibmvnic_remove(struct vio_dev *dev)
4910 {
4911         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4912         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4913
4914         adapter->state = VNIC_REMOVING;
4915         rtnl_lock();
4916         unregister_netdevice(netdev);
4917
4918         release_resources(adapter);
4919         release_sub_crqs(adapter, 1);
4920         release_crq_queue(adapter);
4921
4922         release_stats_token(adapter);
4923         release_stats_buffers(adapter);
4924
4925         adapter->state = VNIC_REMOVED;
4926
4927         rtnl_unlock();
4928         device_remove_file(&dev->dev, &dev_attr_failover);
4929         free_netdev(netdev);
4930         dev_set_drvdata(&dev->dev, NULL);
4931
4932         return 0;
4933 }
4934
4935 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4936                               const char *buf, size_t count)
4937 {
4938         struct net_device *netdev = dev_get_drvdata(dev);
4939         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4940         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4941         __be64 session_token;
4942         long rc;
4943
4944         if (!sysfs_streq(buf, "1"))
4945                 return -EINVAL;
4946
4947         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4948                          H_GET_SESSION_TOKEN, 0, 0, 0);
4949         if (rc) {
4950                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4951                            rc);
4952                 return -EINVAL;
4953         }
4954
4955         session_token = (__be64)retbuf[0];
4956         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4957                    be64_to_cpu(session_token));
4958         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4959                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4960         if (rc) {
4961                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4962                            rc);
4963                 return -EINVAL;
4964         }
4965
4966         return count;
4967 }
4968
4969 static DEVICE_ATTR_WO(failover);
4970
4971 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4972 {
4973         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4974         struct ibmvnic_adapter *adapter;
4975         struct iommu_table *tbl;
4976         unsigned long ret = 0;
4977         int i;
4978
4979         tbl = get_iommu_table_base(&vdev->dev);
4980
4981         /* netdev inits at probe time along with the structures we need below*/
4982         if (!netdev)
4983                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4984
4985         adapter = netdev_priv(netdev);
4986
4987         ret += PAGE_SIZE; /* the crq message queue */
4988         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4989
4990         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4991                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4992
4993         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4994              i++)
4995                 ret += adapter->rx_pool[i].size *
4996                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4997
4998         return ret;
4999 }
5000
5001 static int ibmvnic_resume(struct device *dev)
5002 {
5003         struct net_device *netdev = dev_get_drvdata(dev);
5004         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5005
5006         if (adapter->state != VNIC_OPEN)
5007                 return 0;
5008
5009         tasklet_schedule(&adapter->tasklet);
5010
5011         return 0;
5012 }
5013
5014 static const struct vio_device_id ibmvnic_device_table[] = {
5015         {"network", "IBM,vnic"},
5016         {"", "" }
5017 };
5018 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5019
5020 static const struct dev_pm_ops ibmvnic_pm_ops = {
5021         .resume = ibmvnic_resume
5022 };
5023
5024 static struct vio_driver ibmvnic_driver = {
5025         .id_table       = ibmvnic_device_table,
5026         .probe          = ibmvnic_probe,
5027         .remove         = ibmvnic_remove,
5028         .get_desired_dma = ibmvnic_get_desired_dma,
5029         .name           = ibmvnic_driver_name,
5030         .pm             = &ibmvnic_pm_ops,
5031 };
5032
5033 /* module functions */
5034 static int __init ibmvnic_module_init(void)
5035 {
5036         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5037                 IBMVNIC_DRIVER_VERSION);
5038
5039         return vio_register_driver(&ibmvnic_driver);
5040 }
5041
5042 static void __exit ibmvnic_module_exit(void)
5043 {
5044         vio_unregister_driver(&ibmvnic_driver);
5045 }
5046
5047 module_init(ibmvnic_module_init);
5048 module_exit(ibmvnic_module_exit);