drivers/net/ethernet/ibm/ibmvnic.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /**************************************************************************/
   3 /*                                                                        */
   4 /*  IBM System i and System p Virtual NIC Device Driver                   */
   5 /*  Copyright (C) 2014 IBM Corp.                                          */
   6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   9 /*                                                                        */
  10 /*                                                                        */
  11 /* This module contains the implementation of a virtual ethernet device   */
  12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  15 /*                                                                         */
  16 /* Messages are passed between the VNIC driver and the VNIC server using  */
  17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  18 /* issue and receive commands that initiate communication with the server */
  19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  20 /* are used by the driver to notify the server that a packet is           */
  21 /* ready for transmission or that a buffer has been added to receive a    */
  22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  23 /* driver that a packet transmission has been completed or that a packet  */
  24 /* has been received and placed in a waiting buffer.                      */
  25 /*                                                                        */
  26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  28 /* or receive has been completed, the VNIC driver is required to use      */
  29 /* "long term mapping". This entails that large, continuous DMA mapped    */
  30 /* buffers are allocated on driver initialization and these buffers are   */
  31 /* then continuously reused to pass skbs to and from the VNIC server.     */
  32 /*                                                                        */
  33 /**************************************************************************/
  34
  35 #include <linux/module.h>
  36 #include <linux/moduleparam.h>
  37 #include <linux/types.h>
  38 #include <linux/errno.h>
  39 #include <linux/completion.h>
  40 #include <linux/ioport.h>
  41 #include <linux/dma-mapping.h>
  42 #include <linux/kernel.h>
  43 #include <linux/netdevice.h>
  44 #include <linux/etherdevice.h>
  45 #include <linux/skbuff.h>
  46 #include <linux/init.h>
  47 #include <linux/delay.h>
  48 #include <linux/mm.h>
  49 #include <linux/ethtool.h>
  50 #include <linux/proc_fs.h>
  51 #include <linux/if_arp.h>
  52 #include <linux/in.h>
  53 #include <linux/ip.h>
  54 #include <linux/ipv6.h>
  55 #include <linux/irq.h>
  56 #include <linux/kthread.h>
  57 #include <linux/seq_file.h>
  58 #include <linux/interrupt.h>
  59 #include <net/net_namespace.h>
  60 #include <asm/hvcall.h>
  61 #include <linux/atomic.h>
  62 #include <asm/vio.h>
  63 #include <asm/iommu.h>
  64 #include <linux/uaccess.h>
  65 #include <asm/firmware.h>
  66 #include <linux/workqueue.h>
  67 #include <linux/if_vlan.h>
  68 #include <linux/utsname.h>
  69
  70 #include "ibmvnic.h"
  71
  72 static const char ibmvnic_driver_name[] = "ibmvnic";
  73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  74
  75 MODULE_AUTHOR("Santiago Leon");
  76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  77 MODULE_LICENSE("GPL");
  78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  79
  80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  81 static int ibmvnic_remove(struct vio_dev *);
  82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  88                        union sub_crq *sub_crq);
  89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
  90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
  91 static int enable_scrq_irq(struct ibmvnic_adapter *,
  92                            struct ibmvnic_sub_crq_queue *);
  93 static int disable_scrq_irq(struct ibmvnic_adapter *,
  94                             struct ibmvnic_sub_crq_queue *);
  95 static int pending_scrq(struct ibmvnic_adapter *,
  96                         struct ibmvnic_sub_crq_queue *);
  97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
  98                                         struct ibmvnic_sub_crq_queue *);
  99 static int ibmvnic_poll(struct napi_struct *napi, int data);
 100 static void send_map_query(struct ibmvnic_adapter *adapter);
 101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
 103 static int send_login(struct ibmvnic_adapter *adapter);
 104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 105 static int init_sub_crqs(struct ibmvnic_adapter *);
 106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 107 static int ibmvnic_init(struct ibmvnic_adapter *);
 108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
 109 static void release_crq_queue(struct ibmvnic_adapter *);
 110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
 111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
 113
 114 struct ibmvnic_stat {
 115         char name[ETH_GSTRING_LEN];
 116         int offset;
 117 };
 118
 119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 120                              offsetof(struct ibmvnic_statistics, stat))
 121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 122
 123 static const struct ibmvnic_stat ibmvnic_stats[] = {
 124         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 125         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 126         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 127         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 128         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 129         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 130         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 131         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 132         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 133         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 134         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 135         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 136         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 137         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 138         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 139         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 140         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 141         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 142         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 143         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 144         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 145         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 146 };
 147
 148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 149                           unsigned long length, unsigned long *number,
 150                           unsigned long *irq)
 151 {
 152         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 153         long rc;
 154
 155         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 156         *number = retbuf[0];
 157         *irq = retbuf[1];
 158
 159         return rc;
 160 }
 161
 162 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 163                                 struct ibmvnic_long_term_buff *ltb, int size)
 164 {
 165         struct device *dev = &adapter->vdev->dev;
 166         int rc;
 167
 168         ltb->size = size;
 169         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 170                                        GFP_KERNEL);
 171
 172         if (!ltb->buff) {
 173                 dev_err(dev, "Couldn't alloc long term buffer\n");
 174                 return -ENOMEM;
 175         }
 176         ltb->map_id = adapter->map_id;
 177         adapter->map_id++;
 178
 179         init_completion(&adapter->fw_done);
 180         rc = send_request_map(adapter, ltb->addr,
 181                               ltb->size, ltb->map_id);
 182         if (rc) {
 183                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 184                 return rc;
 185         }
 186         wait_for_completion(&adapter->fw_done);
 187
 188         if (adapter->fw_done_rc) {
 189                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 190                         adapter->fw_done_rc);
 191                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 192                 return -1;
 193         }
 194         return 0;
 195 }
 196
 197 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 198                                 struct ibmvnic_long_term_buff *ltb)
 199 {
 200         struct device *dev = &adapter->vdev->dev;
 201
 202         if (!ltb->buff)
 203                 return;
 204
 205         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 206             adapter->reset_reason != VNIC_RESET_MOBILITY)
 207                 send_request_unmap(adapter, ltb->map_id);
 208         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 209 }
 210
 211 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 212                                 struct ibmvnic_long_term_buff *ltb)
 213 {
 214         int rc;
 215
 216         memset(ltb->buff, 0, ltb->size);
 217
 218         init_completion(&adapter->fw_done);
 219         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 220         if (rc)
 221                 return rc;
 222         wait_for_completion(&adapter->fw_done);
 223
 224         if (adapter->fw_done_rc) {
 225                 dev_info(&adapter->vdev->dev,
 226                          "Reset failed, attempting to free and reallocate buffer\n");
 227                 free_long_term_buff(adapter, ltb);
 228                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 229         }
 230         return 0;
 231 }
 232
 233 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 234 {
 235         int i;
 236
 237         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 238              i++)
 239                 adapter->rx_pool[i].active = 0;
 240 }
 241
 242 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 243                               struct ibmvnic_rx_pool *pool)
 244 {
 245         int count = pool->size - atomic_read(&pool->available);
 246         struct device *dev = &adapter->vdev->dev;
 247         int buffers_added = 0;
 248         unsigned long lpar_rc;
 249         union sub_crq sub_crq;
 250         struct sk_buff *skb;
 251         unsigned int offset;
 252         dma_addr_t dma_addr;
 253         unsigned char *dst;
 254         u64 *handle_array;
 255         int shift = 0;
 256         int index;
 257         int i;
 258
 259         if (!pool->active)
 260                 return;
 261
 262         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 263                                       be32_to_cpu(adapter->login_rsp_buf->
 264                                       off_rxadd_subcrqs));
 265
 266         for (i = 0; i < count; ++i) {
 267                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 268                 if (!skb) {
 269                         dev_err(dev, "Couldn't replenish rx buff\n");
 270                         adapter->replenish_no_mem++;
 271                         break;
 272                 }
 273
 274                 index = pool->free_map[pool->next_free];
 275
 276                 if (pool->rx_buff[index].skb)
 277                         dev_err(dev, "Inconsistent free_map!\n");
 278
 279                 /* Copy the skb to the long term mapped DMA buffer */
 280                 offset = index * pool->buff_size;
 281                 dst = pool->long_term_buff.buff + offset;
 282                 memset(dst, 0, pool->buff_size);
 283                 dma_addr = pool->long_term_buff.addr + offset;
 284                 pool->rx_buff[index].data = dst;
 285
 286                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 287                 pool->rx_buff[index].dma = dma_addr;
 288                 pool->rx_buff[index].skb = skb;
 289                 pool->rx_buff[index].pool_index = pool->index;
 290                 pool->rx_buff[index].size = pool->buff_size;
 291
 292                 memset(&sub_crq, 0, sizeof(sub_crq));
 293                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 294                 sub_crq.rx_add.correlator =
 295                     cpu_to_be64((u64)&pool->rx_buff[index]);
 296                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 297                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 298
 299                 /* The length field of the sCRQ is defined to be 24 bits so the
 300                  * buffer size needs to be left shifted by a byte before it is
 301                  * converted to big endian to prevent the last byte from being
 302                  * truncated.
 303                  */
 304 #ifdef __LITTLE_ENDIAN__
 305                 shift = 8;
 306 #endif
 307                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 308
 309                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 310                                       &sub_crq);
 311                 if (lpar_rc != H_SUCCESS)
 312                         goto failure;
 313
 314                 buffers_added++;
 315                 adapter->replenish_add_buff_success++;
 316                 pool->next_free = (pool->next_free + 1) % pool->size;
 317         }
 318         atomic_add(buffers_added, &pool->available);
 319         return;
 320
 321 failure:
 322         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 323                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 324         pool->free_map[pool->next_free] = index;
 325         pool->rx_buff[index].skb = NULL;
 326
 327         dev_kfree_skb_any(skb);
 328         adapter->replenish_add_buff_failure++;
 329         atomic_add(buffers_added, &pool->available);
 330
 331         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 332                 /* Disable buffer pool replenishment and report carrier off if
 333                  * queue is closed or pending failover.
 334                  * Firmware guarantees that a signal will be sent to the
 335                  * driver, triggering a reset.
 336                  */
 337                 deactivate_rx_pools(adapter);
 338                 netif_carrier_off(adapter->netdev);
 339         }
 340 }
 341
 342 static void replenish_pools(struct ibmvnic_adapter *adapter)
 343 {
 344         int i;
 345
 346         adapter->replenish_task_cycles++;
 347         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 348              i++) {
 349                 if (adapter->rx_pool[i].active)
 350                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 351         }
 352 }
 353
 354 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 355 {
 356         kfree(adapter->tx_stats_buffers);
 357         kfree(adapter->rx_stats_buffers);
 358         adapter->tx_stats_buffers = NULL;
 359         adapter->rx_stats_buffers = NULL;
 360 }
 361
 362 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 363 {
 364         adapter->tx_stats_buffers =
 365                                 kcalloc(IBMVNIC_MAX_QUEUES,
 366                                         sizeof(struct ibmvnic_tx_queue_stats),
 367                                         GFP_KERNEL);
 368         if (!adapter->tx_stats_buffers)
 369                 return -ENOMEM;
 370
 371         adapter->rx_stats_buffers =
 372                                 kcalloc(IBMVNIC_MAX_QUEUES,
 373                                         sizeof(struct ibmvnic_rx_queue_stats),
 374                                         GFP_KERNEL);
 375         if (!adapter->rx_stats_buffers)
 376                 return -ENOMEM;
 377
 378         return 0;
 379 }
 380
 381 static void release_stats_token(struct ibmvnic_adapter *adapter)
 382 {
 383         struct device *dev = &adapter->vdev->dev;
 384
 385         if (!adapter->stats_token)
 386                 return;
 387
 388         dma_unmap_single(dev, adapter->stats_token,
 389                          sizeof(struct ibmvnic_statistics),
 390                          DMA_FROM_DEVICE);
 391         adapter->stats_token = 0;
 392 }
 393
 394 static int init_stats_token(struct ibmvnic_adapter *adapter)
 395 {
 396         struct device *dev = &adapter->vdev->dev;
 397         dma_addr_t stok;
 398
 399         stok = dma_map_single(dev, &adapter->stats,
 400                               sizeof(struct ibmvnic_statistics),
 401                               DMA_FROM_DEVICE);
 402         if (dma_mapping_error(dev, stok)) {
 403                 dev_err(dev, "Couldn't map stats buffer\n");
 404                 return -1;
 405         }
 406
 407         adapter->stats_token = stok;
 408         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 409         return 0;
 410 }
 411
 412 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 413 {
 414         struct ibmvnic_rx_pool *rx_pool;
 415         int rx_scrqs;
 416         int i, j, rc;
 417         u64 *size_array;
 418
 419         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 420                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 421
 422         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 423         for (i = 0; i < rx_scrqs; i++) {
 424                 rx_pool = &adapter->rx_pool[i];
 425
 426                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 427
 428                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 429                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 430                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 431                         rc = alloc_long_term_buff(adapter,
 432                                                   &rx_pool->long_term_buff,
 433                                                   rx_pool->size *
 434                                                   rx_pool->buff_size);
 435                 } else {
 436                         rc = reset_long_term_buff(adapter,
 437                                                   &rx_pool->long_term_buff);
 438                 }
 439
 440                 if (rc)
 441                         return rc;
 442
 443                 for (j = 0; j < rx_pool->size; j++)
 444                         rx_pool->free_map[j] = j;
 445
 446                 memset(rx_pool->rx_buff, 0,
 447                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 448
 449                 atomic_set(&rx_pool->available, 0);
 450                 rx_pool->next_alloc = 0;
 451                 rx_pool->next_free = 0;
 452                 rx_pool->active = 1;
 453         }
 454
 455         return 0;
 456 }
 457
 458 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 459 {
 460         struct ibmvnic_rx_pool *rx_pool;
 461         int i, j;
 462
 463         if (!adapter->rx_pool)
 464                 return;
 465
 466         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 467                 rx_pool = &adapter->rx_pool[i];
 468
 469                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 470
 471                 kfree(rx_pool->free_map);
 472                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 473
 474                 if (!rx_pool->rx_buff)
 475                         continue;
 476
 477                 for (j = 0; j < rx_pool->size; j++) {
 478                         if (rx_pool->rx_buff[j].skb) {
 479                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 480                                 rx_pool->rx_buff[j].skb = NULL;
 481                         }
 482                 }
 483
 484                 kfree(rx_pool->rx_buff);
 485         }
 486
 487         kfree(adapter->rx_pool);
 488         adapter->rx_pool = NULL;
 489         adapter->num_active_rx_pools = 0;
 490 }
 491
 492 static int init_rx_pools(struct net_device *netdev)
 493 {
 494         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 495         struct device *dev = &adapter->vdev->dev;
 496         struct ibmvnic_rx_pool *rx_pool;
 497         int rxadd_subcrqs;
 498         u64 *size_array;
 499         int i, j;
 500
 501         rxadd_subcrqs =
 502                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 503         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 504                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 505
 506         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 507                                    sizeof(struct ibmvnic_rx_pool),
 508                                    GFP_KERNEL);
 509         if (!adapter->rx_pool) {
 510                 dev_err(dev, "Failed to allocate rx pools\n");
 511                 return -1;
 512         }
 513
 514         adapter->num_active_rx_pools = rxadd_subcrqs;
 515
 516         for (i = 0; i < rxadd_subcrqs; i++) {
 517                 rx_pool = &adapter->rx_pool[i];
 518
 519                 netdev_dbg(adapter->netdev,
 520                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 521                            i, adapter->req_rx_add_entries_per_subcrq,
 522                            be64_to_cpu(size_array[i]));
 523
 524                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 525                 rx_pool->index = i;
 526                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 527                 rx_pool->active = 1;
 528
 529                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 530                                             GFP_KERNEL);
 531                 if (!rx_pool->free_map) {
 532                         release_rx_pools(adapter);
 533                         return -1;
 534                 }
 535
 536                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 537                                            sizeof(struct ibmvnic_rx_buff),
 538                                            GFP_KERNEL);
 539                 if (!rx_pool->rx_buff) {
 540                         dev_err(dev, "Couldn't alloc rx buffers\n");
 541                         release_rx_pools(adapter);
 542                         return -1;
 543                 }
 544
 545                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 546                                          rx_pool->size * rx_pool->buff_size)) {
 547                         release_rx_pools(adapter);
 548                         return -1;
 549                 }
 550
 551                 for (j = 0; j < rx_pool->size; ++j)
 552                         rx_pool->free_map[j] = j;
 553
 554                 atomic_set(&rx_pool->available, 0);
 555                 rx_pool->next_alloc = 0;
 556                 rx_pool->next_free = 0;
 557         }
 558
 559         return 0;
 560 }
 561
 562 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 563                              struct ibmvnic_tx_pool *tx_pool)
 564 {
 565         int rc, i;
 566
 567         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 568         if (rc)
 569                 return rc;
 570
 571         memset(tx_pool->tx_buff, 0,
 572                tx_pool->num_buffers *
 573                sizeof(struct ibmvnic_tx_buff));
 574
 575         for (i = 0; i < tx_pool->num_buffers; i++)
 576                 tx_pool->free_map[i] = i;
 577
 578         tx_pool->consumer_index = 0;
 579         tx_pool->producer_index = 0;
 580
 581         return 0;
 582 }
 583
 584 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 585 {
 586         int tx_scrqs;
 587         int i, rc;
 588
 589         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 590         for (i = 0; i < tx_scrqs; i++) {
 591                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 592                 if (rc)
 593                         return rc;
 594                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 595                 if (rc)
 596                         return rc;
 597         }
 598
 599         return 0;
 600 }
 601
 602 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 603 {
 604         if (!adapter->vpd)
 605                 return;
 606
 607         kfree(adapter->vpd->buff);
 608         kfree(adapter->vpd);
 609
 610         adapter->vpd = NULL;
 611 }
 612
 613 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 614                                 struct ibmvnic_tx_pool *tx_pool)
 615 {
 616         kfree(tx_pool->tx_buff);
 617         kfree(tx_pool->free_map);
 618         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 619 }
 620
 621 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 622 {
 623         int i;
 624
 625         if (!adapter->tx_pool)
 626                 return;
 627
 628         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 629                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 630                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 631         }
 632
 633         kfree(adapter->tx_pool);
 634         adapter->tx_pool = NULL;
 635         kfree(adapter->tso_pool);
 636         adapter->tso_pool = NULL;
 637         adapter->num_active_tx_pools = 0;
 638 }
 639
 640 static int init_one_tx_pool(struct net_device *netdev,
 641                             struct ibmvnic_tx_pool *tx_pool,
 642                             int num_entries, int buf_size)
 643 {
 644         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 645         int i;
 646
 647         tx_pool->tx_buff = kcalloc(num_entries,
 648                                    sizeof(struct ibmvnic_tx_buff),
 649                                    GFP_KERNEL);
 650         if (!tx_pool->tx_buff)
 651                 return -1;
 652
 653         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 654                                  num_entries * buf_size))
 655                 return -1;
 656
 657         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 658         if (!tx_pool->free_map)
 659                 return -1;
 660
 661         for (i = 0; i < num_entries; i++)
 662                 tx_pool->free_map[i] = i;
 663
 664         tx_pool->consumer_index = 0;
 665         tx_pool->producer_index = 0;
 666         tx_pool->num_buffers = num_entries;
 667         tx_pool->buf_size = buf_size;
 668
 669         return 0;
 670 }
 671
 672 static int init_tx_pools(struct net_device *netdev)
 673 {
 674         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 675         int tx_subcrqs;
 676         int i, rc;
 677
 678         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 679         adapter->tx_pool = kcalloc(tx_subcrqs,
 680                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 681         if (!adapter->tx_pool)
 682                 return -1;
 683
 684         adapter->tso_pool = kcalloc(tx_subcrqs,
 685                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 686         if (!adapter->tso_pool)
 687                 return -1;
 688
 689         adapter->num_active_tx_pools = tx_subcrqs;
 690
 691         for (i = 0; i < tx_subcrqs; i++) {
 692                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 693                                       adapter->req_tx_entries_per_subcrq,
 694                                       adapter->req_mtu + VLAN_HLEN);
 695                 if (rc) {
 696                         release_tx_pools(adapter);
 697                         return rc;
 698                 }
 699
 700                 rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
 701                                       IBMVNIC_TSO_BUFS,
 702                                       IBMVNIC_TSO_BUF_SZ);
 703                 if (rc) {
 704                         release_tx_pools(adapter);
 705                         return rc;
 706                 }
 707         }
 708
 709         return 0;
 710 }
 711
 712 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 713 {
 714         int i;
 715
 716         if (adapter->napi_enabled)
 717                 return;
 718
 719         for (i = 0; i < adapter->req_rx_queues; i++)
 720                 napi_enable(&adapter->napi[i]);
 721
 722         adapter->napi_enabled = true;
 723 }
 724
 725 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 726 {
 727         int i;
 728
 729         if (!adapter->napi_enabled)
 730                 return;
 731
 732         for (i = 0; i < adapter->req_rx_queues; i++) {
 733                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 734                 napi_disable(&adapter->napi[i]);
 735         }
 736
 737         adapter->napi_enabled = false;
 738 }
 739
 740 static int init_napi(struct ibmvnic_adapter *adapter)
 741 {
 742         int i;
 743
 744         adapter->napi = kcalloc(adapter->req_rx_queues,
 745                                 sizeof(struct napi_struct), GFP_KERNEL);
 746         if (!adapter->napi)
 747                 return -ENOMEM;
 748
 749         for (i = 0; i < adapter->req_rx_queues; i++) {
 750                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 751                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 752                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 753         }
 754
 755         adapter->num_active_rx_napi = adapter->req_rx_queues;
 756         return 0;
 757 }
 758
 759 static void release_napi(struct ibmvnic_adapter *adapter)
 760 {
 761         int i;
 762
 763         if (!adapter->napi)
 764                 return;
 765
 766         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 767                 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
 768                 netif_napi_del(&adapter->napi[i]);
 769         }
 770
 771         kfree(adapter->napi);
 772         adapter->napi = NULL;
 773         adapter->num_active_rx_napi = 0;
 774         adapter->napi_enabled = false;
 775 }
 776
 777 static int ibmvnic_login(struct net_device *netdev)
 778 {
 779         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 780         unsigned long timeout = msecs_to_jiffies(30000);
 781         int retry_count = 0;
 782         bool retry;
 783         int rc;
 784
 785         do {
 786                 retry = false;
 787                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 788                         netdev_warn(netdev, "Login attempts exceeded\n");
 789                         return -1;
 790                 }
 791
 792                 adapter->init_done_rc = 0;
 793                 reinit_completion(&adapter->init_done);
 794                 rc = send_login(adapter);
 795                 if (rc) {
 796                         netdev_warn(netdev, "Unable to login\n");
 797                         return rc;
 798                 }
 799
 800                 if (!wait_for_completion_timeout(&adapter->init_done,
 801                                                  timeout)) {
 802                         netdev_warn(netdev, "Login timed out\n");
 803                         return -1;
 804                 }
 805
 806                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 807                         retry_count++;
 808                         release_sub_crqs(adapter, 1);
 809
 810                         retry = true;
 811                         netdev_dbg(netdev,
 812                                    "Received partial success, retrying...\n");
 813                         adapter->init_done_rc = 0;
 814                         reinit_completion(&adapter->init_done);
 815                         send_cap_queries(adapter);
 816                         if (!wait_for_completion_timeout(&adapter->init_done,
 817                                                          timeout)) {
 818                                 netdev_warn(netdev,
 819                                             "Capabilities query timed out\n");
 820                                 return -1;
 821                         }
 822
 823                         rc = init_sub_crqs(adapter);
 824                         if (rc) {
 825                                 netdev_warn(netdev,
 826                                             "SCRQ initialization failed\n");
 827                                 return -1;
 828                         }
 829
 830                         rc = init_sub_crq_irqs(adapter);
 831                         if (rc) {
 832                                 netdev_warn(netdev,
 833                                             "SCRQ irq initialization failed\n");
 834                                 return -1;
 835                         }
 836                 } else if (adapter->init_done_rc) {
 837                         netdev_warn(netdev, "Adapter login failed\n");
 838                         return -1;
 839                 }
 840         } while (retry);
 841
 842         __ibmvnic_set_mac(netdev, adapter->mac_addr);
 843
 844         return 0;
 845 }
 846
 847 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 848 {
 849         kfree(adapter->login_buf);
 850         adapter->login_buf = NULL;
 851 }
 852
 853 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 854 {
 855         kfree(adapter->login_rsp_buf);
 856         adapter->login_rsp_buf = NULL;
 857 }
 858
 859 static void release_resources(struct ibmvnic_adapter *adapter)
 860 {
 861         release_vpd_data(adapter);
 862
 863         release_tx_pools(adapter);
 864         release_rx_pools(adapter);
 865
 866         release_napi(adapter);
 867         release_login_rsp_buffer(adapter);
 868 }
 869
 870 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 871 {
 872         struct net_device *netdev = adapter->netdev;
 873         unsigned long timeout = msecs_to_jiffies(30000);
 874         union ibmvnic_crq crq;
 875         bool resend;
 876         int rc;
 877
 878         netdev_dbg(netdev, "setting link state %d\n", link_state);
 879
 880         memset(&crq, 0, sizeof(crq));
 881         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 882         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 883         crq.logical_link_state.link_state = link_state;
 884
 885         do {
 886                 resend = false;
 887
 888                 reinit_completion(&adapter->init_done);
 889                 rc = ibmvnic_send_crq(adapter, &crq);
 890                 if (rc) {
 891                         netdev_err(netdev, "Failed to set link state\n");
 892                         return rc;
 893                 }
 894
 895                 if (!wait_for_completion_timeout(&adapter->init_done,
 896                                                  timeout)) {
 897                         netdev_err(netdev, "timeout setting link state\n");
 898                         return -1;
 899                 }
 900
 901                 if (adapter->init_done_rc == 1) {
 902                         /* Partuial success, delay and re-send */
 903                         mdelay(1000);
 904                         resend = true;
 905                 } else if (adapter->init_done_rc) {
 906                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 907                                     adapter->init_done_rc);
 908                         return adapter->init_done_rc;
 909                 }
 910         } while (resend);
 911
 912         return 0;
 913 }
 914
 915 static int set_real_num_queues(struct net_device *netdev)
 916 {
 917         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 918         int rc;
 919
 920         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 921                    adapter->req_tx_queues, adapter->req_rx_queues);
 922
 923         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 924         if (rc) {
 925                 netdev_err(netdev, "failed to set the number of tx queues\n");
 926                 return rc;
 927         }
 928
 929         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 930         if (rc)
 931                 netdev_err(netdev, "failed to set the number of rx queues\n");
 932
 933         return rc;
 934 }
 935
 936 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 937 {
 938         struct device *dev = &adapter->vdev->dev;
 939         union ibmvnic_crq crq;
 940         int len = 0;
 941         int rc;
 942
 943         if (adapter->vpd->buff)
 944                 len = adapter->vpd->len;
 945
 946         init_completion(&adapter->fw_done);
 947         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 948         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 949         rc = ibmvnic_send_crq(adapter, &crq);
 950         if (rc)
 951                 return rc;
 952         wait_for_completion(&adapter->fw_done);
 953
 954         if (!adapter->vpd->len)
 955                 return -ENODATA;
 956
 957         if (!adapter->vpd->buff)
 958                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 959         else if (adapter->vpd->len != len)
 960                 adapter->vpd->buff =
 961                         krealloc(adapter->vpd->buff,
 962                                  adapter->vpd->len, GFP_KERNEL);
 963
 964         if (!adapter->vpd->buff) {
 965                 dev_err(dev, "Could allocate VPD buffer\n");
 966                 return -ENOMEM;
 967         }
 968
 969         adapter->vpd->dma_addr =
 970                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 971                                DMA_FROM_DEVICE);
 972         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 973                 dev_err(dev, "Could not map VPD buffer\n");
 974                 kfree(adapter->vpd->buff);
 975                 adapter->vpd->buff = NULL;
 976                 return -ENOMEM;
 977         }
 978
 979         reinit_completion(&adapter->fw_done);
 980         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 981         crq.get_vpd.cmd = GET_VPD;
 982         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 983         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 984         rc = ibmvnic_send_crq(adapter, &crq);
 985         if (rc) {
 986                 kfree(adapter->vpd->buff);
 987                 adapter->vpd->buff = NULL;
 988                 return rc;
 989         }
 990         wait_for_completion(&adapter->fw_done);
 991
 992         return 0;
 993 }
 994
 995 static int init_resources(struct ibmvnic_adapter *adapter)
 996 {
 997         struct net_device *netdev = adapter->netdev;
 998         int rc;
 999
1000         rc = set_real_num_queues(netdev);
1001         if (rc)
1002                 return rc;
1003
1004         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1005         if (!adapter->vpd)
1006                 return -ENOMEM;
1007
1008         /* Vital Product Data (VPD) */
1009         rc = ibmvnic_get_vpd(adapter);
1010         if (rc) {
1011                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1012                 return rc;
1013         }
1014
1015         adapter->map_id = 1;
1016
1017         rc = init_napi(adapter);
1018         if (rc)
1019                 return rc;
1020
1021         send_map_query(adapter);
1022
1023         rc = init_rx_pools(netdev);
1024         if (rc)
1025                 return rc;
1026
1027         rc = init_tx_pools(netdev);
1028         return rc;
1029 }
1030
1031 static int __ibmvnic_open(struct net_device *netdev)
1032 {
1033         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1034         enum vnic_state prev_state = adapter->state;
1035         int i, rc;
1036
1037         adapter->state = VNIC_OPENING;
1038         replenish_pools(adapter);
1039         ibmvnic_napi_enable(adapter);
1040
1041         /* We're ready to receive frames, enable the sub-crq interrupts and
1042          * set the logical link state to up
1043          */
1044         for (i = 0; i < adapter->req_rx_queues; i++) {
1045                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1046                 if (prev_state == VNIC_CLOSED)
1047                         enable_irq(adapter->rx_scrq[i]->irq);
1048                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1049         }
1050
1051         for (i = 0; i < adapter->req_tx_queues; i++) {
1052                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1053                 if (prev_state == VNIC_CLOSED)
1054                         enable_irq(adapter->tx_scrq[i]->irq);
1055                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1056         }
1057
1058         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1059         if (rc) {
1060                 for (i = 0; i < adapter->req_rx_queues; i++)
1061                         napi_disable(&adapter->napi[i]);
1062                 release_resources(adapter);
1063                 return rc;
1064         }
1065
1066         netif_tx_start_all_queues(netdev);
1067
1068         if (prev_state == VNIC_CLOSED) {
1069                 for (i = 0; i < adapter->req_rx_queues; i++)
1070                         napi_schedule(&adapter->napi[i]);
1071         }
1072
1073         adapter->state = VNIC_OPEN;
1074         return rc;
1075 }
1076
1077 static int ibmvnic_open(struct net_device *netdev)
1078 {
1079         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1080         int rc;
1081
1082         /* If device failover is pending, just set device state and return.
1083          * Device operation will be handled by reset routine.
1084          */
1085         if (adapter->failover_pending) {
1086                 adapter->state = VNIC_OPEN;
1087                 return 0;
1088         }
1089
1090         if (adapter->state != VNIC_CLOSED) {
1091                 rc = ibmvnic_login(netdev);
1092                 if (rc)
1093                         return rc;
1094
1095                 rc = init_resources(adapter);
1096                 if (rc) {
1097                         netdev_err(netdev, "failed to initialize resources\n");
1098                         release_resources(adapter);
1099                         return rc;
1100                 }
1101         }
1102
1103         rc = __ibmvnic_open(netdev);
1104
1105         return rc;
1106 }
1107
1108 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1109 {
1110         struct ibmvnic_rx_pool *rx_pool;
1111         struct ibmvnic_rx_buff *rx_buff;
1112         u64 rx_entries;
1113         int rx_scrqs;
1114         int i, j;
1115
1116         if (!adapter->rx_pool)
1117                 return;
1118
1119         rx_scrqs = adapter->num_active_rx_pools;
1120         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1121
1122         /* Free any remaining skbs in the rx buffer pools */
1123         for (i = 0; i < rx_scrqs; i++) {
1124                 rx_pool = &adapter->rx_pool[i];
1125                 if (!rx_pool || !rx_pool->rx_buff)
1126                         continue;
1127
1128                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1129                 for (j = 0; j < rx_entries; j++) {
1130                         rx_buff = &rx_pool->rx_buff[j];
1131                         if (rx_buff && rx_buff->skb) {
1132                                 dev_kfree_skb_any(rx_buff->skb);
1133                                 rx_buff->skb = NULL;
1134                         }
1135                 }
1136         }
1137 }
1138
1139 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1140                               struct ibmvnic_tx_pool *tx_pool)
1141 {
1142         struct ibmvnic_tx_buff *tx_buff;
1143         u64 tx_entries;
1144         int i;
1145
1146         if (!tx_pool || !tx_pool->tx_buff)
1147                 return;
1148
1149         tx_entries = tx_pool->num_buffers;
1150
1151         for (i = 0; i < tx_entries; i++) {
1152                 tx_buff = &tx_pool->tx_buff[i];
1153                 if (tx_buff && tx_buff->skb) {
1154                         dev_kfree_skb_any(tx_buff->skb);
1155                         tx_buff->skb = NULL;
1156                 }
1157         }
1158 }
1159
1160 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1161 {
1162         int tx_scrqs;
1163         int i;
1164
1165         if (!adapter->tx_pool || !adapter->tso_pool)
1166                 return;
1167
1168         tx_scrqs = adapter->num_active_tx_pools;
1169
1170         /* Free any remaining skbs in the tx buffer pools */
1171         for (i = 0; i < tx_scrqs; i++) {
1172                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1173                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1174                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1175         }
1176 }
1177
1178 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1179 {
1180         struct net_device *netdev = adapter->netdev;
1181         int i;
1182
1183         if (adapter->tx_scrq) {
1184                 for (i = 0; i < adapter->req_tx_queues; i++)
1185                         if (adapter->tx_scrq[i]->irq) {
1186                                 netdev_dbg(netdev,
1187                                            "Disabling tx_scrq[%d] irq\n", i);
1188                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1189                                 disable_irq(adapter->tx_scrq[i]->irq);
1190                         }
1191         }
1192
1193         if (adapter->rx_scrq) {
1194                 for (i = 0; i < adapter->req_rx_queues; i++) {
1195                         if (adapter->rx_scrq[i]->irq) {
1196                                 netdev_dbg(netdev,
1197                                            "Disabling rx_scrq[%d] irq\n", i);
1198                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1199                                 disable_irq(adapter->rx_scrq[i]->irq);
1200                         }
1201                 }
1202         }
1203 }
1204
1205 static void ibmvnic_cleanup(struct net_device *netdev)
1206 {
1207         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1208
1209         /* ensure that transmissions are stopped if called by do_reset */
1210         if (adapter->resetting)
1211                 netif_tx_disable(netdev);
1212         else
1213                 netif_tx_stop_all_queues(netdev);
1214
1215         ibmvnic_napi_disable(adapter);
1216         ibmvnic_disable_irqs(adapter);
1217
1218         clean_rx_pools(adapter);
1219         clean_tx_pools(adapter);
1220 }
1221
1222 static int __ibmvnic_close(struct net_device *netdev)
1223 {
1224         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1225         int rc = 0;
1226
1227         adapter->state = VNIC_CLOSING;
1228         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1229         if (rc)
1230                 return rc;
1231         adapter->state = VNIC_CLOSED;
1232         return 0;
1233 }
1234
1235 static int ibmvnic_close(struct net_device *netdev)
1236 {
1237         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1238         int rc;
1239
1240         /* If device failover is pending, just set device state and return.
1241          * Device operation will be handled by reset routine.
1242          */
1243         if (adapter->failover_pending) {
1244                 adapter->state = VNIC_CLOSED;
1245                 return 0;
1246         }
1247
1248         rc = __ibmvnic_close(netdev);
1249         ibmvnic_cleanup(netdev);
1250
1251         return rc;
1252 }
1253
1254 /**
1255  * build_hdr_data - creates L2/L3/L4 header data buffer
1256  * @hdr_field - bitfield determining needed headers
1257  * @skb - socket buffer
1258  * @hdr_len - array of header lengths
1259  * @tot_len - total length of data
1260  *
1261  * Reads hdr_field to determine which headers are needed by firmware.
1262  * Builds a buffer containing these headers.  Saves individual header
1263  * lengths and total buffer length to be used to build descriptors.
1264  */
1265 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1266                           int *hdr_len, u8 *hdr_data)
1267 {
1268         int len = 0;
1269         u8 *hdr;
1270
1271         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1272                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1273         else
1274                 hdr_len[0] = sizeof(struct ethhdr);
1275
1276         if (skb->protocol == htons(ETH_P_IP)) {
1277                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1278                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1279                         hdr_len[2] = tcp_hdrlen(skb);
1280                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1281                         hdr_len[2] = sizeof(struct udphdr);
1282         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1283                 hdr_len[1] = sizeof(struct ipv6hdr);
1284                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1285                         hdr_len[2] = tcp_hdrlen(skb);
1286                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1287                         hdr_len[2] = sizeof(struct udphdr);
1288         } else if (skb->protocol == htons(ETH_P_ARP)) {
1289                 hdr_len[1] = arp_hdr_len(skb->dev);
1290                 hdr_len[2] = 0;
1291         }
1292
1293         memset(hdr_data, 0, 120);
1294         if ((hdr_field >> 6) & 1) {
1295                 hdr = skb_mac_header(skb);
1296                 memcpy(hdr_data, hdr, hdr_len[0]);
1297                 len += hdr_len[0];
1298         }
1299
1300         if ((hdr_field >> 5) & 1) {
1301                 hdr = skb_network_header(skb);
1302                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1303                 len += hdr_len[1];
1304         }
1305
1306         if ((hdr_field >> 4) & 1) {
1307                 hdr = skb_transport_header(skb);
1308                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1309                 len += hdr_len[2];
1310         }
1311         return len;
1312 }
1313
1314 /**
1315  * create_hdr_descs - create header and header extension descriptors
1316  * @hdr_field - bitfield determining needed headers
1317  * @data - buffer containing header data
1318  * @len - length of data buffer
1319  * @hdr_len - array of individual header lengths
1320  * @scrq_arr - descriptor array
1321  *
1322  * Creates header and, if needed, header extension descriptors and
1323  * places them in a descriptor array, scrq_arr
1324  */
1325
1326 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1327                             union sub_crq *scrq_arr)
1328 {
1329         union sub_crq hdr_desc;
1330         int tmp_len = len;
1331         int num_descs = 0;
1332         u8 *data, *cur;
1333         int tmp;
1334
1335         while (tmp_len > 0) {
1336                 cur = hdr_data + len - tmp_len;
1337
1338                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1339                 if (cur != hdr_data) {
1340                         data = hdr_desc.hdr_ext.data;
1341                         tmp = tmp_len > 29 ? 29 : tmp_len;
1342                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1343                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1344                         hdr_desc.hdr_ext.len = tmp;
1345                 } else {
1346                         data = hdr_desc.hdr.data;
1347                         tmp = tmp_len > 24 ? 24 : tmp_len;
1348                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1349                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1350                         hdr_desc.hdr.len = tmp;
1351                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1352                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1353                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1354                         hdr_desc.hdr.flag = hdr_field << 1;
1355                 }
1356                 memcpy(data, cur, tmp);
1357                 tmp_len -= tmp;
1358                 *scrq_arr = hdr_desc;
1359                 scrq_arr++;
1360                 num_descs++;
1361         }
1362
1363         return num_descs;
1364 }
1365
1366 /**
1367  * build_hdr_descs_arr - build a header descriptor array
1368  * @skb - socket buffer
1369  * @num_entries - number of descriptors to be sent
1370  * @subcrq - first TX descriptor
1371  * @hdr_field - bit field determining which headers will be sent
1372  *
1373  * This function will build a TX descriptor array with applicable
1374  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1375  */
1376
1377 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1378                                 int *num_entries, u8 hdr_field)
1379 {
1380         int hdr_len[3] = {0, 0, 0};
1381         int tot_len;
1382         u8 *hdr_data = txbuff->hdr_data;
1383
1384         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1385                                  txbuff->hdr_data);
1386         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1387                          txbuff->indir_arr + 1);
1388 }
1389
1390 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1391                                     struct net_device *netdev)
1392 {
1393         /* For some backing devices, mishandling of small packets
1394          * can result in a loss of connection or TX stall. Device
1395          * architects recommend that no packet should be smaller
1396          * than the minimum MTU value provided to the driver, so
1397          * pad any packets to that length
1398          */
1399         if (skb->len < netdev->min_mtu)
1400                 return skb_put_padto(skb, netdev->min_mtu);
1401
1402         return 0;
1403 }
1404
1405 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1406 {
1407         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1408         int queue_num = skb_get_queue_mapping(skb);
1409         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1410         struct device *dev = &adapter->vdev->dev;
1411         struct ibmvnic_tx_buff *tx_buff = NULL;
1412         struct ibmvnic_sub_crq_queue *tx_scrq;
1413         struct ibmvnic_tx_pool *tx_pool;
1414         unsigned int tx_send_failed = 0;
1415         unsigned int tx_map_failed = 0;
1416         unsigned int tx_dropped = 0;
1417         unsigned int tx_packets = 0;
1418         unsigned int tx_bytes = 0;
1419         dma_addr_t data_dma_addr;
1420         struct netdev_queue *txq;
1421         unsigned long lpar_rc;
1422         union sub_crq tx_crq;
1423         unsigned int offset;
1424         int num_entries = 1;
1425         unsigned char *dst;
1426         u64 *handle_array;
1427         int index = 0;
1428         u8 proto = 0;
1429         netdev_tx_t ret = NETDEV_TX_OK;
1430
1431         if (adapter->resetting) {
1432                 if (!netif_subqueue_stopped(netdev, skb))
1433                         netif_stop_subqueue(netdev, queue_num);
1434                 dev_kfree_skb_any(skb);
1435
1436                 tx_send_failed++;
1437                 tx_dropped++;
1438                 ret = NETDEV_TX_OK;
1439                 goto out;
1440         }
1441
1442         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1443                 tx_dropped++;
1444                 tx_send_failed++;
1445                 ret = NETDEV_TX_OK;
1446                 goto out;
1447         }
1448         if (skb_is_gso(skb))
1449                 tx_pool = &adapter->tso_pool[queue_num];
1450         else
1451                 tx_pool = &adapter->tx_pool[queue_num];
1452
1453         tx_scrq = adapter->tx_scrq[queue_num];
1454         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1455         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1456                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1457
1458         index = tx_pool->free_map[tx_pool->consumer_index];
1459
1460         if (index == IBMVNIC_INVALID_MAP) {
1461                 dev_kfree_skb_any(skb);
1462                 tx_send_failed++;
1463                 tx_dropped++;
1464                 ret = NETDEV_TX_OK;
1465                 goto out;
1466         }
1467
1468         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1469
1470         offset = index * tx_pool->buf_size;
1471         dst = tx_pool->long_term_buff.buff + offset;
1472         memset(dst, 0, tx_pool->buf_size);
1473         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1474
1475         if (skb_shinfo(skb)->nr_frags) {
1476                 int cur, i;
1477
1478                 /* Copy the head */
1479                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1480                 cur = skb_headlen(skb);
1481
1482                 /* Copy the frags */
1483                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1484                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1485
1486                         memcpy(dst + cur,
1487                                page_address(skb_frag_page(frag)) +
1488                                frag->page_offset, skb_frag_size(frag));
1489                         cur += skb_frag_size(frag);
1490                 }
1491         } else {
1492                 skb_copy_from_linear_data(skb, dst, skb->len);
1493         }
1494
1495         tx_pool->consumer_index =
1496             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1497
1498         tx_buff = &tx_pool->tx_buff[index];
1499         tx_buff->skb = skb;
1500         tx_buff->data_dma[0] = data_dma_addr;
1501         tx_buff->data_len[0] = skb->len;
1502         tx_buff->index = index;
1503         tx_buff->pool_index = queue_num;
1504         tx_buff->last_frag = true;
1505
1506         memset(&tx_crq, 0, sizeof(tx_crq));
1507         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1508         tx_crq.v1.type = IBMVNIC_TX_DESC;
1509         tx_crq.v1.n_crq_elem = 1;
1510         tx_crq.v1.n_sge = 1;
1511         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1512
1513         if (skb_is_gso(skb))
1514                 tx_crq.v1.correlator =
1515                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1516         else
1517                 tx_crq.v1.correlator = cpu_to_be32(index);
1518         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1519         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1520         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1521
1522         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1523                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1524                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1525         }
1526
1527         if (skb->protocol == htons(ETH_P_IP)) {
1528                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1529                 proto = ip_hdr(skb)->protocol;
1530         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1531                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1532                 proto = ipv6_hdr(skb)->nexthdr;
1533         }
1534
1535         if (proto == IPPROTO_TCP)
1536                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1537         else if (proto == IPPROTO_UDP)
1538                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1539
1540         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1541                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1542                 hdrs += 2;
1543         }
1544         if (skb_is_gso(skb)) {
1545                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1546                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1547                 hdrs += 2;
1548         }
1549         /* determine if l2/3/4 headers are sent to firmware */
1550         if ((*hdrs >> 7) & 1) {
1551                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1552                 tx_crq.v1.n_crq_elem = num_entries;
1553                 tx_buff->num_entries = num_entries;
1554                 tx_buff->indir_arr[0] = tx_crq;
1555                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1556                                                     sizeof(tx_buff->indir_arr),
1557                                                     DMA_TO_DEVICE);
1558                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1559                         dev_kfree_skb_any(skb);
1560                         tx_buff->skb = NULL;
1561                         if (!firmware_has_feature(FW_FEATURE_CMO))
1562                                 dev_err(dev, "tx: unable to map descriptor array\n");
1563                         tx_map_failed++;
1564                         tx_dropped++;
1565                         ret = NETDEV_TX_OK;
1566                         goto tx_err_out;
1567                 }
1568                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1569                                                (u64)tx_buff->indir_dma,
1570                                                (u64)num_entries);
1571                 dma_unmap_single(dev, tx_buff->indir_dma,
1572                                  sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1573         } else {
1574                 tx_buff->num_entries = num_entries;
1575                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1576                                       &tx_crq);
1577         }
1578         if (lpar_rc != H_SUCCESS) {
1579                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1580                         dev_err_ratelimited(dev, "tx: send failed\n");
1581                 dev_kfree_skb_any(skb);
1582                 tx_buff->skb = NULL;
1583
1584                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1585                         /* Disable TX and report carrier off if queue is closed
1586                          * or pending failover.
1587                          * Firmware guarantees that a signal will be sent to the
1588                          * driver, triggering a reset or some other action.
1589                          */
1590                         netif_tx_stop_all_queues(netdev);
1591                         netif_carrier_off(netdev);
1592                 }
1593
1594                 tx_send_failed++;
1595                 tx_dropped++;
1596                 ret = NETDEV_TX_OK;
1597                 goto tx_err_out;
1598         }
1599
1600         if (atomic_add_return(num_entries, &tx_scrq->used)
1601                                         >= adapter->req_tx_entries_per_subcrq) {
1602                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1603                 netif_stop_subqueue(netdev, queue_num);
1604         }
1605
1606         tx_packets++;
1607         tx_bytes += skb->len;
1608         txq->trans_start = jiffies;
1609         ret = NETDEV_TX_OK;
1610         goto out;
1611
1612 tx_err_out:
1613         /* roll back consumer index and map array*/
1614         if (tx_pool->consumer_index == 0)
1615                 tx_pool->consumer_index =
1616                         tx_pool->num_buffers - 1;
1617         else
1618                 tx_pool->consumer_index--;
1619         tx_pool->free_map[tx_pool->consumer_index] = index;
1620 out:
1621         netdev->stats.tx_dropped += tx_dropped;
1622         netdev->stats.tx_bytes += tx_bytes;
1623         netdev->stats.tx_packets += tx_packets;
1624         adapter->tx_send_failed += tx_send_failed;
1625         adapter->tx_map_failed += tx_map_failed;
1626         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1627         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1628         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1629
1630         return ret;
1631 }
1632
1633 static void ibmvnic_set_multi(struct net_device *netdev)
1634 {
1635         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1636         struct netdev_hw_addr *ha;
1637         union ibmvnic_crq crq;
1638
1639         memset(&crq, 0, sizeof(crq));
1640         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1641         crq.request_capability.cmd = REQUEST_CAPABILITY;
1642
1643         if (netdev->flags & IFF_PROMISC) {
1644                 if (!adapter->promisc_supported)
1645                         return;
1646         } else {
1647                 if (netdev->flags & IFF_ALLMULTI) {
1648                         /* Accept all multicast */
1649                         memset(&crq, 0, sizeof(crq));
1650                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1651                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1652                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1653                         ibmvnic_send_crq(adapter, &crq);
1654                 } else if (netdev_mc_empty(netdev)) {
1655                         /* Reject all multicast */
1656                         memset(&crq, 0, sizeof(crq));
1657                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1658                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1659                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1660                         ibmvnic_send_crq(adapter, &crq);
1661                 } else {
1662                         /* Accept one or more multicast(s) */
1663                         netdev_for_each_mc_addr(ha, netdev) {
1664                                 memset(&crq, 0, sizeof(crq));
1665                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1666                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1667                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1668                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1669                                                 ha->addr);
1670                                 ibmvnic_send_crq(adapter, &crq);
1671                         }
1672                 }
1673         }
1674 }
1675
1676 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1677 {
1678         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1679         union ibmvnic_crq crq;
1680         int rc;
1681
1682         if (!is_valid_ether_addr(dev_addr)) {
1683                 rc = -EADDRNOTAVAIL;
1684                 goto err;
1685         }
1686
1687         memset(&crq, 0, sizeof(crq));
1688         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1689         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1690         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1691
1692         init_completion(&adapter->fw_done);
1693         rc = ibmvnic_send_crq(adapter, &crq);
1694         if (rc) {
1695                 rc = -EIO;
1696                 goto err;
1697         }
1698
1699         wait_for_completion(&adapter->fw_done);
1700         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1701         if (adapter->fw_done_rc) {
1702                 rc = -EIO;
1703                 goto err;
1704         }
1705
1706         return 0;
1707 err:
1708         ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1709         return rc;
1710 }
1711
1712 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1713 {
1714         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1715         struct sockaddr *addr = p;
1716         int rc;
1717
1718         rc = 0;
1719         ether_addr_copy(adapter->mac_addr, addr->sa_data);
1720         if (adapter->state != VNIC_PROBED)
1721                 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1722
1723         return rc;
1724 }
1725
1726 /**
1727  * do_reset returns zero if we are able to keep processing reset events, or
1728  * non-zero if we hit a fatal error and must halt.
1729  */
1730 static int do_reset(struct ibmvnic_adapter *adapter,
1731                     struct ibmvnic_rwi *rwi, u32 reset_state)
1732 {
1733         u64 old_num_rx_queues, old_num_tx_queues;
1734         u64 old_num_rx_slots, old_num_tx_slots;
1735         struct net_device *netdev = adapter->netdev;
1736         int i, rc;
1737
1738         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1739                    rwi->reset_reason);
1740
1741         netif_carrier_off(netdev);
1742         adapter->reset_reason = rwi->reset_reason;
1743
1744         old_num_rx_queues = adapter->req_rx_queues;
1745         old_num_tx_queues = adapter->req_tx_queues;
1746         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1747         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1748
1749         ibmvnic_cleanup(netdev);
1750
1751         if (reset_state == VNIC_OPEN &&
1752             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1753             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1754                 rc = __ibmvnic_close(netdev);
1755                 if (rc)
1756                         return rc;
1757         }
1758
1759         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1760             adapter->wait_for_reset) {
1761                 release_resources(adapter);
1762                 release_sub_crqs(adapter, 1);
1763                 release_crq_queue(adapter);
1764         }
1765
1766         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1767                 /* remove the closed state so when we call open it appears
1768                  * we are coming from the probed state.
1769                  */
1770                 adapter->state = VNIC_PROBED;
1771
1772                 if (adapter->wait_for_reset) {
1773                         rc = init_crq_queue(adapter);
1774                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1775                         rc = ibmvnic_reenable_crq_queue(adapter);
1776                         release_sub_crqs(adapter, 1);
1777                 } else {
1778                         rc = ibmvnic_reset_crq(adapter);
1779                         if (!rc)
1780                                 rc = vio_enable_interrupts(adapter->vdev);
1781                 }
1782
1783                 if (rc) {
1784                         netdev_err(adapter->netdev,
1785                                    "Couldn't initialize crq. rc=%d\n", rc);
1786                         return rc;
1787                 }
1788
1789                 rc = ibmvnic_reset_init(adapter);
1790                 if (rc)
1791                         return IBMVNIC_INIT_FAILED;
1792
1793                 /* If the adapter was in PROBE state prior to the reset,
1794                  * exit here.
1795                  */
1796                 if (reset_state == VNIC_PROBED)
1797                         return 0;
1798
1799                 rc = ibmvnic_login(netdev);
1800                 if (rc) {
1801                         adapter->state = reset_state;
1802                         return rc;
1803                 }
1804
1805                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1806                     adapter->wait_for_reset) {
1807                         rc = init_resources(adapter);
1808                         if (rc)
1809                                 return rc;
1810                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1811                            adapter->req_tx_queues != old_num_tx_queues ||
1812                            adapter->req_rx_add_entries_per_subcrq !=
1813                                                         old_num_rx_slots ||
1814                            adapter->req_tx_entries_per_subcrq !=
1815                                                         old_num_tx_slots) {
1816                         release_rx_pools(adapter);
1817                         release_tx_pools(adapter);
1818                         release_napi(adapter);
1819                         release_vpd_data(adapter);
1820
1821                         rc = init_resources(adapter);
1822                         if (rc)
1823                                 return rc;
1824
1825                 } else {
1826                         rc = reset_tx_pools(adapter);
1827                         if (rc)
1828                                 return rc;
1829
1830                         rc = reset_rx_pools(adapter);
1831                         if (rc)
1832                                 return rc;
1833                 }
1834                 ibmvnic_disable_irqs(adapter);
1835         }
1836         adapter->state = VNIC_CLOSED;
1837
1838         if (reset_state == VNIC_CLOSED)
1839                 return 0;
1840
1841         rc = __ibmvnic_open(netdev);
1842         if (rc) {
1843                 if (list_empty(&adapter->rwi_list))
1844                         adapter->state = VNIC_CLOSED;
1845                 else
1846                         adapter->state = reset_state;
1847
1848                 return 0;
1849         }
1850
1851         /* refresh device's multicast list */
1852         ibmvnic_set_multi(netdev);
1853
1854         /* kick napi */
1855         for (i = 0; i < adapter->req_rx_queues; i++)
1856                 napi_schedule(&adapter->napi[i]);
1857
1858         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1859             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1860                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1861
1862         return 0;
1863 }
1864
1865 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1866                          struct ibmvnic_rwi *rwi, u32 reset_state)
1867 {
1868         struct net_device *netdev = adapter->netdev;
1869         int rc;
1870
1871         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1872                    rwi->reset_reason);
1873
1874         netif_carrier_off(netdev);
1875         adapter->reset_reason = rwi->reset_reason;
1876
1877         ibmvnic_cleanup(netdev);
1878         release_resources(adapter);
1879         release_sub_crqs(adapter, 0);
1880         release_crq_queue(adapter);
1881
1882         /* remove the closed state so when we call open it appears
1883          * we are coming from the probed state.
1884          */
1885         adapter->state = VNIC_PROBED;
1886
1887         reinit_completion(&adapter->init_done);
1888         rc = init_crq_queue(adapter);
1889         if (rc) {
1890                 netdev_err(adapter->netdev,
1891                            "Couldn't initialize crq. rc=%d\n", rc);
1892                 return rc;
1893         }
1894
1895         rc = ibmvnic_init(adapter);
1896         if (rc)
1897                 return rc;
1898
1899         /* If the adapter was in PROBE state prior to the reset,
1900          * exit here.
1901          */
1902         if (reset_state == VNIC_PROBED)
1903                 return 0;
1904
1905         rc = ibmvnic_login(netdev);
1906         if (rc) {
1907                 adapter->state = VNIC_PROBED;
1908                 return 0;
1909         }
1910
1911         rc = init_resources(adapter);
1912         if (rc)
1913                 return rc;
1914
1915         ibmvnic_disable_irqs(adapter);
1916         adapter->state = VNIC_CLOSED;
1917
1918         if (reset_state == VNIC_CLOSED)
1919                 return 0;
1920
1921         rc = __ibmvnic_open(netdev);
1922         if (rc) {
1923                 if (list_empty(&adapter->rwi_list))
1924                         adapter->state = VNIC_CLOSED;
1925                 else
1926                         adapter->state = reset_state;
1927
1928                 return 0;
1929         }
1930
1931         return 0;
1932 }
1933
1934 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1935 {
1936         struct ibmvnic_rwi *rwi;
1937         unsigned long flags;
1938
1939         spin_lock_irqsave(&adapter->rwi_lock, flags);
1940
1941         if (!list_empty(&adapter->rwi_list)) {
1942                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1943                                        list);
1944                 list_del(&rwi->list);
1945         } else {
1946                 rwi = NULL;
1947         }
1948
1949         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1950         return rwi;
1951 }
1952
1953 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1954 {
1955         struct ibmvnic_rwi *rwi;
1956
1957         rwi = get_next_rwi(adapter);
1958         while (rwi) {
1959                 kfree(rwi);
1960                 rwi = get_next_rwi(adapter);
1961         }
1962 }
1963
1964 static void __ibmvnic_reset(struct work_struct *work)
1965 {
1966         struct ibmvnic_rwi *rwi;
1967         struct ibmvnic_adapter *adapter;
1968         bool we_lock_rtnl = false;
1969         u32 reset_state;
1970         int rc = 0;
1971
1972         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1973
1974         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1975          * unless wait_for_reset is set, in which case the rtnl lock
1976          * has already been taken before initializing the reset
1977          */
1978         if (!adapter->wait_for_reset) {
1979                 rtnl_lock();
1980                 we_lock_rtnl = true;
1981         }
1982         reset_state = adapter->state;
1983
1984         rwi = get_next_rwi(adapter);
1985         while (rwi) {
1986                 if (adapter->state == VNIC_REMOVING ||
1987                     adapter->state == VNIC_REMOVED)
1988                         goto out;
1989
1990                 if (adapter->force_reset_recovery) {
1991                         adapter->force_reset_recovery = false;
1992                         rc = do_hard_reset(adapter, rwi, reset_state);
1993                 } else {
1994                         rc = do_reset(adapter, rwi, reset_state);
1995                 }
1996                 kfree(rwi);
1997                 if (rc && rc != IBMVNIC_INIT_FAILED &&
1998                     !adapter->force_reset_recovery)
1999                         break;
2000
2001                 rwi = get_next_rwi(adapter);
2002         }
2003
2004         if (adapter->wait_for_reset) {
2005                 adapter->wait_for_reset = false;
2006                 adapter->reset_done_rc = rc;
2007                 complete(&adapter->reset_done);
2008         }
2009
2010         if (rc) {
2011                 netdev_dbg(adapter->netdev, "Reset failed\n");
2012                 free_all_rwi(adapter);
2013         }
2014 out:
2015         adapter->resetting = false;
2016         if (we_lock_rtnl)
2017                 rtnl_unlock();
2018 }
2019
2020 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2021                          enum ibmvnic_reset_reason reason)
2022 {
2023         struct list_head *entry, *tmp_entry;
2024         struct ibmvnic_rwi *rwi, *tmp;
2025         struct net_device *netdev = adapter->netdev;
2026         unsigned long flags;
2027         int ret;
2028
2029         if (adapter->state == VNIC_REMOVING ||
2030             adapter->state == VNIC_REMOVED ||
2031             adapter->failover_pending) {
2032                 ret = EBUSY;
2033                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2034                 goto err;
2035         }
2036
2037         if (adapter->state == VNIC_PROBING) {
2038                 netdev_warn(netdev, "Adapter reset during probe\n");
2039                 ret = adapter->init_done_rc = EAGAIN;
2040                 goto err;
2041         }
2042
2043         spin_lock_irqsave(&adapter->rwi_lock, flags);
2044
2045         list_for_each(entry, &adapter->rwi_list) {
2046                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2047                 if (tmp->reset_reason == reason) {
2048                         netdev_dbg(netdev, "Skipping matching reset\n");
2049                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2050                         ret = EBUSY;
2051                         goto err;
2052                 }
2053         }
2054
2055         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2056         if (!rwi) {
2057                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2058                 ibmvnic_close(netdev);
2059                 ret = ENOMEM;
2060                 goto err;
2061         }
2062         /* if we just received a transport event,
2063          * flush reset queue and process this reset
2064          */
2065         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2066                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2067                         list_del(entry);
2068         }
2069         rwi->reset_reason = reason;
2070         list_add_tail(&rwi->list, &adapter->rwi_list);
2071         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2072         adapter->resetting = true;
2073         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2074         schedule_work(&adapter->ibmvnic_reset);
2075
2076         return 0;
2077 err:
2078         if (adapter->wait_for_reset)
2079                 adapter->wait_for_reset = false;
2080         return -ret;
2081 }
2082
2083 static void ibmvnic_tx_timeout(struct net_device *dev)
2084 {
2085         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2086
2087         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2088 }
2089
2090 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2091                                   struct ibmvnic_rx_buff *rx_buff)
2092 {
2093         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2094
2095         rx_buff->skb = NULL;
2096
2097         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2098         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2099
2100         atomic_dec(&pool->available);
2101 }
2102
2103 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2104 {
2105         struct net_device *netdev = napi->dev;
2106         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2107         int scrq_num = (int)(napi - adapter->napi);
2108         int frames_processed = 0;
2109
2110 restart_poll:
2111         while (frames_processed < budget) {
2112                 struct sk_buff *skb;
2113                 struct ibmvnic_rx_buff *rx_buff;
2114                 union sub_crq *next;
2115                 u32 length;
2116                 u16 offset;
2117                 u8 flags = 0;
2118
2119                 if (unlikely(adapter->resetting &&
2120                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2121                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2122                         napi_complete_done(napi, frames_processed);
2123                         return frames_processed;
2124                 }
2125
2126                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2127                         break;
2128                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2129                 rx_buff =
2130                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2131                                                           rx_comp.correlator);
2132                 /* do error checking */
2133                 if (next->rx_comp.rc) {
2134                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2135                                    be16_to_cpu(next->rx_comp.rc));
2136                         /* free the entry */
2137                         next->rx_comp.first = 0;
2138                         dev_kfree_skb_any(rx_buff->skb);
2139                         remove_buff_from_pool(adapter, rx_buff);
2140                         continue;
2141                 } else if (!rx_buff->skb) {
2142                         /* free the entry */
2143                         next->rx_comp.first = 0;
2144                         remove_buff_from_pool(adapter, rx_buff);
2145                         continue;
2146                 }
2147
2148                 length = be32_to_cpu(next->rx_comp.len);
2149                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2150                 flags = next->rx_comp.flags;
2151                 skb = rx_buff->skb;
2152                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2153                                         length);
2154
2155                 /* VLAN Header has been stripped by the system firmware and
2156                  * needs to be inserted by the driver
2157                  */
2158                 if (adapter->rx_vlan_header_insertion &&
2159                     (flags & IBMVNIC_VLAN_STRIPPED))
2160                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2161                                                ntohs(next->rx_comp.vlan_tci));
2162
2163                 /* free the entry */
2164                 next->rx_comp.first = 0;
2165                 remove_buff_from_pool(adapter, rx_buff);
2166
2167                 skb_put(skb, length);
2168                 skb->protocol = eth_type_trans(skb, netdev);
2169                 skb_record_rx_queue(skb, scrq_num);
2170
2171                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2172                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2173                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2174                 }
2175
2176                 length = skb->len;
2177                 napi_gro_receive(napi, skb); /* send it up */
2178                 netdev->stats.rx_packets++;
2179                 netdev->stats.rx_bytes += length;
2180                 adapter->rx_stats_buffers[scrq_num].packets++;
2181                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2182                 frames_processed++;
2183         }
2184
2185         if (adapter->state != VNIC_CLOSING)
2186                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2187
2188         if (frames_processed < budget) {
2189                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2190                 napi_complete_done(napi, frames_processed);
2191                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2192                     napi_reschedule(napi)) {
2193                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2194                         goto restart_poll;
2195                 }
2196         }
2197         return frames_processed;
2198 }
2199
2200 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2201 {
2202         int rc, ret;
2203
2204         adapter->fallback.mtu = adapter->req_mtu;
2205         adapter->fallback.rx_queues = adapter->req_rx_queues;
2206         adapter->fallback.tx_queues = adapter->req_tx_queues;
2207         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2208         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2209
2210         init_completion(&adapter->reset_done);
2211         adapter->wait_for_reset = true;
2212         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2213         if (rc)
2214                 return rc;
2215         wait_for_completion(&adapter->reset_done);
2216
2217         ret = 0;
2218         if (adapter->reset_done_rc) {
2219                 ret = -EIO;
2220                 adapter->desired.mtu = adapter->fallback.mtu;
2221                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2222                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2223                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2224                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2225
2226                 init_completion(&adapter->reset_done);
2227                 adapter->wait_for_reset = true;
2228                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2229                 if (rc)
2230                         return ret;
2231                 wait_for_completion(&adapter->reset_done);
2232         }
2233         adapter->wait_for_reset = false;
2234
2235         return ret;
2236 }
2237
2238 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2239 {
2240         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2241
2242         adapter->desired.mtu = new_mtu + ETH_HLEN;
2243
2244         return wait_for_reset(adapter);
2245 }
2246
2247 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2248                                                 struct net_device *dev,
2249                                                 netdev_features_t features)
2250 {
2251         /* Some backing hardware adapters can not
2252          * handle packets with a MSS less than 224
2253          * or with only one segment.
2254          */
2255         if (skb_is_gso(skb)) {
2256                 if (skb_shinfo(skb)->gso_size < 224 ||
2257                     skb_shinfo(skb)->gso_segs == 1)
2258                         features &= ~NETIF_F_GSO_MASK;
2259         }
2260
2261         return features;
2262 }
2263
2264 static const struct net_device_ops ibmvnic_netdev_ops = {
2265         .ndo_open               = ibmvnic_open,
2266         .ndo_stop               = ibmvnic_close,
2267         .ndo_start_xmit         = ibmvnic_xmit,
2268         .ndo_set_rx_mode        = ibmvnic_set_multi,
2269         .ndo_set_mac_address    = ibmvnic_set_mac,
2270         .ndo_validate_addr      = eth_validate_addr,
2271         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2272         .ndo_change_mtu         = ibmvnic_change_mtu,
2273         .ndo_features_check     = ibmvnic_features_check,
2274 };
2275
2276 /* ethtool functions */
2277
2278 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2279                                       struct ethtool_link_ksettings *cmd)
2280 {
2281         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2282         int rc;
2283
2284         rc = send_query_phys_parms(adapter);
2285         if (rc) {
2286                 adapter->speed = SPEED_UNKNOWN;
2287                 adapter->duplex = DUPLEX_UNKNOWN;
2288         }
2289         cmd->base.speed = adapter->speed;
2290         cmd->base.duplex = adapter->duplex;
2291         cmd->base.port = PORT_FIBRE;
2292         cmd->base.phy_address = 0;
2293         cmd->base.autoneg = AUTONEG_ENABLE;
2294
2295         return 0;
2296 }
2297
2298 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2299                                 struct ethtool_drvinfo *info)
2300 {
2301         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2302
2303         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2304         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2305         strlcpy(info->fw_version, adapter->fw_version,
2306                 sizeof(info->fw_version));
2307 }
2308
2309 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2310 {
2311         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2312
2313         return adapter->msg_enable;
2314 }
2315
2316 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2317 {
2318         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2319
2320         adapter->msg_enable = data;
2321 }
2322
2323 static u32 ibmvnic_get_link(struct net_device *netdev)
2324 {
2325         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2326
2327         /* Don't need to send a query because we request a logical link up at
2328          * init and then we wait for link state indications
2329          */
2330         return adapter->logical_link_state;
2331 }
2332
2333 static void ibmvnic_get_ringparam(struct net_device *netdev,
2334                                   struct ethtool_ringparam *ring)
2335 {
2336         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2337
2338         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2339                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2340                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2341         } else {
2342                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2343                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2344         }
2345         ring->rx_mini_max_pending = 0;
2346         ring->rx_jumbo_max_pending = 0;
2347         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2348         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2349         ring->rx_mini_pending = 0;
2350         ring->rx_jumbo_pending = 0;
2351 }
2352
2353 static int ibmvnic_set_ringparam(struct net_device *netdev,
2354                                  struct ethtool_ringparam *ring)
2355 {
2356         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2357         int ret;
2358
2359         ret = 0;
2360         adapter->desired.rx_entries = ring->rx_pending;
2361         adapter->desired.tx_entries = ring->tx_pending;
2362
2363         ret = wait_for_reset(adapter);
2364
2365         if (!ret &&
2366             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2367              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2368                 netdev_info(netdev,
2369                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2370                             ring->rx_pending, ring->tx_pending,
2371                             adapter->req_rx_add_entries_per_subcrq,
2372                             adapter->req_tx_entries_per_subcrq);
2373         return ret;
2374 }
2375
2376 static void ibmvnic_get_channels(struct net_device *netdev,
2377                                  struct ethtool_channels *channels)
2378 {
2379         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2380
2381         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2382                 channels->max_rx = adapter->max_rx_queues;
2383                 channels->max_tx = adapter->max_tx_queues;
2384         } else {
2385                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2386                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2387         }
2388
2389         channels->max_other = 0;
2390         channels->max_combined = 0;
2391         channels->rx_count = adapter->req_rx_queues;
2392         channels->tx_count = adapter->req_tx_queues;
2393         channels->other_count = 0;
2394         channels->combined_count = 0;
2395 }
2396
2397 static int ibmvnic_set_channels(struct net_device *netdev,
2398                                 struct ethtool_channels *channels)
2399 {
2400         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2401         int ret;
2402
2403         ret = 0;
2404         adapter->desired.rx_queues = channels->rx_count;
2405         adapter->desired.tx_queues = channels->tx_count;
2406
2407         ret = wait_for_reset(adapter);
2408
2409         if (!ret &&
2410             (adapter->req_rx_queues != channels->rx_count ||
2411              adapter->req_tx_queues != channels->tx_count))
2412                 netdev_info(netdev,
2413                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2414                             channels->rx_count, channels->tx_count,
2415                             adapter->req_rx_queues, adapter->req_tx_queues);
2416         return ret;
2417
2418 }
2419
2420 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2421 {
2422         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2423         int i;
2424
2425         switch (stringset) {
2426         case ETH_SS_STATS:
2427                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2428                                 i++, data += ETH_GSTRING_LEN)
2429                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2430
2431                 for (i = 0; i < adapter->req_tx_queues; i++) {
2432                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2433                         data += ETH_GSTRING_LEN;
2434
2435                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2436                         data += ETH_GSTRING_LEN;
2437
2438                         snprintf(data, ETH_GSTRING_LEN,
2439                                  "tx%d_dropped_packets", i);
2440                         data += ETH_GSTRING_LEN;
2441                 }
2442
2443                 for (i = 0; i < adapter->req_rx_queues; i++) {
2444                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2445                         data += ETH_GSTRING_LEN;
2446
2447                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2448                         data += ETH_GSTRING_LEN;
2449
2450                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2451                         data += ETH_GSTRING_LEN;
2452                 }
2453                 break;
2454
2455         case ETH_SS_PRIV_FLAGS:
2456                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2457                         strcpy(data + i * ETH_GSTRING_LEN,
2458                                ibmvnic_priv_flags[i]);
2459                 break;
2460         default:
2461                 return;
2462         }
2463 }
2464
2465 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2466 {
2467         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2468
2469         switch (sset) {
2470         case ETH_SS_STATS:
2471                 return ARRAY_SIZE(ibmvnic_stats) +
2472                        adapter->req_tx_queues * NUM_TX_STATS +
2473                        adapter->req_rx_queues * NUM_RX_STATS;
2474         case ETH_SS_PRIV_FLAGS:
2475                 return ARRAY_SIZE(ibmvnic_priv_flags);
2476         default:
2477                 return -EOPNOTSUPP;
2478         }
2479 }
2480
2481 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2482                                       struct ethtool_stats *stats, u64 *data)
2483 {
2484         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2485         union ibmvnic_crq crq;
2486         int i, j;
2487         int rc;
2488
2489         memset(&crq, 0, sizeof(crq));
2490         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2491         crq.request_statistics.cmd = REQUEST_STATISTICS;
2492         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2493         crq.request_statistics.len =
2494             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2495
2496         /* Wait for data to be written */
2497         init_completion(&adapter->stats_done);
2498         rc = ibmvnic_send_crq(adapter, &crq);
2499         if (rc)
2500                 return;
2501         wait_for_completion(&adapter->stats_done);
2502
2503         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2504                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2505                                                 ibmvnic_stats[i].offset));
2506
2507         for (j = 0; j < adapter->req_tx_queues; j++) {
2508                 data[i] = adapter->tx_stats_buffers[j].packets;
2509                 i++;
2510                 data[i] = adapter->tx_stats_buffers[j].bytes;
2511                 i++;
2512                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2513                 i++;
2514         }
2515
2516         for (j = 0; j < adapter->req_rx_queues; j++) {
2517                 data[i] = adapter->rx_stats_buffers[j].packets;
2518                 i++;
2519                 data[i] = adapter->rx_stats_buffers[j].bytes;
2520                 i++;
2521                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2522                 i++;
2523         }
2524 }
2525
2526 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2527 {
2528         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2529
2530         return adapter->priv_flags;
2531 }
2532
2533 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2534 {
2535         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2536         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2537
2538         if (which_maxes)
2539                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2540         else
2541                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2542
2543         return 0;
2544 }
2545 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2546         .get_drvinfo            = ibmvnic_get_drvinfo,
2547         .get_msglevel           = ibmvnic_get_msglevel,
2548         .set_msglevel           = ibmvnic_set_msglevel,
2549         .get_link               = ibmvnic_get_link,
2550         .get_ringparam          = ibmvnic_get_ringparam,
2551         .set_ringparam          = ibmvnic_set_ringparam,
2552         .get_channels           = ibmvnic_get_channels,
2553         .set_channels           = ibmvnic_set_channels,
2554         .get_strings            = ibmvnic_get_strings,
2555         .get_sset_count         = ibmvnic_get_sset_count,
2556         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2557         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2558         .get_priv_flags         = ibmvnic_get_priv_flags,
2559         .set_priv_flags         = ibmvnic_set_priv_flags,
2560 };
2561
2562 /* Routines for managing CRQs/sCRQs  */
2563
2564 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2565                                    struct ibmvnic_sub_crq_queue *scrq)
2566 {
2567         int rc;
2568
2569         if (scrq->irq) {
2570                 free_irq(scrq->irq, scrq);
2571                 irq_dispose_mapping(scrq->irq);
2572                 scrq->irq = 0;
2573         }
2574
2575         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2576         atomic_set(&scrq->used, 0);
2577         scrq->cur = 0;
2578
2579         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2580                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2581         return rc;
2582 }
2583
2584 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2585 {
2586         int i, rc;
2587
2588         for (i = 0; i < adapter->req_tx_queues; i++) {
2589                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2590                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2591                 if (rc)
2592                         return rc;
2593         }
2594
2595         for (i = 0; i < adapter->req_rx_queues; i++) {
2596                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2597                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2598                 if (rc)
2599                         return rc;
2600         }
2601
2602         return rc;
2603 }
2604
2605 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2606                                   struct ibmvnic_sub_crq_queue *scrq,
2607                                   bool do_h_free)
2608 {
2609         struct device *dev = &adapter->vdev->dev;
2610         long rc;
2611
2612         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2613
2614         if (do_h_free) {
2615                 /* Close the sub-crqs */
2616                 do {
2617                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2618                                                 adapter->vdev->unit_address,
2619                                                 scrq->crq_num);
2620                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2621
2622                 if (rc) {
2623                         netdev_err(adapter->netdev,
2624                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2625                                    scrq->crq_num, rc);
2626                 }
2627         }
2628
2629         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2630                          DMA_BIDIRECTIONAL);
2631         free_pages((unsigned long)scrq->msgs, 2);
2632         kfree(scrq);
2633 }
2634
2635 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2636                                                         *adapter)
2637 {
2638         struct device *dev = &adapter->vdev->dev;
2639         struct ibmvnic_sub_crq_queue *scrq;
2640         int rc;
2641
2642         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2643         if (!scrq)
2644                 return NULL;
2645
2646         scrq->msgs =
2647                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2648         if (!scrq->msgs) {
2649                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2650                 goto zero_page_failed;
2651         }
2652
2653         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2654                                          DMA_BIDIRECTIONAL);
2655         if (dma_mapping_error(dev, scrq->msg_token)) {
2656                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2657                 goto map_failed;
2658         }
2659
2660         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2661                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2662
2663         if (rc == H_RESOURCE)
2664                 rc = ibmvnic_reset_crq(adapter);
2665
2666         if (rc == H_CLOSED) {
2667                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2668         } else if (rc) {
2669                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2670                 goto reg_failed;
2671         }
2672
2673         scrq->adapter = adapter;
2674         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2675         spin_lock_init(&scrq->lock);
2676
2677         netdev_dbg(adapter->netdev,
2678                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2679                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2680
2681         return scrq;
2682
2683 reg_failed:
2684         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2685                          DMA_BIDIRECTIONAL);
2686 map_failed:
2687         free_pages((unsigned long)scrq->msgs, 2);
2688 zero_page_failed:
2689         kfree(scrq);
2690
2691         return NULL;
2692 }
2693
2694 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2695 {
2696         int i;
2697
2698         if (adapter->tx_scrq) {
2699                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2700                         if (!adapter->tx_scrq[i])
2701                                 continue;
2702
2703                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2704                                    i);
2705                         if (adapter->tx_scrq[i]->irq) {
2706                                 free_irq(adapter->tx_scrq[i]->irq,
2707                                          adapter->tx_scrq[i]);
2708                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2709                                 adapter->tx_scrq[i]->irq = 0;
2710                         }
2711
2712                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2713                                               do_h_free);
2714                 }
2715
2716                 kfree(adapter->tx_scrq);
2717                 adapter->tx_scrq = NULL;
2718                 adapter->num_active_tx_scrqs = 0;
2719         }
2720
2721         if (adapter->rx_scrq) {
2722                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2723                         if (!adapter->rx_scrq[i])
2724                                 continue;
2725
2726                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2727                                    i);
2728                         if (adapter->rx_scrq[i]->irq) {
2729                                 free_irq(adapter->rx_scrq[i]->irq,
2730                                          adapter->rx_scrq[i]);
2731                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2732                                 adapter->rx_scrq[i]->irq = 0;
2733                         }
2734
2735                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2736                                               do_h_free);
2737                 }
2738
2739                 kfree(adapter->rx_scrq);
2740                 adapter->rx_scrq = NULL;
2741                 adapter->num_active_rx_scrqs = 0;
2742         }
2743 }
2744
2745 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2746                             struct ibmvnic_sub_crq_queue *scrq)
2747 {
2748         struct device *dev = &adapter->vdev->dev;
2749         unsigned long rc;
2750
2751         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2752                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2753         if (rc)
2754                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2755                         scrq->hw_irq, rc);
2756         return rc;
2757 }
2758
2759 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2760                            struct ibmvnic_sub_crq_queue *scrq)
2761 {
2762         struct device *dev = &adapter->vdev->dev;
2763         unsigned long rc;
2764
2765         if (scrq->hw_irq > 0x100000000ULL) {
2766                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2767                 return 1;
2768         }
2769
2770         if (adapter->resetting &&
2771             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2772                 u64 val = (0xff000000) | scrq->hw_irq;
2773
2774                 rc = plpar_hcall_norets(H_EOI, val);
2775                 if (rc)
2776                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2777                                 val, rc);
2778         }
2779
2780         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2781                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2782         if (rc)
2783                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2784                         scrq->hw_irq, rc);
2785         return rc;
2786 }
2787
2788 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2789                                struct ibmvnic_sub_crq_queue *scrq)
2790 {
2791         struct device *dev = &adapter->vdev->dev;
2792         struct ibmvnic_tx_pool *tx_pool;
2793         struct ibmvnic_tx_buff *txbuff;
2794         union sub_crq *next;
2795         int index;
2796         int i, j;
2797
2798 restart_loop:
2799         while (pending_scrq(adapter, scrq)) {
2800                 unsigned int pool = scrq->pool_index;
2801                 int num_entries = 0;
2802
2803                 next = ibmvnic_next_scrq(adapter, scrq);
2804                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2805                         if (next->tx_comp.rcs[i]) {
2806                                 dev_err(dev, "tx error %x\n",
2807                                         next->tx_comp.rcs[i]);
2808                                 continue;
2809                         }
2810                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2811                         if (index & IBMVNIC_TSO_POOL_MASK) {
2812                                 tx_pool = &adapter->tso_pool[pool];
2813                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2814                         } else {
2815                                 tx_pool = &adapter->tx_pool[pool];
2816                         }
2817
2818                         txbuff = &tx_pool->tx_buff[index];
2819
2820                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2821                                 if (!txbuff->data_dma[j])
2822                                         continue;
2823
2824                                 txbuff->data_dma[j] = 0;
2825                         }
2826
2827                         if (txbuff->last_frag) {
2828                                 dev_kfree_skb_any(txbuff->skb);
2829                                 txbuff->skb = NULL;
2830                         }
2831
2832                         num_entries += txbuff->num_entries;
2833
2834                         tx_pool->free_map[tx_pool->producer_index] = index;
2835                         tx_pool->producer_index =
2836                                 (tx_pool->producer_index + 1) %
2837                                         tx_pool->num_buffers;
2838                 }
2839                 /* remove tx_comp scrq*/
2840                 next->tx_comp.first = 0;
2841
2842                 if (atomic_sub_return(num_entries, &scrq->used) <=
2843                     (adapter->req_tx_entries_per_subcrq / 2) &&
2844                     __netif_subqueue_stopped(adapter->netdev,
2845                                              scrq->pool_index)) {
2846                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2847                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2848                                    scrq->pool_index);
2849                 }
2850         }
2851
2852         enable_scrq_irq(adapter, scrq);
2853
2854         if (pending_scrq(adapter, scrq)) {
2855                 disable_scrq_irq(adapter, scrq);
2856                 goto restart_loop;
2857         }
2858
2859         return 0;
2860 }
2861
2862 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2863 {
2864         struct ibmvnic_sub_crq_queue *scrq = instance;
2865         struct ibmvnic_adapter *adapter = scrq->adapter;
2866
2867         disable_scrq_irq(adapter, scrq);
2868         ibmvnic_complete_tx(adapter, scrq);
2869
2870         return IRQ_HANDLED;
2871 }
2872
2873 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2874 {
2875         struct ibmvnic_sub_crq_queue *scrq = instance;
2876         struct ibmvnic_adapter *adapter = scrq->adapter;
2877
2878         /* When booting a kdump kernel we can hit pending interrupts
2879          * prior to completing driver initialization.
2880          */
2881         if (unlikely(adapter->state != VNIC_OPEN))
2882                 return IRQ_NONE;
2883
2884         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2885
2886         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2887                 disable_scrq_irq(adapter, scrq);
2888                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2889         }
2890
2891         return IRQ_HANDLED;
2892 }
2893
2894 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2895 {
2896         struct device *dev = &adapter->vdev->dev;
2897         struct ibmvnic_sub_crq_queue *scrq;
2898         int i = 0, j = 0;
2899         int rc = 0;
2900
2901         for (i = 0; i < adapter->req_tx_queues; i++) {
2902                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2903                            i);
2904                 scrq = adapter->tx_scrq[i];
2905                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2906
2907                 if (!scrq->irq) {
2908                         rc = -EINVAL;
2909                         dev_err(dev, "Error mapping irq\n");
2910                         goto req_tx_irq_failed;
2911                 }
2912
2913                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
2914                          adapter->vdev->unit_address, i);
2915                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2916                                  0, scrq->name, scrq);
2917
2918                 if (rc) {
2919                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2920                                 scrq->irq, rc);
2921                         irq_dispose_mapping(scrq->irq);
2922                         goto req_tx_irq_failed;
2923                 }
2924         }
2925
2926         for (i = 0; i < adapter->req_rx_queues; i++) {
2927                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2928                            i);
2929                 scrq = adapter->rx_scrq[i];
2930                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2931                 if (!scrq->irq) {
2932                         rc = -EINVAL;
2933                         dev_err(dev, "Error mapping irq\n");
2934                         goto req_rx_irq_failed;
2935                 }
2936                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
2937                          adapter->vdev->unit_address, i);
2938                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2939                                  0, scrq->name, scrq);
2940                 if (rc) {
2941                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2942                                 scrq->irq, rc);
2943                         irq_dispose_mapping(scrq->irq);
2944                         goto req_rx_irq_failed;
2945                 }
2946         }
2947         return rc;
2948
2949 req_rx_irq_failed:
2950         for (j = 0; j < i; j++) {
2951                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2952                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2953         }
2954         i = adapter->req_tx_queues;
2955 req_tx_irq_failed:
2956         for (j = 0; j < i; j++) {
2957                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2958                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2959         }
2960         release_sub_crqs(adapter, 1);
2961         return rc;
2962 }
2963
2964 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2965 {
2966         struct device *dev = &adapter->vdev->dev;
2967         struct ibmvnic_sub_crq_queue **allqueues;
2968         int registered_queues = 0;
2969         int total_queues;
2970         int more = 0;
2971         int i;
2972
2973         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2974
2975         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2976         if (!allqueues)
2977                 return -1;
2978
2979         for (i = 0; i < total_queues; i++) {
2980                 allqueues[i] = init_sub_crq_queue(adapter);
2981                 if (!allqueues[i]) {
2982                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2983                         break;
2984                 }
2985                 registered_queues++;
2986         }
2987
2988         /* Make sure we were able to register the minimum number of queues */
2989         if (registered_queues <
2990             adapter->min_tx_queues + adapter->min_rx_queues) {
2991                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2992                 goto tx_failed;
2993         }
2994
2995         /* Distribute the failed allocated queues*/
2996         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2997                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2998                 switch (i % 3) {
2999                 case 0:
3000                         if (adapter->req_rx_queues > adapter->min_rx_queues)
3001                                 adapter->req_rx_queues--;
3002                         else
3003                                 more++;
3004                         break;
3005                 case 1:
3006                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3007                                 adapter->req_tx_queues--;
3008                         else
3009                                 more++;
3010                         break;
3011                 }
3012         }
3013
3014         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3015                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3016         if (!adapter->tx_scrq)
3017                 goto tx_failed;
3018
3019         for (i = 0; i < adapter->req_tx_queues; i++) {
3020                 adapter->tx_scrq[i] = allqueues[i];
3021                 adapter->tx_scrq[i]->pool_index = i;
3022                 adapter->num_active_tx_scrqs++;
3023         }
3024
3025         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3026                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3027         if (!adapter->rx_scrq)
3028                 goto rx_failed;
3029
3030         for (i = 0; i < adapter->req_rx_queues; i++) {
3031                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3032                 adapter->rx_scrq[i]->scrq_num = i;
3033                 adapter->num_active_rx_scrqs++;
3034         }
3035
3036         kfree(allqueues);
3037         return 0;
3038
3039 rx_failed:
3040         kfree(adapter->tx_scrq);
3041         adapter->tx_scrq = NULL;
3042 tx_failed:
3043         for (i = 0; i < registered_queues; i++)
3044                 release_sub_crq_queue(adapter, allqueues[i], 1);
3045         kfree(allqueues);
3046         return -1;
3047 }
3048
3049 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3050 {
3051         struct device *dev = &adapter->vdev->dev;
3052         union ibmvnic_crq crq;
3053         int max_entries;
3054
3055         if (!retry) {
3056                 /* Sub-CRQ entries are 32 byte long */
3057                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3058
3059                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3060                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3061                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3062                         return;
3063                 }
3064
3065                 if (adapter->desired.mtu)
3066                         adapter->req_mtu = adapter->desired.mtu;
3067                 else
3068                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3069
3070                 if (!adapter->desired.tx_entries)
3071                         adapter->desired.tx_entries =
3072                                         adapter->max_tx_entries_per_subcrq;
3073                 if (!adapter->desired.rx_entries)
3074                         adapter->desired.rx_entries =
3075                                         adapter->max_rx_add_entries_per_subcrq;
3076
3077                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3078                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3079
3080                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3081                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3082                         adapter->desired.tx_entries = max_entries;
3083                 }
3084
3085                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3086                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3087                         adapter->desired.rx_entries = max_entries;
3088                 }
3089
3090                 if (adapter->desired.tx_entries)
3091                         adapter->req_tx_entries_per_subcrq =
3092                                         adapter->desired.tx_entries;
3093                 else
3094                         adapter->req_tx_entries_per_subcrq =
3095                                         adapter->max_tx_entries_per_subcrq;
3096
3097                 if (adapter->desired.rx_entries)
3098                         adapter->req_rx_add_entries_per_subcrq =
3099                                         adapter->desired.rx_entries;
3100                 else
3101                         adapter->req_rx_add_entries_per_subcrq =
3102                                         adapter->max_rx_add_entries_per_subcrq;
3103
3104                 if (adapter->desired.tx_queues)
3105                         adapter->req_tx_queues =
3106                                         adapter->desired.tx_queues;
3107                 else
3108                         adapter->req_tx_queues =
3109                                         adapter->opt_tx_comp_sub_queues;
3110
3111                 if (adapter->desired.rx_queues)
3112                         adapter->req_rx_queues =
3113                                         adapter->desired.rx_queues;
3114                 else
3115                         adapter->req_rx_queues =
3116                                         adapter->opt_rx_comp_queues;
3117
3118                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3119         }
3120
3121         memset(&crq, 0, sizeof(crq));
3122         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3123         crq.request_capability.cmd = REQUEST_CAPABILITY;
3124
3125         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3126         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3127         atomic_inc(&adapter->running_cap_crqs);
3128         ibmvnic_send_crq(adapter, &crq);
3129
3130         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3131         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3132         atomic_inc(&adapter->running_cap_crqs);
3133         ibmvnic_send_crq(adapter, &crq);
3134
3135         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3136         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3137         atomic_inc(&adapter->running_cap_crqs);
3138         ibmvnic_send_crq(adapter, &crq);
3139
3140         crq.request_capability.capability =
3141             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3142         crq.request_capability.number =
3143             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3144         atomic_inc(&adapter->running_cap_crqs);
3145         ibmvnic_send_crq(adapter, &crq);
3146
3147         crq.request_capability.capability =
3148             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3149         crq.request_capability.number =
3150             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3151         atomic_inc(&adapter->running_cap_crqs);
3152         ibmvnic_send_crq(adapter, &crq);
3153
3154         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3155         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3156         atomic_inc(&adapter->running_cap_crqs);
3157         ibmvnic_send_crq(adapter, &crq);
3158
3159         if (adapter->netdev->flags & IFF_PROMISC) {
3160                 if (adapter->promisc_supported) {
3161                         crq.request_capability.capability =
3162                             cpu_to_be16(PROMISC_REQUESTED);
3163                         crq.request_capability.number = cpu_to_be64(1);
3164                         atomic_inc(&adapter->running_cap_crqs);
3165                         ibmvnic_send_crq(adapter, &crq);
3166                 }
3167         } else {
3168                 crq.request_capability.capability =
3169                     cpu_to_be16(PROMISC_REQUESTED);
3170                 crq.request_capability.number = cpu_to_be64(0);
3171                 atomic_inc(&adapter->running_cap_crqs);
3172                 ibmvnic_send_crq(adapter, &crq);
3173         }
3174 }
3175
3176 static int pending_scrq(struct ibmvnic_adapter *adapter,
3177                         struct ibmvnic_sub_crq_queue *scrq)
3178 {
3179         union sub_crq *entry = &scrq->msgs[scrq->cur];
3180
3181         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3182                 return 1;
3183         else
3184                 return 0;
3185 }
3186
3187 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3188                                         struct ibmvnic_sub_crq_queue *scrq)
3189 {
3190         union sub_crq *entry;
3191         unsigned long flags;
3192
3193         spin_lock_irqsave(&scrq->lock, flags);
3194         entry = &scrq->msgs[scrq->cur];
3195         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3196                 if (++scrq->cur == scrq->size)
3197                         scrq->cur = 0;
3198         } else {
3199                 entry = NULL;
3200         }
3201         spin_unlock_irqrestore(&scrq->lock, flags);
3202
3203         return entry;
3204 }
3205
3206 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3207 {
3208         struct ibmvnic_crq_queue *queue = &adapter->crq;
3209         union ibmvnic_crq *crq;
3210
3211         crq = &queue->msgs[queue->cur];
3212         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3213                 if (++queue->cur == queue->size)
3214                         queue->cur = 0;
3215         } else {
3216                 crq = NULL;
3217         }
3218
3219         return crq;
3220 }
3221
3222 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3223 {
3224         switch (rc) {
3225         case H_PARAMETER:
3226                 dev_warn_ratelimited(dev,
3227                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3228                                      func, rc);
3229                 break;
3230         case H_CLOSED:
3231                 dev_warn_ratelimited(dev,
3232                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3233                                      func, rc);
3234                 break;
3235         default:
3236                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3237                 break;
3238         }
3239 }
3240
3241 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3242                        union sub_crq *sub_crq)
3243 {
3244         unsigned int ua = adapter->vdev->unit_address;
3245         struct device *dev = &adapter->vdev->dev;
3246         u64 *u64_crq = (u64 *)sub_crq;
3247         int rc;
3248
3249         netdev_dbg(adapter->netdev,
3250                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3251                    (unsigned long int)cpu_to_be64(remote_handle),
3252                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3253                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3254                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3255                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3256
3257         /* Make sure the hypervisor sees the complete request */
3258         mb();
3259
3260         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3261                                 cpu_to_be64(remote_handle),
3262                                 cpu_to_be64(u64_crq[0]),
3263                                 cpu_to_be64(u64_crq[1]),
3264                                 cpu_to_be64(u64_crq[2]),
3265                                 cpu_to_be64(u64_crq[3]));
3266
3267         if (rc)
3268                 print_subcrq_error(dev, rc, __func__);
3269
3270         return rc;
3271 }
3272
3273 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3274                                 u64 remote_handle, u64 ioba, u64 num_entries)
3275 {
3276         unsigned int ua = adapter->vdev->unit_address;
3277         struct device *dev = &adapter->vdev->dev;
3278         int rc;
3279
3280         /* Make sure the hypervisor sees the complete request */
3281         mb();
3282         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3283                                 cpu_to_be64(remote_handle),
3284                                 ioba, num_entries);
3285
3286         if (rc)
3287                 print_subcrq_error(dev, rc, __func__);
3288
3289         return rc;
3290 }
3291
3292 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3293                             union ibmvnic_crq *crq)
3294 {
3295         unsigned int ua = adapter->vdev->unit_address;
3296         struct device *dev = &adapter->vdev->dev;
3297         u64 *u64_crq = (u64 *)crq;
3298         int rc;
3299
3300         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3301                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3302                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3303
3304         if (!adapter->crq.active &&
3305             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3306                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3307                 return -EINVAL;
3308         }
3309
3310         /* Make sure the hypervisor sees the complete request */
3311         mb();
3312
3313         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3314                                 cpu_to_be64(u64_crq[0]),
3315                                 cpu_to_be64(u64_crq[1]));
3316
3317         if (rc) {
3318                 if (rc == H_CLOSED) {
3319                         dev_warn(dev, "CRQ Queue closed\n");
3320                         if (adapter->resetting)
3321                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3322                 }
3323
3324                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3325         }
3326
3327         return rc;
3328 }
3329
3330 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3331 {
3332         union ibmvnic_crq crq;
3333
3334         memset(&crq, 0, sizeof(crq));
3335         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3336         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3337         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3338
3339         return ibmvnic_send_crq(adapter, &crq);
3340 }
3341
3342 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3343 {
3344         union ibmvnic_crq crq;
3345
3346         memset(&crq, 0, sizeof(crq));
3347         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3348         crq.version_exchange.cmd = VERSION_EXCHANGE;
3349         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3350
3351         return ibmvnic_send_crq(adapter, &crq);
3352 }
3353
3354 struct vnic_login_client_data {
3355         u8      type;
3356         __be16  len;
3357         char    name[];
3358 } __packed;
3359
3360 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3361 {
3362         int len;
3363
3364         /* Calculate the amount of buffer space needed for the
3365          * vnic client data in the login buffer. There are four entries,
3366          * OS name, LPAR name, device name, and a null last entry.
3367          */
3368         len = 4 * sizeof(struct vnic_login_client_data);
3369         len += 6; /* "Linux" plus NULL */
3370         len += strlen(utsname()->nodename) + 1;
3371         len += strlen(adapter->netdev->name) + 1;
3372
3373         return len;
3374 }
3375
3376 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3377                                  struct vnic_login_client_data *vlcd)
3378 {
3379         const char *os_name = "Linux";
3380         int len;
3381
3382         /* Type 1 - LPAR OS */
3383         vlcd->type = 1;
3384         len = strlen(os_name) + 1;
3385         vlcd->len = cpu_to_be16(len);
3386         strncpy(vlcd->name, os_name, len);
3387         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3388
3389         /* Type 2 - LPAR name */
3390         vlcd->type = 2;
3391         len = strlen(utsname()->nodename) + 1;
3392         vlcd->len = cpu_to_be16(len);
3393         strncpy(vlcd->name, utsname()->nodename, len);
3394         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3395
3396         /* Type 3 - device name */
3397         vlcd->type = 3;
3398         len = strlen(adapter->netdev->name) + 1;
3399         vlcd->len = cpu_to_be16(len);
3400         strncpy(vlcd->name, adapter->netdev->name, len);
3401 }
3402
3403 static int send_login(struct ibmvnic_adapter *adapter)
3404 {
3405         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3406         struct ibmvnic_login_buffer *login_buffer;
3407         struct device *dev = &adapter->vdev->dev;
3408         dma_addr_t rsp_buffer_token;
3409         dma_addr_t buffer_token;
3410         size_t rsp_buffer_size;
3411         union ibmvnic_crq crq;
3412         size_t buffer_size;
3413         __be64 *tx_list_p;
3414         __be64 *rx_list_p;
3415         int client_data_len;
3416         struct vnic_login_client_data *vlcd;
3417         int i;
3418
3419         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3420                 netdev_err(adapter->netdev,
3421                            "RX or TX queues are not allocated, device login failed\n");
3422                 return -1;
3423         }
3424
3425         release_login_rsp_buffer(adapter);
3426         client_data_len = vnic_client_data_len(adapter);
3427
3428         buffer_size =
3429             sizeof(struct ibmvnic_login_buffer) +
3430             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3431             client_data_len;
3432
3433         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3434         if (!login_buffer)
3435                 goto buf_alloc_failed;
3436
3437         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3438                                       DMA_TO_DEVICE);
3439         if (dma_mapping_error(dev, buffer_token)) {
3440                 dev_err(dev, "Couldn't map login buffer\n");
3441                 goto buf_map_failed;
3442         }
3443
3444         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3445                           sizeof(u64) * adapter->req_tx_queues +
3446                           sizeof(u64) * adapter->req_rx_queues +
3447                           sizeof(u64) * adapter->req_rx_queues +
3448                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3449
3450         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3451         if (!login_rsp_buffer)
3452                 goto buf_rsp_alloc_failed;
3453
3454         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3455                                           rsp_buffer_size, DMA_FROM_DEVICE);
3456         if (dma_mapping_error(dev, rsp_buffer_token)) {
3457                 dev_err(dev, "Couldn't map login rsp buffer\n");
3458                 goto buf_rsp_map_failed;
3459         }
3460
3461         adapter->login_buf = login_buffer;
3462         adapter->login_buf_token = buffer_token;
3463         adapter->login_buf_sz = buffer_size;
3464         adapter->login_rsp_buf = login_rsp_buffer;
3465         adapter->login_rsp_buf_token = rsp_buffer_token;
3466         adapter->login_rsp_buf_sz = rsp_buffer_size;
3467
3468         login_buffer->len = cpu_to_be32(buffer_size);
3469         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3470         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3471         login_buffer->off_txcomp_subcrqs =
3472             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3473         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3474         login_buffer->off_rxcomp_subcrqs =
3475             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3476                         sizeof(u64) * adapter->req_tx_queues);
3477         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3478         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3479
3480         tx_list_p = (__be64 *)((char *)login_buffer +
3481                                       sizeof(struct ibmvnic_login_buffer));
3482         rx_list_p = (__be64 *)((char *)login_buffer +
3483                                       sizeof(struct ibmvnic_login_buffer) +
3484                                       sizeof(u64) * adapter->req_tx_queues);
3485
3486         for (i = 0; i < adapter->req_tx_queues; i++) {
3487                 if (adapter->tx_scrq[i]) {
3488                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3489                                                    crq_num);
3490                 }
3491         }
3492
3493         for (i = 0; i < adapter->req_rx_queues; i++) {
3494                 if (adapter->rx_scrq[i]) {
3495                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3496                                                    crq_num);
3497                 }
3498         }
3499
3500         /* Insert vNIC login client data */
3501         vlcd = (struct vnic_login_client_data *)
3502                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3503         login_buffer->client_data_offset =
3504                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3505         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3506
3507         vnic_add_client_data(adapter, vlcd);
3508
3509         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3510         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3511                 netdev_dbg(adapter->netdev, "%016lx\n",
3512                            ((unsigned long int *)(adapter->login_buf))[i]);
3513         }
3514
3515         memset(&crq, 0, sizeof(crq));
3516         crq.login.first = IBMVNIC_CRQ_CMD;
3517         crq.login.cmd = LOGIN;
3518         crq.login.ioba = cpu_to_be32(buffer_token);
3519         crq.login.len = cpu_to_be32(buffer_size);
3520         ibmvnic_send_crq(adapter, &crq);
3521
3522         return 0;
3523
3524 buf_rsp_map_failed:
3525         kfree(login_rsp_buffer);
3526 buf_rsp_alloc_failed:
3527         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3528 buf_map_failed:
3529         kfree(login_buffer);
3530 buf_alloc_failed:
3531         return -1;
3532 }
3533
3534 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3535                             u32 len, u8 map_id)
3536 {
3537         union ibmvnic_crq crq;
3538
3539         memset(&crq, 0, sizeof(crq));
3540         crq.request_map.first = IBMVNIC_CRQ_CMD;
3541         crq.request_map.cmd = REQUEST_MAP;
3542         crq.request_map.map_id = map_id;
3543         crq.request_map.ioba = cpu_to_be32(addr);
3544         crq.request_map.len = cpu_to_be32(len);
3545         return ibmvnic_send_crq(adapter, &crq);
3546 }
3547
3548 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3549 {
3550         union ibmvnic_crq crq;
3551
3552         memset(&crq, 0, sizeof(crq));
3553         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3554         crq.request_unmap.cmd = REQUEST_UNMAP;
3555         crq.request_unmap.map_id = map_id;
3556         return ibmvnic_send_crq(adapter, &crq);
3557 }
3558
3559 static void send_map_query(struct ibmvnic_adapter *adapter)
3560 {
3561         union ibmvnic_crq crq;
3562
3563         memset(&crq, 0, sizeof(crq));
3564         crq.query_map.first = IBMVNIC_CRQ_CMD;
3565         crq.query_map.cmd = QUERY_MAP;
3566         ibmvnic_send_crq(adapter, &crq);
3567 }
3568
3569 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3570 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3571 {
3572         union ibmvnic_crq crq;
3573
3574         atomic_set(&adapter->running_cap_crqs, 0);
3575         memset(&crq, 0, sizeof(crq));
3576         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3577         crq.query_capability.cmd = QUERY_CAPABILITY;
3578
3579         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3580         atomic_inc(&adapter->running_cap_crqs);
3581         ibmvnic_send_crq(adapter, &crq);
3582
3583         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3584         atomic_inc(&adapter->running_cap_crqs);
3585         ibmvnic_send_crq(adapter, &crq);
3586
3587         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3588         atomic_inc(&adapter->running_cap_crqs);
3589         ibmvnic_send_crq(adapter, &crq);
3590
3591         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3592         atomic_inc(&adapter->running_cap_crqs);
3593         ibmvnic_send_crq(adapter, &crq);
3594
3595         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3596         atomic_inc(&adapter->running_cap_crqs);
3597         ibmvnic_send_crq(adapter, &crq);
3598
3599         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3600         atomic_inc(&adapter->running_cap_crqs);
3601         ibmvnic_send_crq(adapter, &crq);
3602
3603         crq.query_capability.capability =
3604             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3605         atomic_inc(&adapter->running_cap_crqs);
3606         ibmvnic_send_crq(adapter, &crq);
3607
3608         crq.query_capability.capability =
3609             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3610         atomic_inc(&adapter->running_cap_crqs);
3611         ibmvnic_send_crq(adapter, &crq);
3612
3613         crq.query_capability.capability =
3614             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3615         atomic_inc(&adapter->running_cap_crqs);
3616         ibmvnic_send_crq(adapter, &crq);
3617
3618         crq.query_capability.capability =
3619             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3620         atomic_inc(&adapter->running_cap_crqs);
3621         ibmvnic_send_crq(adapter, &crq);
3622
3623         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3624         atomic_inc(&adapter->running_cap_crqs);
3625         ibmvnic_send_crq(adapter, &crq);
3626
3627         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3628         atomic_inc(&adapter->running_cap_crqs);
3629         ibmvnic_send_crq(adapter, &crq);
3630
3631         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3632         atomic_inc(&adapter->running_cap_crqs);
3633         ibmvnic_send_crq(adapter, &crq);
3634
3635         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3636         atomic_inc(&adapter->running_cap_crqs);
3637         ibmvnic_send_crq(adapter, &crq);
3638
3639         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3640         atomic_inc(&adapter->running_cap_crqs);
3641         ibmvnic_send_crq(adapter, &crq);
3642
3643         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3644         atomic_inc(&adapter->running_cap_crqs);
3645         ibmvnic_send_crq(adapter, &crq);
3646
3647         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3648         atomic_inc(&adapter->running_cap_crqs);
3649         ibmvnic_send_crq(adapter, &crq);
3650
3651         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3652         atomic_inc(&adapter->running_cap_crqs);
3653         ibmvnic_send_crq(adapter, &crq);
3654
3655         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3656         atomic_inc(&adapter->running_cap_crqs);
3657         ibmvnic_send_crq(adapter, &crq);
3658
3659         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3660         atomic_inc(&adapter->running_cap_crqs);
3661         ibmvnic_send_crq(adapter, &crq);
3662
3663         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3664         atomic_inc(&adapter->running_cap_crqs);
3665         ibmvnic_send_crq(adapter, &crq);
3666
3667         crq.query_capability.capability =
3668                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3669         atomic_inc(&adapter->running_cap_crqs);
3670         ibmvnic_send_crq(adapter, &crq);
3671
3672         crq.query_capability.capability =
3673                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3674         atomic_inc(&adapter->running_cap_crqs);
3675         ibmvnic_send_crq(adapter, &crq);
3676
3677         crq.query_capability.capability =
3678                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3679         atomic_inc(&adapter->running_cap_crqs);
3680         ibmvnic_send_crq(adapter, &crq);
3681
3682         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3683         atomic_inc(&adapter->running_cap_crqs);
3684         ibmvnic_send_crq(adapter, &crq);
3685 }
3686
3687 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3688                                 struct ibmvnic_adapter *adapter)
3689 {
3690         struct device *dev = &adapter->vdev->dev;
3691
3692         if (crq->get_vpd_size_rsp.rc.code) {
3693                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3694                         crq->get_vpd_size_rsp.rc.code);
3695                 complete(&adapter->fw_done);
3696                 return;
3697         }
3698
3699         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3700         complete(&adapter->fw_done);
3701 }
3702
3703 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3704                            struct ibmvnic_adapter *adapter)
3705 {
3706         struct device *dev = &adapter->vdev->dev;
3707         unsigned char *substr = NULL;
3708         u8 fw_level_len = 0;
3709
3710         memset(adapter->fw_version, 0, 32);
3711
3712         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3713                          DMA_FROM_DEVICE);
3714
3715         if (crq->get_vpd_rsp.rc.code) {
3716                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3717                         crq->get_vpd_rsp.rc.code);
3718                 goto complete;
3719         }
3720
3721         /* get the position of the firmware version info
3722          * located after the ASCII 'RM' substring in the buffer
3723          */
3724         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3725         if (!substr) {
3726                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3727                 goto complete;
3728         }
3729
3730         /* get length of firmware level ASCII substring */
3731         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3732                 fw_level_len = *(substr + 2);
3733         } else {
3734                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3735                 goto complete;
3736         }
3737
3738         /* copy firmware version string from vpd into adapter */
3739         if ((substr + 3 + fw_level_len) <
3740             (adapter->vpd->buff + adapter->vpd->len)) {
3741                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3742         } else {
3743                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3744         }
3745
3746 complete:
3747         if (adapter->fw_version[0] == '\0')
3748                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3749         complete(&adapter->fw_done);
3750 }
3751
3752 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3753 {
3754         struct device *dev = &adapter->vdev->dev;
3755         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3756         netdev_features_t old_hw_features = 0;
3757         union ibmvnic_crq crq;
3758         int i;
3759
3760         dma_unmap_single(dev, adapter->ip_offload_tok,
3761                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3762
3763         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3764         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3765                 netdev_dbg(adapter->netdev, "%016lx\n",
3766                            ((unsigned long int *)(buf))[i]);
3767
3768         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3769         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3770         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3771                    buf->tcp_ipv4_chksum);
3772         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3773                    buf->tcp_ipv6_chksum);
3774         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3775                    buf->udp_ipv4_chksum);
3776         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3777                    buf->udp_ipv6_chksum);
3778         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3779                    buf->large_tx_ipv4);
3780         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3781                    buf->large_tx_ipv6);
3782         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3783                    buf->large_rx_ipv4);
3784         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3785                    buf->large_rx_ipv6);
3786         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3787                    buf->max_ipv4_header_size);
3788         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3789                    buf->max_ipv6_header_size);
3790         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3791                    buf->max_tcp_header_size);
3792         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3793                    buf->max_udp_header_size);
3794         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3795                    buf->max_large_tx_size);
3796         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3797                    buf->max_large_rx_size);
3798         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3799                    buf->ipv6_extension_header);
3800         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3801                    buf->tcp_pseudosum_req);
3802         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3803                    buf->num_ipv6_ext_headers);
3804         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3805                    buf->off_ipv6_ext_headers);
3806
3807         adapter->ip_offload_ctrl_tok =
3808             dma_map_single(dev, &adapter->ip_offload_ctrl,
3809                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3810
3811         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3812                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3813                 return;
3814         }
3815
3816         adapter->ip_offload_ctrl.len =
3817             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3818         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3819         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3820         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3821         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3822         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3823         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3824         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3825         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3826         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3827
3828         /* large_rx disabled for now, additional features needed */
3829         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3830         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3831
3832         if (adapter->state != VNIC_PROBING) {
3833                 old_hw_features = adapter->netdev->hw_features;
3834                 adapter->netdev->hw_features = 0;
3835         }
3836
3837         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
3838
3839         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3840                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
3841
3842         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3843                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
3844
3845         if ((adapter->netdev->features &
3846             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3847                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
3848
3849         if (buf->large_tx_ipv4)
3850                 adapter->netdev->hw_features |= NETIF_F_TSO;
3851         if (buf->large_tx_ipv6)
3852                 adapter->netdev->hw_features |= NETIF_F_TSO6;
3853
3854         if (adapter->state == VNIC_PROBING) {
3855                 adapter->netdev->features |= adapter->netdev->hw_features;
3856         } else if (old_hw_features != adapter->netdev->hw_features) {
3857                 netdev_features_t tmp = 0;
3858
3859                 /* disable features no longer supported */
3860                 adapter->netdev->features &= adapter->netdev->hw_features;
3861                 /* turn on features now supported if previously enabled */
3862                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
3863                         adapter->netdev->hw_features;
3864                 adapter->netdev->features |=
3865                                 tmp & adapter->netdev->wanted_features;
3866         }
3867
3868         memset(&crq, 0, sizeof(crq));
3869         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3870         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3871         crq.control_ip_offload.len =
3872             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3873         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3874         ibmvnic_send_crq(adapter, &crq);
3875 }
3876
3877 static const char *ibmvnic_fw_err_cause(u16 cause)
3878 {
3879         switch (cause) {
3880         case ADAPTER_PROBLEM:
3881                 return "adapter problem";
3882         case BUS_PROBLEM:
3883                 return "bus problem";
3884         case FW_PROBLEM:
3885                 return "firmware problem";
3886         case DD_PROBLEM:
3887                 return "device driver problem";
3888         case EEH_RECOVERY:
3889                 return "EEH recovery";
3890         case FW_UPDATED:
3891                 return "firmware updated";
3892         case LOW_MEMORY:
3893                 return "low Memory";
3894         default:
3895                 return "unknown";
3896         }
3897 }
3898
3899 static void handle_error_indication(union ibmvnic_crq *crq,
3900                                     struct ibmvnic_adapter *adapter)
3901 {
3902         struct device *dev = &adapter->vdev->dev;
3903         u16 cause;
3904
3905         cause = be16_to_cpu(crq->error_indication.error_cause);
3906
3907         dev_warn_ratelimited(dev,
3908                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3909                              crq->error_indication.flags
3910                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3911                              ibmvnic_fw_err_cause(cause));
3912
3913         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3914                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3915         else
3916                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3917 }
3918
3919 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3920                                  struct ibmvnic_adapter *adapter)
3921 {
3922         struct net_device *netdev = adapter->netdev;
3923         struct device *dev = &adapter->vdev->dev;
3924         long rc;
3925
3926         rc = crq->change_mac_addr_rsp.rc.code;
3927         if (rc) {
3928                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3929                 goto out;
3930         }
3931         ether_addr_copy(netdev->dev_addr,
3932                         &crq->change_mac_addr_rsp.mac_addr[0]);
3933 out:
3934         complete(&adapter->fw_done);
3935         return rc;
3936 }
3937
3938 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3939                                    struct ibmvnic_adapter *adapter)
3940 {
3941         struct device *dev = &adapter->vdev->dev;
3942         u64 *req_value;
3943         char *name;
3944
3945         atomic_dec(&adapter->running_cap_crqs);
3946         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3947         case REQ_TX_QUEUES:
3948                 req_value = &adapter->req_tx_queues;
3949                 name = "tx";
3950                 break;
3951         case REQ_RX_QUEUES:
3952                 req_value = &adapter->req_rx_queues;
3953                 name = "rx";
3954                 break;
3955         case REQ_RX_ADD_QUEUES:
3956                 req_value = &adapter->req_rx_add_queues;
3957                 name = "rx_add";
3958                 break;
3959         case REQ_TX_ENTRIES_PER_SUBCRQ:
3960                 req_value = &adapter->req_tx_entries_per_subcrq;
3961                 name = "tx_entries_per_subcrq";
3962                 break;
3963         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3964                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3965                 name = "rx_add_entries_per_subcrq";
3966                 break;
3967         case REQ_MTU:
3968                 req_value = &adapter->req_mtu;
3969                 name = "mtu";
3970                 break;
3971         case PROMISC_REQUESTED:
3972                 req_value = &adapter->promisc;
3973                 name = "promisc";
3974                 break;
3975         default:
3976                 dev_err(dev, "Got invalid cap request rsp %d\n",
3977                         crq->request_capability.capability);
3978                 return;
3979         }
3980
3981         switch (crq->request_capability_rsp.rc.code) {
3982         case SUCCESS:
3983                 break;
3984         case PARTIALSUCCESS:
3985                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3986                          *req_value,
3987                          (long int)be64_to_cpu(crq->request_capability_rsp.
3988                                                number), name);
3989
3990                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3991                     REQ_MTU) {
3992                         pr_err("mtu of %llu is not supported. Reverting.\n",
3993                                *req_value);
3994                         *req_value = adapter->fallback.mtu;
3995                 } else {
3996                         *req_value =
3997                                 be64_to_cpu(crq->request_capability_rsp.number);
3998                 }
3999
4000                 ibmvnic_send_req_caps(adapter, 1);
4001                 return;
4002         default:
4003                 dev_err(dev, "Error %d in request cap rsp\n",
4004                         crq->request_capability_rsp.rc.code);
4005                 return;
4006         }
4007
4008         /* Done receiving requested capabilities, query IP offload support */
4009         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4010                 union ibmvnic_crq newcrq;
4011                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4012                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4013                     &adapter->ip_offload_buf;
4014
4015                 adapter->wait_capability = false;
4016                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4017                                                          buf_sz,
4018                                                          DMA_FROM_DEVICE);
4019
4020                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4021                         if (!firmware_has_feature(FW_FEATURE_CMO))
4022                                 dev_err(dev, "Couldn't map offload buffer\n");
4023                         return;
4024                 }
4025
4026                 memset(&newcrq, 0, sizeof(newcrq));
4027                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4028                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4029                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4030                 newcrq.query_ip_offload.ioba =
4031                     cpu_to_be32(adapter->ip_offload_tok);
4032
4033                 ibmvnic_send_crq(adapter, &newcrq);
4034         }
4035 }
4036
4037 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4038                             struct ibmvnic_adapter *adapter)
4039 {
4040         struct device *dev = &adapter->vdev->dev;
4041         struct net_device *netdev = adapter->netdev;
4042         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4043         struct ibmvnic_login_buffer *login = adapter->login_buf;
4044         int i;
4045
4046         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4047                          DMA_TO_DEVICE);
4048         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4049                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4050
4051         /* If the number of queues requested can't be allocated by the
4052          * server, the login response will return with code 1. We will need
4053          * to resend the login buffer with fewer queues requested.
4054          */
4055         if (login_rsp_crq->generic.rc.code) {
4056                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4057                 complete(&adapter->init_done);
4058                 return 0;
4059         }
4060
4061         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4062
4063         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4064         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4065                 netdev_dbg(adapter->netdev, "%016lx\n",
4066                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4067         }
4068
4069         /* Sanity checks */
4070         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4071             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4072              adapter->req_rx_add_queues !=
4073              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4074                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4075                 ibmvnic_remove(adapter->vdev);
4076                 return -EIO;
4077         }
4078         release_login_buffer(adapter);
4079         complete(&adapter->init_done);
4080
4081         return 0;
4082 }
4083
4084 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4085                                      struct ibmvnic_adapter *adapter)
4086 {
4087         struct device *dev = &adapter->vdev->dev;
4088         long rc;
4089
4090         rc = crq->request_unmap_rsp.rc.code;
4091         if (rc)
4092                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4093 }
4094
4095 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4096                                  struct ibmvnic_adapter *adapter)
4097 {
4098         struct net_device *netdev = adapter->netdev;
4099         struct device *dev = &adapter->vdev->dev;
4100         long rc;
4101
4102         rc = crq->query_map_rsp.rc.code;
4103         if (rc) {
4104                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4105                 return;
4106         }
4107         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4108                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4109                    crq->query_map_rsp.free_pages);
4110 }
4111
4112 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4113                                  struct ibmvnic_adapter *adapter)
4114 {
4115         struct net_device *netdev = adapter->netdev;
4116         struct device *dev = &adapter->vdev->dev;
4117         long rc;
4118
4119         atomic_dec(&adapter->running_cap_crqs);
4120         netdev_dbg(netdev, "Outstanding queries: %d\n",
4121                    atomic_read(&adapter->running_cap_crqs));
4122         rc = crq->query_capability.rc.code;
4123         if (rc) {
4124                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4125                 goto out;
4126         }
4127
4128         switch (be16_to_cpu(crq->query_capability.capability)) {
4129         case MIN_TX_QUEUES:
4130                 adapter->min_tx_queues =
4131                     be64_to_cpu(crq->query_capability.number);
4132                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4133                            adapter->min_tx_queues);
4134                 break;
4135         case MIN_RX_QUEUES:
4136                 adapter->min_rx_queues =
4137                     be64_to_cpu(crq->query_capability.number);
4138                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4139                            adapter->min_rx_queues);
4140                 break;
4141         case MIN_RX_ADD_QUEUES:
4142                 adapter->min_rx_add_queues =
4143                     be64_to_cpu(crq->query_capability.number);
4144                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4145                            adapter->min_rx_add_queues);
4146                 break;
4147         case MAX_TX_QUEUES:
4148                 adapter->max_tx_queues =
4149                     be64_to_cpu(crq->query_capability.number);
4150                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4151                            adapter->max_tx_queues);
4152                 break;
4153         case MAX_RX_QUEUES:
4154                 adapter->max_rx_queues =
4155                     be64_to_cpu(crq->query_capability.number);
4156                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4157                            adapter->max_rx_queues);
4158                 break;
4159         case MAX_RX_ADD_QUEUES:
4160                 adapter->max_rx_add_queues =
4161                     be64_to_cpu(crq->query_capability.number);
4162                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4163                            adapter->max_rx_add_queues);
4164                 break;
4165         case MIN_TX_ENTRIES_PER_SUBCRQ:
4166                 adapter->min_tx_entries_per_subcrq =
4167                     be64_to_cpu(crq->query_capability.number);
4168                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4169                            adapter->min_tx_entries_per_subcrq);
4170                 break;
4171         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4172                 adapter->min_rx_add_entries_per_subcrq =
4173                     be64_to_cpu(crq->query_capability.number);
4174                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4175                            adapter->min_rx_add_entries_per_subcrq);
4176                 break;
4177         case MAX_TX_ENTRIES_PER_SUBCRQ:
4178                 adapter->max_tx_entries_per_subcrq =
4179                     be64_to_cpu(crq->query_capability.number);
4180                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4181                            adapter->max_tx_entries_per_subcrq);
4182                 break;
4183         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4184                 adapter->max_rx_add_entries_per_subcrq =
4185                     be64_to_cpu(crq->query_capability.number);
4186                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4187                            adapter->max_rx_add_entries_per_subcrq);
4188                 break;
4189         case TCP_IP_OFFLOAD:
4190                 adapter->tcp_ip_offload =
4191                     be64_to_cpu(crq->query_capability.number);
4192                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4193                            adapter->tcp_ip_offload);
4194                 break;
4195         case PROMISC_SUPPORTED:
4196                 adapter->promisc_supported =
4197                     be64_to_cpu(crq->query_capability.number);
4198                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4199                            adapter->promisc_supported);
4200                 break;
4201         case MIN_MTU:
4202                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4203                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4204                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4205                 break;
4206         case MAX_MTU:
4207                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4208                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4209                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4210                 break;
4211         case MAX_MULTICAST_FILTERS:
4212                 adapter->max_multicast_filters =
4213                     be64_to_cpu(crq->query_capability.number);
4214                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4215                            adapter->max_multicast_filters);
4216                 break;
4217         case VLAN_HEADER_INSERTION:
4218                 adapter->vlan_header_insertion =
4219                     be64_to_cpu(crq->query_capability.number);
4220                 if (adapter->vlan_header_insertion)
4221                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4222                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4223                            adapter->vlan_header_insertion);
4224                 break;
4225         case RX_VLAN_HEADER_INSERTION:
4226                 adapter->rx_vlan_header_insertion =
4227                     be64_to_cpu(crq->query_capability.number);
4228                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4229                            adapter->rx_vlan_header_insertion);
4230                 break;
4231         case MAX_TX_SG_ENTRIES:
4232                 adapter->max_tx_sg_entries =
4233                     be64_to_cpu(crq->query_capability.number);
4234                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4235                            adapter->max_tx_sg_entries);
4236                 break;
4237         case RX_SG_SUPPORTED:
4238                 adapter->rx_sg_supported =
4239                     be64_to_cpu(crq->query_capability.number);
4240                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4241                            adapter->rx_sg_supported);
4242                 break;
4243         case OPT_TX_COMP_SUB_QUEUES:
4244                 adapter->opt_tx_comp_sub_queues =
4245                     be64_to_cpu(crq->query_capability.number);
4246                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4247                            adapter->opt_tx_comp_sub_queues);
4248                 break;
4249         case OPT_RX_COMP_QUEUES:
4250                 adapter->opt_rx_comp_queues =
4251                     be64_to_cpu(crq->query_capability.number);
4252                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4253                            adapter->opt_rx_comp_queues);
4254                 break;
4255         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4256                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4257                     be64_to_cpu(crq->query_capability.number);
4258                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4259                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4260                 break;
4261         case OPT_TX_ENTRIES_PER_SUBCRQ:
4262                 adapter->opt_tx_entries_per_subcrq =
4263                     be64_to_cpu(crq->query_capability.number);
4264                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4265                            adapter->opt_tx_entries_per_subcrq);
4266                 break;
4267         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4268                 adapter->opt_rxba_entries_per_subcrq =
4269                     be64_to_cpu(crq->query_capability.number);
4270                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4271                            adapter->opt_rxba_entries_per_subcrq);
4272                 break;
4273         case TX_RX_DESC_REQ:
4274                 adapter->tx_rx_desc_req = crq->query_capability.number;
4275                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4276                            adapter->tx_rx_desc_req);
4277                 break;
4278
4279         default:
4280                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4281                            crq->query_capability.capability);
4282         }
4283
4284 out:
4285         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4286                 adapter->wait_capability = false;
4287                 ibmvnic_send_req_caps(adapter, 0);
4288         }
4289 }
4290
4291 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4292 {
4293         union ibmvnic_crq crq;
4294         int rc;
4295
4296         memset(&crq, 0, sizeof(crq));
4297         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4298         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4299         init_completion(&adapter->fw_done);
4300         rc = ibmvnic_send_crq(adapter, &crq);
4301         if (rc)
4302                 return rc;
4303         wait_for_completion(&adapter->fw_done);
4304         return adapter->fw_done_rc ? -EIO : 0;
4305 }
4306
4307 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4308                                        struct ibmvnic_adapter *adapter)
4309 {
4310         struct net_device *netdev = adapter->netdev;
4311         int rc;
4312
4313         rc = crq->query_phys_parms_rsp.rc.code;
4314         if (rc) {
4315                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4316                 return rc;
4317         }
4318         switch (cpu_to_be32(crq->query_phys_parms_rsp.speed)) {
4319         case IBMVNIC_10MBPS:
4320                 adapter->speed = SPEED_10;
4321                 break;
4322         case IBMVNIC_100MBPS:
4323                 adapter->speed = SPEED_100;
4324                 break;
4325         case IBMVNIC_1GBPS:
4326                 adapter->speed = SPEED_1000;
4327                 break;
4328         case IBMVNIC_10GBP:
4329                 adapter->speed = SPEED_10000;
4330                 break;
4331         case IBMVNIC_25GBPS:
4332                 adapter->speed = SPEED_25000;
4333                 break;
4334         case IBMVNIC_40GBPS:
4335                 adapter->speed = SPEED_40000;
4336                 break;
4337         case IBMVNIC_50GBPS:
4338                 adapter->speed = SPEED_50000;
4339                 break;
4340         case IBMVNIC_100GBPS:
4341                 adapter->speed = SPEED_100000;
4342                 break;
4343         default:
4344                 netdev_warn(netdev, "Unknown speed 0x%08x\n",
4345                             cpu_to_be32(crq->query_phys_parms_rsp.speed));
4346                 adapter->speed = SPEED_UNKNOWN;
4347         }
4348         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4349                 adapter->duplex = DUPLEX_FULL;
4350         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4351                 adapter->duplex = DUPLEX_HALF;
4352         else
4353                 adapter->duplex = DUPLEX_UNKNOWN;
4354
4355         return rc;
4356 }
4357
4358 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4359                                struct ibmvnic_adapter *adapter)
4360 {
4361         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4362         struct net_device *netdev = adapter->netdev;
4363         struct device *dev = &adapter->vdev->dev;
4364         u64 *u64_crq = (u64 *)crq;
4365         long rc;
4366
4367         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4368                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4369                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4370         switch (gen_crq->first) {
4371         case IBMVNIC_CRQ_INIT_RSP:
4372                 switch (gen_crq->cmd) {
4373                 case IBMVNIC_CRQ_INIT:
4374                         dev_info(dev, "Partner initialized\n");
4375                         adapter->from_passive_init = true;
4376                         adapter->failover_pending = false;
4377                         if (!completion_done(&adapter->init_done)) {
4378                                 complete(&adapter->init_done);
4379                                 adapter->init_done_rc = -EIO;
4380                         }
4381                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4382                         break;
4383                 case IBMVNIC_CRQ_INIT_COMPLETE:
4384                         dev_info(dev, "Partner initialization complete\n");
4385                         adapter->crq.active = true;
4386                         send_version_xchg(adapter);
4387                         break;
4388                 default:
4389                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4390                 }
4391                 return;
4392         case IBMVNIC_CRQ_XPORT_EVENT:
4393                 netif_carrier_off(netdev);
4394                 adapter->crq.active = false;
4395                 if (adapter->resetting)
4396                         adapter->force_reset_recovery = true;
4397                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4398                         dev_info(dev, "Migrated, re-enabling adapter\n");
4399                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4400                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4401                         dev_info(dev, "Backing device failover detected\n");
4402                         adapter->failover_pending = true;
4403                 } else {
4404                         /* The adapter lost the connection */
4405                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4406                                 gen_crq->cmd);
4407                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4408                 }
4409                 return;
4410         case IBMVNIC_CRQ_CMD_RSP:
4411                 break;
4412         default:
4413                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4414                         gen_crq->first);
4415                 return;
4416         }
4417
4418         switch (gen_crq->cmd) {
4419         case VERSION_EXCHANGE_RSP:
4420                 rc = crq->version_exchange_rsp.rc.code;
4421                 if (rc) {
4422                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4423                         break;
4424                 }
4425                 dev_info(dev, "Partner protocol version is %d\n",
4426                          crq->version_exchange_rsp.version);
4427                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4428                     ibmvnic_version)
4429                         ibmvnic_version =
4430                             be16_to_cpu(crq->version_exchange_rsp.version);
4431                 send_cap_queries(adapter);
4432                 break;
4433         case QUERY_CAPABILITY_RSP:
4434                 handle_query_cap_rsp(crq, adapter);
4435                 break;
4436         case QUERY_MAP_RSP:
4437                 handle_query_map_rsp(crq, adapter);
4438                 break;
4439         case REQUEST_MAP_RSP:
4440                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4441                 complete(&adapter->fw_done);
4442                 break;
4443         case REQUEST_UNMAP_RSP:
4444                 handle_request_unmap_rsp(crq, adapter);
4445                 break;
4446         case REQUEST_CAPABILITY_RSP:
4447                 handle_request_cap_rsp(crq, adapter);
4448                 break;
4449         case LOGIN_RSP:
4450                 netdev_dbg(netdev, "Got Login Response\n");
4451                 handle_login_rsp(crq, adapter);
4452                 break;
4453         case LOGICAL_LINK_STATE_RSP:
4454                 netdev_dbg(netdev,
4455                            "Got Logical Link State Response, state: %d rc: %d\n",
4456                            crq->logical_link_state_rsp.link_state,
4457                            crq->logical_link_state_rsp.rc.code);
4458                 adapter->logical_link_state =
4459                     crq->logical_link_state_rsp.link_state;
4460                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4461                 complete(&adapter->init_done);
4462                 break;
4463         case LINK_STATE_INDICATION:
4464                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4465                 adapter->phys_link_state =
4466                     crq->link_state_indication.phys_link_state;
4467                 adapter->logical_link_state =
4468                     crq->link_state_indication.logical_link_state;
4469                 if (adapter->phys_link_state && adapter->logical_link_state)
4470                         netif_carrier_on(netdev);
4471                 else
4472                         netif_carrier_off(netdev);
4473                 break;
4474         case CHANGE_MAC_ADDR_RSP:
4475                 netdev_dbg(netdev, "Got MAC address change Response\n");
4476                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4477                 break;
4478         case ERROR_INDICATION:
4479                 netdev_dbg(netdev, "Got Error Indication\n");
4480                 handle_error_indication(crq, adapter);
4481                 break;
4482         case REQUEST_STATISTICS_RSP:
4483                 netdev_dbg(netdev, "Got Statistics Response\n");
4484                 complete(&adapter->stats_done);
4485                 break;
4486         case QUERY_IP_OFFLOAD_RSP:
4487                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4488                 handle_query_ip_offload_rsp(adapter);
4489                 break;
4490         case MULTICAST_CTRL_RSP:
4491                 netdev_dbg(netdev, "Got multicast control Response\n");
4492                 break;
4493         case CONTROL_IP_OFFLOAD_RSP:
4494                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4495                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4496                                  sizeof(adapter->ip_offload_ctrl),
4497                                  DMA_TO_DEVICE);
4498                 complete(&adapter->init_done);
4499                 break;
4500         case COLLECT_FW_TRACE_RSP:
4501                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4502                 complete(&adapter->fw_done);
4503                 break;
4504         case GET_VPD_SIZE_RSP:
4505                 handle_vpd_size_rsp(crq, adapter);
4506                 break;
4507         case GET_VPD_RSP:
4508                 handle_vpd_rsp(crq, adapter);
4509                 break;
4510         case QUERY_PHYS_PARMS_RSP:
4511                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4512                 complete(&adapter->fw_done);
4513                 break;
4514         default:
4515                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4516                            gen_crq->cmd);
4517         }
4518 }
4519
4520 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4521 {
4522         struct ibmvnic_adapter *adapter = instance;
4523
4524         tasklet_schedule(&adapter->tasklet);
4525         return IRQ_HANDLED;
4526 }
4527
4528 static void ibmvnic_tasklet(void *data)
4529 {
4530         struct ibmvnic_adapter *adapter = data;
4531         struct ibmvnic_crq_queue *queue = &adapter->crq;
4532         union ibmvnic_crq *crq;
4533         unsigned long flags;
4534         bool done = false;
4535
4536         spin_lock_irqsave(&queue->lock, flags);
4537         while (!done) {
4538                 /* Pull all the valid messages off the CRQ */
4539                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4540                         ibmvnic_handle_crq(crq, adapter);
4541                         crq->generic.first = 0;
4542                 }
4543
4544                 /* remain in tasklet until all
4545                  * capabilities responses are received
4546                  */
4547                 if (!adapter->wait_capability)
4548                         done = true;
4549         }
4550         /* if capabilities CRQ's were sent in this tasklet, the following
4551          * tasklet must wait until all responses are received
4552          */
4553         if (atomic_read(&adapter->running_cap_crqs) != 0)
4554                 adapter->wait_capability = true;
4555         spin_unlock_irqrestore(&queue->lock, flags);
4556 }
4557
4558 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4559 {
4560         struct vio_dev *vdev = adapter->vdev;
4561         int rc;
4562
4563         do {
4564                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4565         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4566
4567         if (rc)
4568                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4569
4570         return rc;
4571 }
4572
4573 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4574 {
4575         struct ibmvnic_crq_queue *crq = &adapter->crq;
4576         struct device *dev = &adapter->vdev->dev;
4577         struct vio_dev *vdev = adapter->vdev;
4578         int rc;
4579
4580         /* Close the CRQ */
4581         do {
4582                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4583         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4584
4585         /* Clean out the queue */
4586         memset(crq->msgs, 0, PAGE_SIZE);
4587         crq->cur = 0;
4588         crq->active = false;
4589
4590         /* And re-open it again */
4591         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4592                                 crq->msg_token, PAGE_SIZE);
4593
4594         if (rc == H_CLOSED)
4595                 /* Adapter is good, but other end is not ready */
4596                 dev_warn(dev, "Partner adapter not ready\n");
4597         else if (rc != 0)
4598                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4599
4600         return rc;
4601 }
4602
4603 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4604 {
4605         struct ibmvnic_crq_queue *crq = &adapter->crq;
4606         struct vio_dev *vdev = adapter->vdev;
4607         long rc;
4608
4609         if (!crq->msgs)
4610                 return;
4611
4612         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4613         free_irq(vdev->irq, adapter);
4614         tasklet_kill(&adapter->tasklet);
4615         do {
4616                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4617         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4618
4619         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4620                          DMA_BIDIRECTIONAL);
4621         free_page((unsigned long)crq->msgs);
4622         crq->msgs = NULL;
4623         crq->active = false;
4624 }
4625
4626 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4627 {
4628         struct ibmvnic_crq_queue *crq = &adapter->crq;
4629         struct device *dev = &adapter->vdev->dev;
4630         struct vio_dev *vdev = adapter->vdev;
4631         int rc, retrc = -ENOMEM;
4632
4633         if (crq->msgs)
4634                 return 0;
4635
4636         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4637         /* Should we allocate more than one page? */
4638
4639         if (!crq->msgs)
4640                 return -ENOMEM;
4641
4642         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4643         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4644                                         DMA_BIDIRECTIONAL);
4645         if (dma_mapping_error(dev, crq->msg_token))
4646                 goto map_failed;
4647
4648         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4649                                 crq->msg_token, PAGE_SIZE);
4650
4651         if (rc == H_RESOURCE)
4652                 /* maybe kexecing and resource is busy. try a reset */
4653                 rc = ibmvnic_reset_crq(adapter);
4654         retrc = rc;
4655
4656         if (rc == H_CLOSED) {
4657                 dev_warn(dev, "Partner adapter not ready\n");
4658         } else if (rc) {
4659                 dev_warn(dev, "Error %d opening adapter\n", rc);
4660                 goto reg_crq_failed;
4661         }
4662
4663         retrc = 0;
4664
4665         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4666                      (unsigned long)adapter);
4667
4668         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4669         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4670                  adapter->vdev->unit_address);
4671         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4672         if (rc) {
4673                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4674                         vdev->irq, rc);
4675                 goto req_irq_failed;
4676         }
4677
4678         rc = vio_enable_interrupts(vdev);
4679         if (rc) {
4680                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4681                 goto req_irq_failed;
4682         }
4683
4684         crq->cur = 0;
4685         spin_lock_init(&crq->lock);
4686
4687         return retrc;
4688
4689 req_irq_failed:
4690         tasklet_kill(&adapter->tasklet);
4691         do {
4692                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4693         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4694 reg_crq_failed:
4695         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4696 map_failed:
4697         free_page((unsigned long)crq->msgs);
4698         crq->msgs = NULL;
4699         return retrc;
4700 }
4701
4702 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4703 {
4704         struct device *dev = &adapter->vdev->dev;
4705         unsigned long timeout = msecs_to_jiffies(30000);
4706         u64 old_num_rx_queues, old_num_tx_queues;
4707         int rc;
4708
4709         adapter->from_passive_init = false;
4710
4711         old_num_rx_queues = adapter->req_rx_queues;
4712         old_num_tx_queues = adapter->req_tx_queues;
4713
4714         reinit_completion(&adapter->init_done);
4715         adapter->init_done_rc = 0;
4716         ibmvnic_send_crq_init(adapter);
4717         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4718                 dev_err(dev, "Initialization sequence timed out\n");
4719                 return -1;
4720         }
4721
4722         if (adapter->init_done_rc) {
4723                 release_crq_queue(adapter);
4724                 return adapter->init_done_rc;
4725         }
4726
4727         if (adapter->from_passive_init) {
4728                 adapter->state = VNIC_OPEN;
4729                 adapter->from_passive_init = false;
4730                 return -1;
4731         }
4732
4733         if (adapter->resetting && !adapter->wait_for_reset &&
4734             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4735                 if (adapter->req_rx_queues != old_num_rx_queues ||
4736                     adapter->req_tx_queues != old_num_tx_queues) {
4737                         release_sub_crqs(adapter, 0);
4738                         rc = init_sub_crqs(adapter);
4739                 } else {
4740                         rc = reset_sub_crq_queues(adapter);
4741                 }
4742         } else {
4743                 rc = init_sub_crqs(adapter);
4744         }
4745
4746         if (rc) {
4747                 dev_err(dev, "Initialization of sub crqs failed\n");
4748                 release_crq_queue(adapter);
4749                 return rc;
4750         }
4751
4752         rc = init_sub_crq_irqs(adapter);
4753         if (rc) {
4754                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4755                 release_crq_queue(adapter);
4756         }
4757
4758         return rc;
4759 }
4760
4761 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4762 {
4763         struct device *dev = &adapter->vdev->dev;
4764         unsigned long timeout = msecs_to_jiffies(30000);
4765         int rc;
4766
4767         adapter->from_passive_init = false;
4768
4769         adapter->init_done_rc = 0;
4770         ibmvnic_send_crq_init(adapter);
4771         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4772                 dev_err(dev, "Initialization sequence timed out\n");
4773                 return -1;
4774         }
4775
4776         if (adapter->init_done_rc) {
4777                 release_crq_queue(adapter);
4778                 return adapter->init_done_rc;
4779         }
4780
4781         if (adapter->from_passive_init) {
4782                 adapter->state = VNIC_OPEN;
4783                 adapter->from_passive_init = false;
4784                 return -1;
4785         }
4786
4787         rc = init_sub_crqs(adapter);
4788         if (rc) {
4789                 dev_err(dev, "Initialization of sub crqs failed\n");
4790                 release_crq_queue(adapter);
4791                 return rc;
4792         }
4793
4794         rc = init_sub_crq_irqs(adapter);
4795         if (rc) {
4796                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4797                 release_crq_queue(adapter);
4798         }
4799
4800         return rc;
4801 }
4802
4803 static struct device_attribute dev_attr_failover;
4804
4805 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4806 {
4807         struct ibmvnic_adapter *adapter;
4808         struct net_device *netdev;
4809         unsigned char *mac_addr_p;
4810         int rc;
4811
4812         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4813                 dev->unit_address);
4814
4815         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4816                                                         VETH_MAC_ADDR, NULL);
4817         if (!mac_addr_p) {
4818                 dev_err(&dev->dev,
4819                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4820                         __FILE__, __LINE__);
4821                 return 0;
4822         }
4823
4824         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4825                                    IBMVNIC_MAX_QUEUES);
4826         if (!netdev)
4827                 return -ENOMEM;
4828
4829         adapter = netdev_priv(netdev);
4830         adapter->state = VNIC_PROBING;
4831         dev_set_drvdata(&dev->dev, netdev);
4832         adapter->vdev = dev;
4833         adapter->netdev = netdev;
4834
4835         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4836         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4837         netdev->irq = dev->irq;
4838         netdev->netdev_ops = &ibmvnic_netdev_ops;
4839         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4840         SET_NETDEV_DEV(netdev, &dev->dev);
4841
4842         spin_lock_init(&adapter->stats_lock);
4843
4844         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4845         INIT_LIST_HEAD(&adapter->rwi_list);
4846         spin_lock_init(&adapter->rwi_lock);
4847         init_completion(&adapter->init_done);
4848         adapter->resetting = false;
4849
4850         do {
4851                 rc = init_crq_queue(adapter);
4852                 if (rc) {
4853                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4854                                 rc);
4855                         goto ibmvnic_init_fail;
4856                 }
4857
4858                 rc = ibmvnic_init(adapter);
4859                 if (rc && rc != EAGAIN)
4860                         goto ibmvnic_init_fail;
4861         } while (rc == EAGAIN);
4862
4863         rc = init_stats_buffers(adapter);
4864         if (rc)
4865                 goto ibmvnic_init_fail;
4866
4867         rc = init_stats_token(adapter);
4868         if (rc)
4869                 goto ibmvnic_stats_fail;
4870
4871         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4872         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4873         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4874
4875         rc = device_create_file(&dev->dev, &dev_attr_failover);
4876         if (rc)
4877                 goto ibmvnic_dev_file_err;
4878
4879         netif_carrier_off(netdev);
4880         rc = register_netdev(netdev);
4881         if (rc) {
4882                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4883                 goto ibmvnic_register_fail;
4884         }
4885         dev_info(&dev->dev, "ibmvnic registered\n");
4886
4887         adapter->state = VNIC_PROBED;
4888
4889         adapter->wait_for_reset = false;
4890
4891         return 0;
4892
4893 ibmvnic_register_fail:
4894         device_remove_file(&dev->dev, &dev_attr_failover);
4895
4896 ibmvnic_dev_file_err:
4897         release_stats_token(adapter);
4898
4899 ibmvnic_stats_fail:
4900         release_stats_buffers(adapter);
4901
4902 ibmvnic_init_fail:
4903         release_sub_crqs(adapter, 1);
4904         release_crq_queue(adapter);
4905         free_netdev(netdev);
4906
4907         return rc;
4908 }
4909
4910 static int ibmvnic_remove(struct vio_dev *dev)
4911 {
4912         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4913         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4914
4915         adapter->state = VNIC_REMOVING;
4916         rtnl_lock();
4917         unregister_netdevice(netdev);
4918
4919         release_resources(adapter);
4920         release_sub_crqs(adapter, 1);
4921         release_crq_queue(adapter);
4922
4923         release_stats_token(adapter);
4924         release_stats_buffers(adapter);
4925
4926         adapter->state = VNIC_REMOVED;
4927
4928         rtnl_unlock();
4929         device_remove_file(&dev->dev, &dev_attr_failover);
4930         free_netdev(netdev);
4931         dev_set_drvdata(&dev->dev, NULL);
4932
4933         return 0;
4934 }
4935
4936 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4937                               const char *buf, size_t count)
4938 {
4939         struct net_device *netdev = dev_get_drvdata(dev);
4940         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4941         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4942         __be64 session_token;
4943         long rc;
4944
4945         if (!sysfs_streq(buf, "1"))
4946                 return -EINVAL;
4947
4948         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4949                          H_GET_SESSION_TOKEN, 0, 0, 0);
4950         if (rc) {
4951                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4952                            rc);
4953                 return -EINVAL;
4954         }
4955
4956         session_token = (__be64)retbuf[0];
4957         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4958                    be64_to_cpu(session_token));
4959         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4960                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4961         if (rc) {
4962                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4963                            rc);
4964                 return -EINVAL;
4965         }
4966
4967         return count;
4968 }
4969
4970 static DEVICE_ATTR_WO(failover);
4971
4972 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4973 {
4974         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4975         struct ibmvnic_adapter *adapter;
4976         struct iommu_table *tbl;
4977         unsigned long ret = 0;
4978         int i;
4979
4980         tbl = get_iommu_table_base(&vdev->dev);
4981
4982         /* netdev inits at probe time along with the structures we need below*/
4983         if (!netdev)
4984                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4985
4986         adapter = netdev_priv(netdev);
4987
4988         ret += PAGE_SIZE; /* the crq message queue */
4989         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4990
4991         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4992                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4993
4994         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4995              i++)
4996                 ret += adapter->rx_pool[i].size *
4997                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4998
4999         return ret;
5000 }
5001
5002 static int ibmvnic_resume(struct device *dev)
5003 {
5004         struct net_device *netdev = dev_get_drvdata(dev);
5005         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5006
5007         if (adapter->state != VNIC_OPEN)
5008                 return 0;
5009
5010         tasklet_schedule(&adapter->tasklet);
5011
5012         return 0;
5013 }
5014
5015 static const struct vio_device_id ibmvnic_device_table[] = {
5016         {"network", "IBM,vnic"},
5017         {"", "" }
5018 };
5019 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5020
5021 static const struct dev_pm_ops ibmvnic_pm_ops = {
5022         .resume = ibmvnic_resume
5023 };
5024
5025 static struct vio_driver ibmvnic_driver = {
5026         .id_table       = ibmvnic_device_table,
5027         .probe          = ibmvnic_probe,
5028         .remove         = ibmvnic_remove,
5029         .get_desired_dma = ibmvnic_get_desired_dma,
5030         .name           = ibmvnic_driver_name,
5031         .pm             = &ibmvnic_pm_ops,
5032 };
5033
5034 /* module functions */
5035 static int __init ibmvnic_module_init(void)
5036 {
5037         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5038                 IBMVNIC_DRIVER_VERSION);
5039
5040         return vio_register_driver(&ibmvnic_driver);
5041 }
5042
5043 static void __exit ibmvnic_module_exit(void)
5044 {
5045         vio_unregister_driver(&ibmvnic_driver);
5046 }
5047
5048 module_init(ibmvnic_module_init);
5049 module_exit(ibmvnic_module_exit);