drivers/net/ethernet/ibm/ibmvnic.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /**************************************************************************/
   3 /*                                                                        */
   4 /*  IBM System i and System p Virtual NIC Device Driver                   */
   5 /*  Copyright (C) 2014 IBM Corp.                                          */
   6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   9 /*                                                                        */
  10 /*                                                                        */
  11 /* This module contains the implementation of a virtual ethernet device   */
  12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  15 /*                                                                         */
  16 /* Messages are passed between the VNIC driver and the VNIC server using  */
  17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  18 /* issue and receive commands that initiate communication with the server */
  19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  20 /* are used by the driver to notify the server that a packet is           */
  21 /* ready for transmission or that a buffer has been added to receive a    */
  22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  23 /* driver that a packet transmission has been completed or that a packet  */
  24 /* has been received and placed in a waiting buffer.                      */
  25 /*                                                                        */
  26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  28 /* or receive has been completed, the VNIC driver is required to use      */
  29 /* "long term mapping". This entails that large, continuous DMA mapped    */
  30 /* buffers are allocated on driver initialization and these buffers are   */
  31 /* then continuously reused to pass skbs to and from the VNIC server.     */
  32 /*                                                                        */
  33 /**************************************************************************/
  34
  35 #include <linux/module.h>
  36 #include <linux/moduleparam.h>
  37 #include <linux/types.h>
  38 #include <linux/errno.h>
  39 #include <linux/completion.h>
  40 #include <linux/ioport.h>
  41 #include <linux/dma-mapping.h>
  42 #include <linux/kernel.h>
  43 #include <linux/netdevice.h>
  44 #include <linux/etherdevice.h>
  45 #include <linux/skbuff.h>
  46 #include <linux/init.h>
  47 #include <linux/delay.h>
  48 #include <linux/mm.h>
  49 #include <linux/ethtool.h>
  50 #include <linux/proc_fs.h>
  51 #include <linux/if_arp.h>
  52 #include <linux/in.h>
  53 #include <linux/ip.h>
  54 #include <linux/ipv6.h>
  55 #include <linux/irq.h>
  56 #include <linux/kthread.h>
  57 #include <linux/seq_file.h>
  58 #include <linux/interrupt.h>
  59 #include <net/net_namespace.h>
  60 #include <asm/hvcall.h>
  61 #include <linux/atomic.h>
  62 #include <asm/vio.h>
  63 #include <asm/iommu.h>
  64 #include <linux/uaccess.h>
  65 #include <asm/firmware.h>
  66 #include <linux/workqueue.h>
  67 #include <linux/if_vlan.h>
  68 #include <linux/utsname.h>
  69
  70 #include "ibmvnic.h"
  71
  72 static const char ibmvnic_driver_name[] = "ibmvnic";
  73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  74
  75 MODULE_AUTHOR("Santiago Leon");
  76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  77 MODULE_LICENSE("GPL");
  78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  79
  80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  81 static int ibmvnic_remove(struct vio_dev *);
  82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  88                        union sub_crq *sub_crq);
  89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
  90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
  91 static int enable_scrq_irq(struct ibmvnic_adapter *,
  92                            struct ibmvnic_sub_crq_queue *);
  93 static int disable_scrq_irq(struct ibmvnic_adapter *,
  94                             struct ibmvnic_sub_crq_queue *);
  95 static int pending_scrq(struct ibmvnic_adapter *,
  96                         struct ibmvnic_sub_crq_queue *);
  97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
  98                                         struct ibmvnic_sub_crq_queue *);
  99 static int ibmvnic_poll(struct napi_struct *napi, int data);
 100 static void send_map_query(struct ibmvnic_adapter *adapter);
 101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
 103 static int send_login(struct ibmvnic_adapter *adapter);
 104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 105 static int init_sub_crqs(struct ibmvnic_adapter *);
 106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 107 static int ibmvnic_init(struct ibmvnic_adapter *);
 108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
 109 static void release_crq_queue(struct ibmvnic_adapter *);
 110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
 111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
 113
 114 struct ibmvnic_stat {
 115         char name[ETH_GSTRING_LEN];
 116         int offset;
 117 };
 118
 119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 120                              offsetof(struct ibmvnic_statistics, stat))
 121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 122
 123 static const struct ibmvnic_stat ibmvnic_stats[] = {
 124         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 125         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 126         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 127         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 128         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 129         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 130         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 131         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 132         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 133         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 134         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 135         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 136         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 137         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 138         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 139         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 140         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 141         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 142         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 143         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 144         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 145         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 146 };
 147
 148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 149                           unsigned long length, unsigned long *number,
 150                           unsigned long *irq)
 151 {
 152         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 153         long rc;
 154
 155         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 156         *number = retbuf[0];
 157         *irq = retbuf[1];
 158
 159         return rc;
 160 }
 161
 162 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 163                                 struct ibmvnic_long_term_buff *ltb, int size)
 164 {
 165         struct device *dev = &adapter->vdev->dev;
 166         int rc;
 167
 168         ltb->size = size;
 169         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 170                                        GFP_KERNEL);
 171
 172         if (!ltb->buff) {
 173                 dev_err(dev, "Couldn't alloc long term buffer\n");
 174                 return -ENOMEM;
 175         }
 176         ltb->map_id = adapter->map_id;
 177         adapter->map_id++;
 178
 179         init_completion(&adapter->fw_done);
 180         rc = send_request_map(adapter, ltb->addr,
 181                               ltb->size, ltb->map_id);
 182         if (rc) {
 183                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 184                 return rc;
 185         }
 186         wait_for_completion(&adapter->fw_done);
 187
 188         if (adapter->fw_done_rc) {
 189                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 190                         adapter->fw_done_rc);
 191                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 192                 return -1;
 193         }
 194         return 0;
 195 }
 196
 197 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 198                                 struct ibmvnic_long_term_buff *ltb)
 199 {
 200         struct device *dev = &adapter->vdev->dev;
 201
 202         if (!ltb->buff)
 203                 return;
 204
 205         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 206             adapter->reset_reason != VNIC_RESET_MOBILITY)
 207                 send_request_unmap(adapter, ltb->map_id);
 208         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 209 }
 210
 211 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 212                                 struct ibmvnic_long_term_buff *ltb)
 213 {
 214         int rc;
 215
 216         memset(ltb->buff, 0, ltb->size);
 217
 218         init_completion(&adapter->fw_done);
 219         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 220         if (rc)
 221                 return rc;
 222         wait_for_completion(&adapter->fw_done);
 223
 224         if (adapter->fw_done_rc) {
 225                 dev_info(&adapter->vdev->dev,
 226                          "Reset failed, attempting to free and reallocate buffer\n");
 227                 free_long_term_buff(adapter, ltb);
 228                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 229         }
 230         return 0;
 231 }
 232
 233 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 234 {
 235         int i;
 236
 237         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 238              i++)
 239                 adapter->rx_pool[i].active = 0;
 240 }
 241
 242 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 243                               struct ibmvnic_rx_pool *pool)
 244 {
 245         int count = pool->size - atomic_read(&pool->available);
 246         struct device *dev = &adapter->vdev->dev;
 247         int buffers_added = 0;
 248         unsigned long lpar_rc;
 249         union sub_crq sub_crq;
 250         struct sk_buff *skb;
 251         unsigned int offset;
 252         dma_addr_t dma_addr;
 253         unsigned char *dst;
 254         u64 *handle_array;
 255         int shift = 0;
 256         int index;
 257         int i;
 258
 259         if (!pool->active)
 260                 return;
 261
 262         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 263                                       be32_to_cpu(adapter->login_rsp_buf->
 264                                       off_rxadd_subcrqs));
 265
 266         for (i = 0; i < count; ++i) {
 267                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 268                 if (!skb) {
 269                         dev_err(dev, "Couldn't replenish rx buff\n");
 270                         adapter->replenish_no_mem++;
 271                         break;
 272                 }
 273
 274                 index = pool->free_map[pool->next_free];
 275
 276                 if (pool->rx_buff[index].skb)
 277                         dev_err(dev, "Inconsistent free_map!\n");
 278
 279                 /* Copy the skb to the long term mapped DMA buffer */
 280                 offset = index * pool->buff_size;
 281                 dst = pool->long_term_buff.buff + offset;
 282                 memset(dst, 0, pool->buff_size);
 283                 dma_addr = pool->long_term_buff.addr + offset;
 284                 pool->rx_buff[index].data = dst;
 285
 286                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 287                 pool->rx_buff[index].dma = dma_addr;
 288                 pool->rx_buff[index].skb = skb;
 289                 pool->rx_buff[index].pool_index = pool->index;
 290                 pool->rx_buff[index].size = pool->buff_size;
 291
 292                 memset(&sub_crq, 0, sizeof(sub_crq));
 293                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 294                 sub_crq.rx_add.correlator =
 295                     cpu_to_be64((u64)&pool->rx_buff[index]);
 296                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 297                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 298
 299                 /* The length field of the sCRQ is defined to be 24 bits so the
 300                  * buffer size needs to be left shifted by a byte before it is
 301                  * converted to big endian to prevent the last byte from being
 302                  * truncated.
 303                  */
 304 #ifdef __LITTLE_ENDIAN__
 305                 shift = 8;
 306 #endif
 307                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 308
 309                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 310                                       &sub_crq);
 311                 if (lpar_rc != H_SUCCESS)
 312                         goto failure;
 313
 314                 buffers_added++;
 315                 adapter->replenish_add_buff_success++;
 316                 pool->next_free = (pool->next_free + 1) % pool->size;
 317         }
 318         atomic_add(buffers_added, &pool->available);
 319         return;
 320
 321 failure:
 322         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 323                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 324         pool->free_map[pool->next_free] = index;
 325         pool->rx_buff[index].skb = NULL;
 326
 327         dev_kfree_skb_any(skb);
 328         adapter->replenish_add_buff_failure++;
 329         atomic_add(buffers_added, &pool->available);
 330
 331         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 332                 /* Disable buffer pool replenishment and report carrier off if
 333                  * queue is closed or pending failover.
 334                  * Firmware guarantees that a signal will be sent to the
 335                  * driver, triggering a reset.
 336                  */
 337                 deactivate_rx_pools(adapter);
 338                 netif_carrier_off(adapter->netdev);
 339         }
 340 }
 341
 342 static void replenish_pools(struct ibmvnic_adapter *adapter)
 343 {
 344         int i;
 345
 346         adapter->replenish_task_cycles++;
 347         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 348              i++) {
 349                 if (adapter->rx_pool[i].active)
 350                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 351         }
 352 }
 353
 354 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 355 {
 356         kfree(adapter->tx_stats_buffers);
 357         kfree(adapter->rx_stats_buffers);
 358         adapter->tx_stats_buffers = NULL;
 359         adapter->rx_stats_buffers = NULL;
 360 }
 361
 362 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 363 {
 364         adapter->tx_stats_buffers =
 365                                 kcalloc(IBMVNIC_MAX_QUEUES,
 366                                         sizeof(struct ibmvnic_tx_queue_stats),
 367                                         GFP_KERNEL);
 368         if (!adapter->tx_stats_buffers)
 369                 return -ENOMEM;
 370
 371         adapter->rx_stats_buffers =
 372                                 kcalloc(IBMVNIC_MAX_QUEUES,
 373                                         sizeof(struct ibmvnic_rx_queue_stats),
 374                                         GFP_KERNEL);
 375         if (!adapter->rx_stats_buffers)
 376                 return -ENOMEM;
 377
 378         return 0;
 379 }
 380
 381 static void release_stats_token(struct ibmvnic_adapter *adapter)
 382 {
 383         struct device *dev = &adapter->vdev->dev;
 384
 385         if (!adapter->stats_token)
 386                 return;
 387
 388         dma_unmap_single(dev, adapter->stats_token,
 389                          sizeof(struct ibmvnic_statistics),
 390                          DMA_FROM_DEVICE);
 391         adapter->stats_token = 0;
 392 }
 393
 394 static int init_stats_token(struct ibmvnic_adapter *adapter)
 395 {
 396         struct device *dev = &adapter->vdev->dev;
 397         dma_addr_t stok;
 398
 399         stok = dma_map_single(dev, &adapter->stats,
 400                               sizeof(struct ibmvnic_statistics),
 401                               DMA_FROM_DEVICE);
 402         if (dma_mapping_error(dev, stok)) {
 403                 dev_err(dev, "Couldn't map stats buffer\n");
 404                 return -1;
 405         }
 406
 407         adapter->stats_token = stok;
 408         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 409         return 0;
 410 }
 411
 412 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 413 {
 414         struct ibmvnic_rx_pool *rx_pool;
 415         int rx_scrqs;
 416         int i, j, rc;
 417         u64 *size_array;
 418
 419         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 420                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 421
 422         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 423         for (i = 0; i < rx_scrqs; i++) {
 424                 rx_pool = &adapter->rx_pool[i];
 425
 426                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 427
 428                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 429                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 430                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 431                         alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 432                                              rx_pool->size *
 433                                              rx_pool->buff_size);
 434                 } else {
 435                         rc = reset_long_term_buff(adapter,
 436                                                   &rx_pool->long_term_buff);
 437                 }
 438
 439                 if (rc)
 440                         return rc;
 441
 442                 for (j = 0; j < rx_pool->size; j++)
 443                         rx_pool->free_map[j] = j;
 444
 445                 memset(rx_pool->rx_buff, 0,
 446                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 447
 448                 atomic_set(&rx_pool->available, 0);
 449                 rx_pool->next_alloc = 0;
 450                 rx_pool->next_free = 0;
 451                 rx_pool->active = 1;
 452         }
 453
 454         return 0;
 455 }
 456
 457 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 458 {
 459         struct ibmvnic_rx_pool *rx_pool;
 460         int i, j;
 461
 462         if (!adapter->rx_pool)
 463                 return;
 464
 465         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 466                 rx_pool = &adapter->rx_pool[i];
 467
 468                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 469
 470                 kfree(rx_pool->free_map);
 471                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 472
 473                 if (!rx_pool->rx_buff)
 474                         continue;
 475
 476                 for (j = 0; j < rx_pool->size; j++) {
 477                         if (rx_pool->rx_buff[j].skb) {
 478                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 479                                 rx_pool->rx_buff[j].skb = NULL;
 480                         }
 481                 }
 482
 483                 kfree(rx_pool->rx_buff);
 484         }
 485
 486         kfree(adapter->rx_pool);
 487         adapter->rx_pool = NULL;
 488         adapter->num_active_rx_pools = 0;
 489 }
 490
 491 static int init_rx_pools(struct net_device *netdev)
 492 {
 493         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 494         struct device *dev = &adapter->vdev->dev;
 495         struct ibmvnic_rx_pool *rx_pool;
 496         int rxadd_subcrqs;
 497         u64 *size_array;
 498         int i, j;
 499
 500         rxadd_subcrqs =
 501                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 502         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 503                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 504
 505         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 506                                    sizeof(struct ibmvnic_rx_pool),
 507                                    GFP_KERNEL);
 508         if (!adapter->rx_pool) {
 509                 dev_err(dev, "Failed to allocate rx pools\n");
 510                 return -1;
 511         }
 512
 513         adapter->num_active_rx_pools = rxadd_subcrqs;
 514
 515         for (i = 0; i < rxadd_subcrqs; i++) {
 516                 rx_pool = &adapter->rx_pool[i];
 517
 518                 netdev_dbg(adapter->netdev,
 519                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 520                            i, adapter->req_rx_add_entries_per_subcrq,
 521                            be64_to_cpu(size_array[i]));
 522
 523                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 524                 rx_pool->index = i;
 525                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 526                 rx_pool->active = 1;
 527
 528                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 529                                             GFP_KERNEL);
 530                 if (!rx_pool->free_map) {
 531                         release_rx_pools(adapter);
 532                         return -1;
 533                 }
 534
 535                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 536                                            sizeof(struct ibmvnic_rx_buff),
 537                                            GFP_KERNEL);
 538                 if (!rx_pool->rx_buff) {
 539                         dev_err(dev, "Couldn't alloc rx buffers\n");
 540                         release_rx_pools(adapter);
 541                         return -1;
 542                 }
 543
 544                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 545                                          rx_pool->size * rx_pool->buff_size)) {
 546                         release_rx_pools(adapter);
 547                         return -1;
 548                 }
 549
 550                 for (j = 0; j < rx_pool->size; ++j)
 551                         rx_pool->free_map[j] = j;
 552
 553                 atomic_set(&rx_pool->available, 0);
 554                 rx_pool->next_alloc = 0;
 555                 rx_pool->next_free = 0;
 556         }
 557
 558         return 0;
 559 }
 560
 561 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 562                              struct ibmvnic_tx_pool *tx_pool)
 563 {
 564         int rc, i;
 565
 566         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 567         if (rc)
 568                 return rc;
 569
 570         memset(tx_pool->tx_buff, 0,
 571                tx_pool->num_buffers *
 572                sizeof(struct ibmvnic_tx_buff));
 573
 574         for (i = 0; i < tx_pool->num_buffers; i++)
 575                 tx_pool->free_map[i] = i;
 576
 577         tx_pool->consumer_index = 0;
 578         tx_pool->producer_index = 0;
 579
 580         return 0;
 581 }
 582
 583 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 584 {
 585         int tx_scrqs;
 586         int i, rc;
 587
 588         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 589         for (i = 0; i < tx_scrqs; i++) {
 590                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 591                 if (rc)
 592                         return rc;
 593                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 594                 if (rc)
 595                         return rc;
 596         }
 597
 598         return 0;
 599 }
 600
 601 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 602 {
 603         if (!adapter->vpd)
 604                 return;
 605
 606         kfree(adapter->vpd->buff);
 607         kfree(adapter->vpd);
 608
 609         adapter->vpd = NULL;
 610 }
 611
 612 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 613                                 struct ibmvnic_tx_pool *tx_pool)
 614 {
 615         kfree(tx_pool->tx_buff);
 616         kfree(tx_pool->free_map);
 617         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 618 }
 619
 620 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 621 {
 622         int i;
 623
 624         if (!adapter->tx_pool)
 625                 return;
 626
 627         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 628                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 629                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 630         }
 631
 632         kfree(adapter->tx_pool);
 633         adapter->tx_pool = NULL;
 634         kfree(adapter->tso_pool);
 635         adapter->tso_pool = NULL;
 636         adapter->num_active_tx_pools = 0;
 637 }
 638
 639 static int init_one_tx_pool(struct net_device *netdev,
 640                             struct ibmvnic_tx_pool *tx_pool,
 641                             int num_entries, int buf_size)
 642 {
 643         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 644         int i;
 645
 646         tx_pool->tx_buff = kcalloc(num_entries,
 647                                    sizeof(struct ibmvnic_tx_buff),
 648                                    GFP_KERNEL);
 649         if (!tx_pool->tx_buff)
 650                 return -1;
 651
 652         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 653                                  num_entries * buf_size))
 654                 return -1;
 655
 656         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 657         if (!tx_pool->free_map)
 658                 return -1;
 659
 660         for (i = 0; i < num_entries; i++)
 661                 tx_pool->free_map[i] = i;
 662
 663         tx_pool->consumer_index = 0;
 664         tx_pool->producer_index = 0;
 665         tx_pool->num_buffers = num_entries;
 666         tx_pool->buf_size = buf_size;
 667
 668         return 0;
 669 }
 670
 671 static int init_tx_pools(struct net_device *netdev)
 672 {
 673         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 674         int tx_subcrqs;
 675         int i, rc;
 676
 677         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 678         adapter->tx_pool = kcalloc(tx_subcrqs,
 679                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 680         if (!adapter->tx_pool)
 681                 return -1;
 682
 683         adapter->tso_pool = kcalloc(tx_subcrqs,
 684                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 685         if (!adapter->tso_pool)
 686                 return -1;
 687
 688         adapter->num_active_tx_pools = tx_subcrqs;
 689
 690         for (i = 0; i < tx_subcrqs; i++) {
 691                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 692                                       adapter->req_tx_entries_per_subcrq,
 693                                       adapter->req_mtu + VLAN_HLEN);
 694                 if (rc) {
 695                         release_tx_pools(adapter);
 696                         return rc;
 697                 }
 698
 699                 init_one_tx_pool(netdev, &adapter->tso_pool[i],
 700                                  IBMVNIC_TSO_BUFS,
 701                                  IBMVNIC_TSO_BUF_SZ);
 702                 if (rc) {
 703                         release_tx_pools(adapter);
 704                         return rc;
 705                 }
 706         }
 707
 708         return 0;
 709 }
 710
 711 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 712 {
 713         int i;
 714
 715         if (adapter->napi_enabled)
 716                 return;
 717
 718         for (i = 0; i < adapter->req_rx_queues; i++)
 719                 napi_enable(&adapter->napi[i]);
 720
 721         adapter->napi_enabled = true;
 722 }
 723
 724 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 725 {
 726         int i;
 727
 728         if (!adapter->napi_enabled)
 729                 return;
 730
 731         for (i = 0; i < adapter->req_rx_queues; i++) {
 732                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 733                 napi_disable(&adapter->napi[i]);
 734         }
 735
 736         adapter->napi_enabled = false;
 737 }
 738
 739 static int init_napi(struct ibmvnic_adapter *adapter)
 740 {
 741         int i;
 742
 743         adapter->napi = kcalloc(adapter->req_rx_queues,
 744                                 sizeof(struct napi_struct), GFP_KERNEL);
 745         if (!adapter->napi)
 746                 return -ENOMEM;
 747
 748         for (i = 0; i < adapter->req_rx_queues; i++) {
 749                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 750                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 751                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 752         }
 753
 754         adapter->num_active_rx_napi = adapter->req_rx_queues;
 755         return 0;
 756 }
 757
 758 static void release_napi(struct ibmvnic_adapter *adapter)
 759 {
 760         int i;
 761
 762         if (!adapter->napi)
 763                 return;
 764
 765         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 766                 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
 767                 netif_napi_del(&adapter->napi[i]);
 768         }
 769
 770         kfree(adapter->napi);
 771         adapter->napi = NULL;
 772         adapter->num_active_rx_napi = 0;
 773         adapter->napi_enabled = false;
 774 }
 775
 776 static int ibmvnic_login(struct net_device *netdev)
 777 {
 778         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 779         unsigned long timeout = msecs_to_jiffies(30000);
 780         int retry_count = 0;
 781         bool retry;
 782         int rc;
 783
 784         do {
 785                 retry = false;
 786                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 787                         netdev_warn(netdev, "Login attempts exceeded\n");
 788                         return -1;
 789                 }
 790
 791                 adapter->init_done_rc = 0;
 792                 reinit_completion(&adapter->init_done);
 793                 rc = send_login(adapter);
 794                 if (rc) {
 795                         netdev_warn(netdev, "Unable to login\n");
 796                         return rc;
 797                 }
 798
 799                 if (!wait_for_completion_timeout(&adapter->init_done,
 800                                                  timeout)) {
 801                         netdev_warn(netdev, "Login timed out\n");
 802                         return -1;
 803                 }
 804
 805                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 806                         retry_count++;
 807                         release_sub_crqs(adapter, 1);
 808
 809                         retry = true;
 810                         netdev_dbg(netdev,
 811                                    "Received partial success, retrying...\n");
 812                         adapter->init_done_rc = 0;
 813                         reinit_completion(&adapter->init_done);
 814                         send_cap_queries(adapter);
 815                         if (!wait_for_completion_timeout(&adapter->init_done,
 816                                                          timeout)) {
 817                                 netdev_warn(netdev,
 818                                             "Capabilities query timed out\n");
 819                                 return -1;
 820                         }
 821
 822                         rc = init_sub_crqs(adapter);
 823                         if (rc) {
 824                                 netdev_warn(netdev,
 825                                             "SCRQ initialization failed\n");
 826                                 return -1;
 827                         }
 828
 829                         rc = init_sub_crq_irqs(adapter);
 830                         if (rc) {
 831                                 netdev_warn(netdev,
 832                                             "SCRQ irq initialization failed\n");
 833                                 return -1;
 834                         }
 835                 } else if (adapter->init_done_rc) {
 836                         netdev_warn(netdev, "Adapter login failed\n");
 837                         return -1;
 838                 }
 839         } while (retry);
 840
 841         __ibmvnic_set_mac(netdev, adapter->mac_addr);
 842
 843         return 0;
 844 }
 845
 846 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 847 {
 848         kfree(adapter->login_buf);
 849         adapter->login_buf = NULL;
 850 }
 851
 852 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 853 {
 854         kfree(adapter->login_rsp_buf);
 855         adapter->login_rsp_buf = NULL;
 856 }
 857
 858 static void release_resources(struct ibmvnic_adapter *adapter)
 859 {
 860         release_vpd_data(adapter);
 861
 862         release_tx_pools(adapter);
 863         release_rx_pools(adapter);
 864
 865         release_napi(adapter);
 866         release_login_rsp_buffer(adapter);
 867 }
 868
 869 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 870 {
 871         struct net_device *netdev = adapter->netdev;
 872         unsigned long timeout = msecs_to_jiffies(30000);
 873         union ibmvnic_crq crq;
 874         bool resend;
 875         int rc;
 876
 877         netdev_dbg(netdev, "setting link state %d\n", link_state);
 878
 879         memset(&crq, 0, sizeof(crq));
 880         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 881         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 882         crq.logical_link_state.link_state = link_state;
 883
 884         do {
 885                 resend = false;
 886
 887                 reinit_completion(&adapter->init_done);
 888                 rc = ibmvnic_send_crq(adapter, &crq);
 889                 if (rc) {
 890                         netdev_err(netdev, "Failed to set link state\n");
 891                         return rc;
 892                 }
 893
 894                 if (!wait_for_completion_timeout(&adapter->init_done,
 895                                                  timeout)) {
 896                         netdev_err(netdev, "timeout setting link state\n");
 897                         return -1;
 898                 }
 899
 900                 if (adapter->init_done_rc == 1) {
 901                         /* Partuial success, delay and re-send */
 902                         mdelay(1000);
 903                         resend = true;
 904                 } else if (adapter->init_done_rc) {
 905                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 906                                     adapter->init_done_rc);
 907                         return adapter->init_done_rc;
 908                 }
 909         } while (resend);
 910
 911         return 0;
 912 }
 913
 914 static int set_real_num_queues(struct net_device *netdev)
 915 {
 916         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 917         int rc;
 918
 919         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 920                    adapter->req_tx_queues, adapter->req_rx_queues);
 921
 922         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 923         if (rc) {
 924                 netdev_err(netdev, "failed to set the number of tx queues\n");
 925                 return rc;
 926         }
 927
 928         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 929         if (rc)
 930                 netdev_err(netdev, "failed to set the number of rx queues\n");
 931
 932         return rc;
 933 }
 934
 935 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 936 {
 937         struct device *dev = &adapter->vdev->dev;
 938         union ibmvnic_crq crq;
 939         int len = 0;
 940         int rc;
 941
 942         if (adapter->vpd->buff)
 943                 len = adapter->vpd->len;
 944
 945         init_completion(&adapter->fw_done);
 946         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 947         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 948         rc = ibmvnic_send_crq(adapter, &crq);
 949         if (rc)
 950                 return rc;
 951         wait_for_completion(&adapter->fw_done);
 952
 953         if (!adapter->vpd->len)
 954                 return -ENODATA;
 955
 956         if (!adapter->vpd->buff)
 957                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 958         else if (adapter->vpd->len != len)
 959                 adapter->vpd->buff =
 960                         krealloc(adapter->vpd->buff,
 961                                  adapter->vpd->len, GFP_KERNEL);
 962
 963         if (!adapter->vpd->buff) {
 964                 dev_err(dev, "Could allocate VPD buffer\n");
 965                 return -ENOMEM;
 966         }
 967
 968         adapter->vpd->dma_addr =
 969                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 970                                DMA_FROM_DEVICE);
 971         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 972                 dev_err(dev, "Could not map VPD buffer\n");
 973                 kfree(adapter->vpd->buff);
 974                 adapter->vpd->buff = NULL;
 975                 return -ENOMEM;
 976         }
 977
 978         reinit_completion(&adapter->fw_done);
 979         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 980         crq.get_vpd.cmd = GET_VPD;
 981         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 982         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 983         rc = ibmvnic_send_crq(adapter, &crq);
 984         if (rc) {
 985                 kfree(adapter->vpd->buff);
 986                 adapter->vpd->buff = NULL;
 987                 return rc;
 988         }
 989         wait_for_completion(&adapter->fw_done);
 990
 991         return 0;
 992 }
 993
 994 static int init_resources(struct ibmvnic_adapter *adapter)
 995 {
 996         struct net_device *netdev = adapter->netdev;
 997         int rc;
 998
 999         rc = set_real_num_queues(netdev);
1000         if (rc)
1001                 return rc;
1002
1003         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1004         if (!adapter->vpd)
1005                 return -ENOMEM;
1006
1007         /* Vital Product Data (VPD) */
1008         rc = ibmvnic_get_vpd(adapter);
1009         if (rc) {
1010                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1011                 return rc;
1012         }
1013
1014         adapter->map_id = 1;
1015
1016         rc = init_napi(adapter);
1017         if (rc)
1018                 return rc;
1019
1020         send_map_query(adapter);
1021
1022         rc = init_rx_pools(netdev);
1023         if (rc)
1024                 return rc;
1025
1026         rc = init_tx_pools(netdev);
1027         return rc;
1028 }
1029
1030 static int __ibmvnic_open(struct net_device *netdev)
1031 {
1032         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1033         enum vnic_state prev_state = adapter->state;
1034         int i, rc;
1035
1036         adapter->state = VNIC_OPENING;
1037         replenish_pools(adapter);
1038         ibmvnic_napi_enable(adapter);
1039
1040         /* We're ready to receive frames, enable the sub-crq interrupts and
1041          * set the logical link state to up
1042          */
1043         for (i = 0; i < adapter->req_rx_queues; i++) {
1044                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1045                 if (prev_state == VNIC_CLOSED)
1046                         enable_irq(adapter->rx_scrq[i]->irq);
1047                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1048         }
1049
1050         for (i = 0; i < adapter->req_tx_queues; i++) {
1051                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1052                 if (prev_state == VNIC_CLOSED)
1053                         enable_irq(adapter->tx_scrq[i]->irq);
1054                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1055         }
1056
1057         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1058         if (rc) {
1059                 for (i = 0; i < adapter->req_rx_queues; i++)
1060                         napi_disable(&adapter->napi[i]);
1061                 release_resources(adapter);
1062                 return rc;
1063         }
1064
1065         netif_tx_start_all_queues(netdev);
1066
1067         if (prev_state == VNIC_CLOSED) {
1068                 for (i = 0; i < adapter->req_rx_queues; i++)
1069                         napi_schedule(&adapter->napi[i]);
1070         }
1071
1072         adapter->state = VNIC_OPEN;
1073         return rc;
1074 }
1075
1076 static int ibmvnic_open(struct net_device *netdev)
1077 {
1078         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1079         int rc;
1080
1081         /* If device failover is pending, just set device state and return.
1082          * Device operation will be handled by reset routine.
1083          */
1084         if (adapter->failover_pending) {
1085                 adapter->state = VNIC_OPEN;
1086                 return 0;
1087         }
1088
1089         if (adapter->state != VNIC_CLOSED) {
1090                 rc = ibmvnic_login(netdev);
1091                 if (rc)
1092                         return rc;
1093
1094                 rc = init_resources(adapter);
1095                 if (rc) {
1096                         netdev_err(netdev, "failed to initialize resources\n");
1097                         release_resources(adapter);
1098                         return rc;
1099                 }
1100         }
1101
1102         rc = __ibmvnic_open(netdev);
1103
1104         return rc;
1105 }
1106
1107 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1108 {
1109         struct ibmvnic_rx_pool *rx_pool;
1110         struct ibmvnic_rx_buff *rx_buff;
1111         u64 rx_entries;
1112         int rx_scrqs;
1113         int i, j;
1114
1115         if (!adapter->rx_pool)
1116                 return;
1117
1118         rx_scrqs = adapter->num_active_rx_pools;
1119         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1120
1121         /* Free any remaining skbs in the rx buffer pools */
1122         for (i = 0; i < rx_scrqs; i++) {
1123                 rx_pool = &adapter->rx_pool[i];
1124                 if (!rx_pool || !rx_pool->rx_buff)
1125                         continue;
1126
1127                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1128                 for (j = 0; j < rx_entries; j++) {
1129                         rx_buff = &rx_pool->rx_buff[j];
1130                         if (rx_buff && rx_buff->skb) {
1131                                 dev_kfree_skb_any(rx_buff->skb);
1132                                 rx_buff->skb = NULL;
1133                         }
1134                 }
1135         }
1136 }
1137
1138 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1139                               struct ibmvnic_tx_pool *tx_pool)
1140 {
1141         struct ibmvnic_tx_buff *tx_buff;
1142         u64 tx_entries;
1143         int i;
1144
1145         if (!tx_pool || !tx_pool->tx_buff)
1146                 return;
1147
1148         tx_entries = tx_pool->num_buffers;
1149
1150         for (i = 0; i < tx_entries; i++) {
1151                 tx_buff = &tx_pool->tx_buff[i];
1152                 if (tx_buff && tx_buff->skb) {
1153                         dev_kfree_skb_any(tx_buff->skb);
1154                         tx_buff->skb = NULL;
1155                 }
1156         }
1157 }
1158
1159 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1160 {
1161         int tx_scrqs;
1162         int i;
1163
1164         if (!adapter->tx_pool || !adapter->tso_pool)
1165                 return;
1166
1167         tx_scrqs = adapter->num_active_tx_pools;
1168
1169         /* Free any remaining skbs in the tx buffer pools */
1170         for (i = 0; i < tx_scrqs; i++) {
1171                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1172                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1173                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1174         }
1175 }
1176
1177 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1178 {
1179         struct net_device *netdev = adapter->netdev;
1180         int i;
1181
1182         if (adapter->tx_scrq) {
1183                 for (i = 0; i < adapter->req_tx_queues; i++)
1184                         if (adapter->tx_scrq[i]->irq) {
1185                                 netdev_dbg(netdev,
1186                                            "Disabling tx_scrq[%d] irq\n", i);
1187                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1188                                 disable_irq(adapter->tx_scrq[i]->irq);
1189                         }
1190         }
1191
1192         if (adapter->rx_scrq) {
1193                 for (i = 0; i < adapter->req_rx_queues; i++) {
1194                         if (adapter->rx_scrq[i]->irq) {
1195                                 netdev_dbg(netdev,
1196                                            "Disabling rx_scrq[%d] irq\n", i);
1197                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1198                                 disable_irq(adapter->rx_scrq[i]->irq);
1199                         }
1200                 }
1201         }
1202 }
1203
1204 static void ibmvnic_cleanup(struct net_device *netdev)
1205 {
1206         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1207
1208         /* ensure that transmissions are stopped if called by do_reset */
1209         if (adapter->resetting)
1210                 netif_tx_disable(netdev);
1211         else
1212                 netif_tx_stop_all_queues(netdev);
1213
1214         ibmvnic_napi_disable(adapter);
1215         ibmvnic_disable_irqs(adapter);
1216
1217         clean_rx_pools(adapter);
1218         clean_tx_pools(adapter);
1219 }
1220
1221 static int __ibmvnic_close(struct net_device *netdev)
1222 {
1223         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1224         int rc = 0;
1225
1226         adapter->state = VNIC_CLOSING;
1227         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1228         if (rc)
1229                 return rc;
1230         adapter->state = VNIC_CLOSED;
1231         return 0;
1232 }
1233
1234 static int ibmvnic_close(struct net_device *netdev)
1235 {
1236         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1237         int rc;
1238
1239         /* If device failover is pending, just set device state and return.
1240          * Device operation will be handled by reset routine.
1241          */
1242         if (adapter->failover_pending) {
1243                 adapter->state = VNIC_CLOSED;
1244                 return 0;
1245         }
1246
1247         rc = __ibmvnic_close(netdev);
1248         ibmvnic_cleanup(netdev);
1249
1250         return rc;
1251 }
1252
1253 /**
1254  * build_hdr_data - creates L2/L3/L4 header data buffer
1255  * @hdr_field - bitfield determining needed headers
1256  * @skb - socket buffer
1257  * @hdr_len - array of header lengths
1258  * @tot_len - total length of data
1259  *
1260  * Reads hdr_field to determine which headers are needed by firmware.
1261  * Builds a buffer containing these headers.  Saves individual header
1262  * lengths and total buffer length to be used to build descriptors.
1263  */
1264 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1265                           int *hdr_len, u8 *hdr_data)
1266 {
1267         int len = 0;
1268         u8 *hdr;
1269
1270         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1271                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1272         else
1273                 hdr_len[0] = sizeof(struct ethhdr);
1274
1275         if (skb->protocol == htons(ETH_P_IP)) {
1276                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1277                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1278                         hdr_len[2] = tcp_hdrlen(skb);
1279                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1280                         hdr_len[2] = sizeof(struct udphdr);
1281         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1282                 hdr_len[1] = sizeof(struct ipv6hdr);
1283                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1284                         hdr_len[2] = tcp_hdrlen(skb);
1285                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1286                         hdr_len[2] = sizeof(struct udphdr);
1287         } else if (skb->protocol == htons(ETH_P_ARP)) {
1288                 hdr_len[1] = arp_hdr_len(skb->dev);
1289                 hdr_len[2] = 0;
1290         }
1291
1292         memset(hdr_data, 0, 120);
1293         if ((hdr_field >> 6) & 1) {
1294                 hdr = skb_mac_header(skb);
1295                 memcpy(hdr_data, hdr, hdr_len[0]);
1296                 len += hdr_len[0];
1297         }
1298
1299         if ((hdr_field >> 5) & 1) {
1300                 hdr = skb_network_header(skb);
1301                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1302                 len += hdr_len[1];
1303         }
1304
1305         if ((hdr_field >> 4) & 1) {
1306                 hdr = skb_transport_header(skb);
1307                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1308                 len += hdr_len[2];
1309         }
1310         return len;
1311 }
1312
1313 /**
1314  * create_hdr_descs - create header and header extension descriptors
1315  * @hdr_field - bitfield determining needed headers
1316  * @data - buffer containing header data
1317  * @len - length of data buffer
1318  * @hdr_len - array of individual header lengths
1319  * @scrq_arr - descriptor array
1320  *
1321  * Creates header and, if needed, header extension descriptors and
1322  * places them in a descriptor array, scrq_arr
1323  */
1324
1325 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1326                             union sub_crq *scrq_arr)
1327 {
1328         union sub_crq hdr_desc;
1329         int tmp_len = len;
1330         int num_descs = 0;
1331         u8 *data, *cur;
1332         int tmp;
1333
1334         while (tmp_len > 0) {
1335                 cur = hdr_data + len - tmp_len;
1336
1337                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1338                 if (cur != hdr_data) {
1339                         data = hdr_desc.hdr_ext.data;
1340                         tmp = tmp_len > 29 ? 29 : tmp_len;
1341                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1342                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1343                         hdr_desc.hdr_ext.len = tmp;
1344                 } else {
1345                         data = hdr_desc.hdr.data;
1346                         tmp = tmp_len > 24 ? 24 : tmp_len;
1347                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1348                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1349                         hdr_desc.hdr.len = tmp;
1350                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1351                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1352                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1353                         hdr_desc.hdr.flag = hdr_field << 1;
1354                 }
1355                 memcpy(data, cur, tmp);
1356                 tmp_len -= tmp;
1357                 *scrq_arr = hdr_desc;
1358                 scrq_arr++;
1359                 num_descs++;
1360         }
1361
1362         return num_descs;
1363 }
1364
1365 /**
1366  * build_hdr_descs_arr - build a header descriptor array
1367  * @skb - socket buffer
1368  * @num_entries - number of descriptors to be sent
1369  * @subcrq - first TX descriptor
1370  * @hdr_field - bit field determining which headers will be sent
1371  *
1372  * This function will build a TX descriptor array with applicable
1373  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1374  */
1375
1376 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1377                                 int *num_entries, u8 hdr_field)
1378 {
1379         int hdr_len[3] = {0, 0, 0};
1380         int tot_len;
1381         u8 *hdr_data = txbuff->hdr_data;
1382
1383         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1384                                  txbuff->hdr_data);
1385         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1386                          txbuff->indir_arr + 1);
1387 }
1388
1389 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1390                                     struct net_device *netdev)
1391 {
1392         /* For some backing devices, mishandling of small packets
1393          * can result in a loss of connection or TX stall. Device
1394          * architects recommend that no packet should be smaller
1395          * than the minimum MTU value provided to the driver, so
1396          * pad any packets to that length
1397          */
1398         if (skb->len < netdev->min_mtu)
1399                 return skb_put_padto(skb, netdev->min_mtu);
1400
1401         return 0;
1402 }
1403
1404 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1405 {
1406         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1407         int queue_num = skb_get_queue_mapping(skb);
1408         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1409         struct device *dev = &adapter->vdev->dev;
1410         struct ibmvnic_tx_buff *tx_buff = NULL;
1411         struct ibmvnic_sub_crq_queue *tx_scrq;
1412         struct ibmvnic_tx_pool *tx_pool;
1413         unsigned int tx_send_failed = 0;
1414         unsigned int tx_map_failed = 0;
1415         unsigned int tx_dropped = 0;
1416         unsigned int tx_packets = 0;
1417         unsigned int tx_bytes = 0;
1418         dma_addr_t data_dma_addr;
1419         struct netdev_queue *txq;
1420         unsigned long lpar_rc;
1421         union sub_crq tx_crq;
1422         unsigned int offset;
1423         int num_entries = 1;
1424         unsigned char *dst;
1425         u64 *handle_array;
1426         int index = 0;
1427         u8 proto = 0;
1428         netdev_tx_t ret = NETDEV_TX_OK;
1429
1430         if (adapter->resetting) {
1431                 if (!netif_subqueue_stopped(netdev, skb))
1432                         netif_stop_subqueue(netdev, queue_num);
1433                 dev_kfree_skb_any(skb);
1434
1435                 tx_send_failed++;
1436                 tx_dropped++;
1437                 ret = NETDEV_TX_OK;
1438                 goto out;
1439         }
1440
1441         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1442                 tx_dropped++;
1443                 tx_send_failed++;
1444                 ret = NETDEV_TX_OK;
1445                 goto out;
1446         }
1447         if (skb_is_gso(skb))
1448                 tx_pool = &adapter->tso_pool[queue_num];
1449         else
1450                 tx_pool = &adapter->tx_pool[queue_num];
1451
1452         tx_scrq = adapter->tx_scrq[queue_num];
1453         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1454         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1455                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1456
1457         index = tx_pool->free_map[tx_pool->consumer_index];
1458
1459         if (index == IBMVNIC_INVALID_MAP) {
1460                 dev_kfree_skb_any(skb);
1461                 tx_send_failed++;
1462                 tx_dropped++;
1463                 ret = NETDEV_TX_OK;
1464                 goto out;
1465         }
1466
1467         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1468
1469         offset = index * tx_pool->buf_size;
1470         dst = tx_pool->long_term_buff.buff + offset;
1471         memset(dst, 0, tx_pool->buf_size);
1472         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1473
1474         if (skb_shinfo(skb)->nr_frags) {
1475                 int cur, i;
1476
1477                 /* Copy the head */
1478                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1479                 cur = skb_headlen(skb);
1480
1481                 /* Copy the frags */
1482                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1483                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1484
1485                         memcpy(dst + cur,
1486                                page_address(skb_frag_page(frag)) +
1487                                frag->page_offset, skb_frag_size(frag));
1488                         cur += skb_frag_size(frag);
1489                 }
1490         } else {
1491                 skb_copy_from_linear_data(skb, dst, skb->len);
1492         }
1493
1494         tx_pool->consumer_index =
1495             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1496
1497         tx_buff = &tx_pool->tx_buff[index];
1498         tx_buff->skb = skb;
1499         tx_buff->data_dma[0] = data_dma_addr;
1500         tx_buff->data_len[0] = skb->len;
1501         tx_buff->index = index;
1502         tx_buff->pool_index = queue_num;
1503         tx_buff->last_frag = true;
1504
1505         memset(&tx_crq, 0, sizeof(tx_crq));
1506         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1507         tx_crq.v1.type = IBMVNIC_TX_DESC;
1508         tx_crq.v1.n_crq_elem = 1;
1509         tx_crq.v1.n_sge = 1;
1510         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1511
1512         if (skb_is_gso(skb))
1513                 tx_crq.v1.correlator =
1514                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1515         else
1516                 tx_crq.v1.correlator = cpu_to_be32(index);
1517         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1518         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1519         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1520
1521         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1522                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1523                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1524         }
1525
1526         if (skb->protocol == htons(ETH_P_IP)) {
1527                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1528                 proto = ip_hdr(skb)->protocol;
1529         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1530                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1531                 proto = ipv6_hdr(skb)->nexthdr;
1532         }
1533
1534         if (proto == IPPROTO_TCP)
1535                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1536         else if (proto == IPPROTO_UDP)
1537                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1538
1539         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1540                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1541                 hdrs += 2;
1542         }
1543         if (skb_is_gso(skb)) {
1544                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1545                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1546                 hdrs += 2;
1547         }
1548         /* determine if l2/3/4 headers are sent to firmware */
1549         if ((*hdrs >> 7) & 1) {
1550                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1551                 tx_crq.v1.n_crq_elem = num_entries;
1552                 tx_buff->num_entries = num_entries;
1553                 tx_buff->indir_arr[0] = tx_crq;
1554                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1555                                                     sizeof(tx_buff->indir_arr),
1556                                                     DMA_TO_DEVICE);
1557                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1558                         dev_kfree_skb_any(skb);
1559                         tx_buff->skb = NULL;
1560                         if (!firmware_has_feature(FW_FEATURE_CMO))
1561                                 dev_err(dev, "tx: unable to map descriptor array\n");
1562                         tx_map_failed++;
1563                         tx_dropped++;
1564                         ret = NETDEV_TX_OK;
1565                         goto tx_err_out;
1566                 }
1567                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1568                                                (u64)tx_buff->indir_dma,
1569                                                (u64)num_entries);
1570         } else {
1571                 tx_buff->num_entries = num_entries;
1572                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1573                                       &tx_crq);
1574         }
1575         if (lpar_rc != H_SUCCESS) {
1576                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1577                         dev_err_ratelimited(dev, "tx: send failed\n");
1578                 dev_kfree_skb_any(skb);
1579                 tx_buff->skb = NULL;
1580
1581                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1582                         /* Disable TX and report carrier off if queue is closed
1583                          * or pending failover.
1584                          * Firmware guarantees that a signal will be sent to the
1585                          * driver, triggering a reset or some other action.
1586                          */
1587                         netif_tx_stop_all_queues(netdev);
1588                         netif_carrier_off(netdev);
1589                 }
1590
1591                 tx_send_failed++;
1592                 tx_dropped++;
1593                 ret = NETDEV_TX_OK;
1594                 goto tx_err_out;
1595         }
1596
1597         if (atomic_add_return(num_entries, &tx_scrq->used)
1598                                         >= adapter->req_tx_entries_per_subcrq) {
1599                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1600                 netif_stop_subqueue(netdev, queue_num);
1601         }
1602
1603         tx_packets++;
1604         tx_bytes += skb->len;
1605         txq->trans_start = jiffies;
1606         ret = NETDEV_TX_OK;
1607         goto out;
1608
1609 tx_err_out:
1610         /* roll back consumer index and map array*/
1611         if (tx_pool->consumer_index == 0)
1612                 tx_pool->consumer_index =
1613                         tx_pool->num_buffers - 1;
1614         else
1615                 tx_pool->consumer_index--;
1616         tx_pool->free_map[tx_pool->consumer_index] = index;
1617 out:
1618         netdev->stats.tx_dropped += tx_dropped;
1619         netdev->stats.tx_bytes += tx_bytes;
1620         netdev->stats.tx_packets += tx_packets;
1621         adapter->tx_send_failed += tx_send_failed;
1622         adapter->tx_map_failed += tx_map_failed;
1623         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1624         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1625         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1626
1627         return ret;
1628 }
1629
1630 static void ibmvnic_set_multi(struct net_device *netdev)
1631 {
1632         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1633         struct netdev_hw_addr *ha;
1634         union ibmvnic_crq crq;
1635
1636         memset(&crq, 0, sizeof(crq));
1637         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1638         crq.request_capability.cmd = REQUEST_CAPABILITY;
1639
1640         if (netdev->flags & IFF_PROMISC) {
1641                 if (!adapter->promisc_supported)
1642                         return;
1643         } else {
1644                 if (netdev->flags & IFF_ALLMULTI) {
1645                         /* Accept all multicast */
1646                         memset(&crq, 0, sizeof(crq));
1647                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1648                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1649                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1650                         ibmvnic_send_crq(adapter, &crq);
1651                 } else if (netdev_mc_empty(netdev)) {
1652                         /* Reject all multicast */
1653                         memset(&crq, 0, sizeof(crq));
1654                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1655                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1656                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1657                         ibmvnic_send_crq(adapter, &crq);
1658                 } else {
1659                         /* Accept one or more multicast(s) */
1660                         netdev_for_each_mc_addr(ha, netdev) {
1661                                 memset(&crq, 0, sizeof(crq));
1662                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1663                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1664                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1665                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1666                                                 ha->addr);
1667                                 ibmvnic_send_crq(adapter, &crq);
1668                         }
1669                 }
1670         }
1671 }
1672
1673 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1674 {
1675         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1676         union ibmvnic_crq crq;
1677         int rc;
1678
1679         if (!is_valid_ether_addr(dev_addr)) {
1680                 rc = -EADDRNOTAVAIL;
1681                 goto err;
1682         }
1683
1684         memset(&crq, 0, sizeof(crq));
1685         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1686         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1687         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1688
1689         init_completion(&adapter->fw_done);
1690         rc = ibmvnic_send_crq(adapter, &crq);
1691         if (rc) {
1692                 rc = -EIO;
1693                 goto err;
1694         }
1695
1696         wait_for_completion(&adapter->fw_done);
1697         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1698         if (adapter->fw_done_rc) {
1699                 rc = -EIO;
1700                 goto err;
1701         }
1702
1703         return 0;
1704 err:
1705         ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1706         return rc;
1707 }
1708
1709 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1710 {
1711         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1712         struct sockaddr *addr = p;
1713         int rc;
1714
1715         rc = 0;
1716         ether_addr_copy(adapter->mac_addr, addr->sa_data);
1717         if (adapter->state != VNIC_PROBED)
1718                 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1719
1720         return rc;
1721 }
1722
1723 /**
1724  * do_reset returns zero if we are able to keep processing reset events, or
1725  * non-zero if we hit a fatal error and must halt.
1726  */
1727 static int do_reset(struct ibmvnic_adapter *adapter,
1728                     struct ibmvnic_rwi *rwi, u32 reset_state)
1729 {
1730         u64 old_num_rx_queues, old_num_tx_queues;
1731         u64 old_num_rx_slots, old_num_tx_slots;
1732         struct net_device *netdev = adapter->netdev;
1733         int i, rc;
1734
1735         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1736                    rwi->reset_reason);
1737
1738         netif_carrier_off(netdev);
1739         adapter->reset_reason = rwi->reset_reason;
1740
1741         old_num_rx_queues = adapter->req_rx_queues;
1742         old_num_tx_queues = adapter->req_tx_queues;
1743         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1744         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1745
1746         ibmvnic_cleanup(netdev);
1747
1748         if (reset_state == VNIC_OPEN &&
1749             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1750             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1751                 rc = __ibmvnic_close(netdev);
1752                 if (rc)
1753                         return rc;
1754         }
1755
1756         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1757             adapter->wait_for_reset) {
1758                 release_resources(adapter);
1759                 release_sub_crqs(adapter, 1);
1760                 release_crq_queue(adapter);
1761         }
1762
1763         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1764                 /* remove the closed state so when we call open it appears
1765                  * we are coming from the probed state.
1766                  */
1767                 adapter->state = VNIC_PROBED;
1768
1769                 if (adapter->wait_for_reset) {
1770                         rc = init_crq_queue(adapter);
1771                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1772                         rc = ibmvnic_reenable_crq_queue(adapter);
1773                         release_sub_crqs(adapter, 1);
1774                 } else {
1775                         rc = ibmvnic_reset_crq(adapter);
1776                         if (!rc)
1777                                 rc = vio_enable_interrupts(adapter->vdev);
1778                 }
1779
1780                 if (rc) {
1781                         netdev_err(adapter->netdev,
1782                                    "Couldn't initialize crq. rc=%d\n", rc);
1783                         return rc;
1784                 }
1785
1786                 rc = ibmvnic_reset_init(adapter);
1787                 if (rc)
1788                         return IBMVNIC_INIT_FAILED;
1789
1790                 /* If the adapter was in PROBE state prior to the reset,
1791                  * exit here.
1792                  */
1793                 if (reset_state == VNIC_PROBED)
1794                         return 0;
1795
1796                 rc = ibmvnic_login(netdev);
1797                 if (rc) {
1798                         adapter->state = reset_state;
1799                         return rc;
1800                 }
1801
1802                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1803                     adapter->wait_for_reset) {
1804                         rc = init_resources(adapter);
1805                         if (rc)
1806                                 return rc;
1807                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1808                            adapter->req_tx_queues != old_num_tx_queues ||
1809                            adapter->req_rx_add_entries_per_subcrq !=
1810                                                         old_num_rx_slots ||
1811                            adapter->req_tx_entries_per_subcrq !=
1812                                                         old_num_tx_slots) {
1813                         release_rx_pools(adapter);
1814                         release_tx_pools(adapter);
1815                         release_napi(adapter);
1816                         release_vpd_data(adapter);
1817
1818                         rc = init_resources(adapter);
1819                         if (rc)
1820                                 return rc;
1821
1822                 } else {
1823                         rc = reset_tx_pools(adapter);
1824                         if (rc)
1825                                 return rc;
1826
1827                         rc = reset_rx_pools(adapter);
1828                         if (rc)
1829                                 return rc;
1830                 }
1831                 ibmvnic_disable_irqs(adapter);
1832         }
1833         adapter->state = VNIC_CLOSED;
1834
1835         if (reset_state == VNIC_CLOSED)
1836                 return 0;
1837
1838         rc = __ibmvnic_open(netdev);
1839         if (rc) {
1840                 if (list_empty(&adapter->rwi_list))
1841                         adapter->state = VNIC_CLOSED;
1842                 else
1843                         adapter->state = reset_state;
1844
1845                 return 0;
1846         }
1847
1848         /* refresh device's multicast list */
1849         ibmvnic_set_multi(netdev);
1850
1851         /* kick napi */
1852         for (i = 0; i < adapter->req_rx_queues; i++)
1853                 napi_schedule(&adapter->napi[i]);
1854
1855         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1856             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1857                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1858
1859         return 0;
1860 }
1861
1862 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1863                          struct ibmvnic_rwi *rwi, u32 reset_state)
1864 {
1865         struct net_device *netdev = adapter->netdev;
1866         int rc;
1867
1868         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1869                    rwi->reset_reason);
1870
1871         netif_carrier_off(netdev);
1872         adapter->reset_reason = rwi->reset_reason;
1873
1874         ibmvnic_cleanup(netdev);
1875         release_resources(adapter);
1876         release_sub_crqs(adapter, 0);
1877         release_crq_queue(adapter);
1878
1879         /* remove the closed state so when we call open it appears
1880          * we are coming from the probed state.
1881          */
1882         adapter->state = VNIC_PROBED;
1883
1884         reinit_completion(&adapter->init_done);
1885         rc = init_crq_queue(adapter);
1886         if (rc) {
1887                 netdev_err(adapter->netdev,
1888                            "Couldn't initialize crq. rc=%d\n", rc);
1889                 return rc;
1890         }
1891
1892         rc = ibmvnic_init(adapter);
1893         if (rc)
1894                 return rc;
1895
1896         /* If the adapter was in PROBE state prior to the reset,
1897          * exit here.
1898          */
1899         if (reset_state == VNIC_PROBED)
1900                 return 0;
1901
1902         rc = ibmvnic_login(netdev);
1903         if (rc) {
1904                 adapter->state = VNIC_PROBED;
1905                 return 0;
1906         }
1907
1908         rc = init_resources(adapter);
1909         if (rc)
1910                 return rc;
1911
1912         ibmvnic_disable_irqs(adapter);
1913         adapter->state = VNIC_CLOSED;
1914
1915         if (reset_state == VNIC_CLOSED)
1916                 return 0;
1917
1918         rc = __ibmvnic_open(netdev);
1919         if (rc) {
1920                 if (list_empty(&adapter->rwi_list))
1921                         adapter->state = VNIC_CLOSED;
1922                 else
1923                         adapter->state = reset_state;
1924
1925                 return 0;
1926         }
1927
1928         return 0;
1929 }
1930
1931 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1932 {
1933         struct ibmvnic_rwi *rwi;
1934         unsigned long flags;
1935
1936         spin_lock_irqsave(&adapter->rwi_lock, flags);
1937
1938         if (!list_empty(&adapter->rwi_list)) {
1939                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1940                                        list);
1941                 list_del(&rwi->list);
1942         } else {
1943                 rwi = NULL;
1944         }
1945
1946         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1947         return rwi;
1948 }
1949
1950 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1951 {
1952         struct ibmvnic_rwi *rwi;
1953
1954         rwi = get_next_rwi(adapter);
1955         while (rwi) {
1956                 kfree(rwi);
1957                 rwi = get_next_rwi(adapter);
1958         }
1959 }
1960
1961 static void __ibmvnic_reset(struct work_struct *work)
1962 {
1963         struct ibmvnic_rwi *rwi;
1964         struct ibmvnic_adapter *adapter;
1965         bool we_lock_rtnl = false;
1966         u32 reset_state;
1967         int rc = 0;
1968
1969         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1970
1971         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1972          * unless wait_for_reset is set, in which case the rtnl lock
1973          * has already been taken before initializing the reset
1974          */
1975         if (!adapter->wait_for_reset) {
1976                 rtnl_lock();
1977                 we_lock_rtnl = true;
1978         }
1979         reset_state = adapter->state;
1980
1981         rwi = get_next_rwi(adapter);
1982         while (rwi) {
1983                 if (adapter->force_reset_recovery) {
1984                         adapter->force_reset_recovery = false;
1985                         rc = do_hard_reset(adapter, rwi, reset_state);
1986                 } else {
1987                         rc = do_reset(adapter, rwi, reset_state);
1988                 }
1989                 kfree(rwi);
1990                 if (rc && rc != IBMVNIC_INIT_FAILED &&
1991                     !adapter->force_reset_recovery)
1992                         break;
1993
1994                 rwi = get_next_rwi(adapter);
1995         }
1996
1997         if (adapter->wait_for_reset) {
1998                 adapter->wait_for_reset = false;
1999                 adapter->reset_done_rc = rc;
2000                 complete(&adapter->reset_done);
2001         }
2002
2003         if (rc) {
2004                 netdev_dbg(adapter->netdev, "Reset failed\n");
2005                 free_all_rwi(adapter);
2006         }
2007
2008         adapter->resetting = false;
2009         if (we_lock_rtnl)
2010                 rtnl_unlock();
2011 }
2012
2013 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2014                          enum ibmvnic_reset_reason reason)
2015 {
2016         struct list_head *entry, *tmp_entry;
2017         struct ibmvnic_rwi *rwi, *tmp;
2018         struct net_device *netdev = adapter->netdev;
2019         unsigned long flags;
2020         int ret;
2021
2022         if (adapter->state == VNIC_REMOVING ||
2023             adapter->state == VNIC_REMOVED ||
2024             adapter->failover_pending) {
2025                 ret = EBUSY;
2026                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2027                 goto err;
2028         }
2029
2030         if (adapter->state == VNIC_PROBING) {
2031                 netdev_warn(netdev, "Adapter reset during probe\n");
2032                 ret = adapter->init_done_rc = EAGAIN;
2033                 goto err;
2034         }
2035
2036         spin_lock_irqsave(&adapter->rwi_lock, flags);
2037
2038         list_for_each(entry, &adapter->rwi_list) {
2039                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2040                 if (tmp->reset_reason == reason) {
2041                         netdev_dbg(netdev, "Skipping matching reset\n");
2042                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2043                         ret = EBUSY;
2044                         goto err;
2045                 }
2046         }
2047
2048         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2049         if (!rwi) {
2050                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2051                 ibmvnic_close(netdev);
2052                 ret = ENOMEM;
2053                 goto err;
2054         }
2055         /* if we just received a transport event,
2056          * flush reset queue and process this reset
2057          */
2058         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2059                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2060                         list_del(entry);
2061         }
2062         rwi->reset_reason = reason;
2063         list_add_tail(&rwi->list, &adapter->rwi_list);
2064         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2065         adapter->resetting = true;
2066         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2067         schedule_work(&adapter->ibmvnic_reset);
2068
2069         return 0;
2070 err:
2071         if (adapter->wait_for_reset)
2072                 adapter->wait_for_reset = false;
2073         return -ret;
2074 }
2075
2076 static void ibmvnic_tx_timeout(struct net_device *dev)
2077 {
2078         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2079
2080         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2081 }
2082
2083 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2084                                   struct ibmvnic_rx_buff *rx_buff)
2085 {
2086         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2087
2088         rx_buff->skb = NULL;
2089
2090         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2091         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2092
2093         atomic_dec(&pool->available);
2094 }
2095
2096 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2097 {
2098         struct net_device *netdev = napi->dev;
2099         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2100         int scrq_num = (int)(napi - adapter->napi);
2101         int frames_processed = 0;
2102
2103 restart_poll:
2104         while (frames_processed < budget) {
2105                 struct sk_buff *skb;
2106                 struct ibmvnic_rx_buff *rx_buff;
2107                 union sub_crq *next;
2108                 u32 length;
2109                 u16 offset;
2110                 u8 flags = 0;
2111
2112                 if (unlikely(adapter->resetting &&
2113                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2114                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2115                         napi_complete_done(napi, frames_processed);
2116                         return frames_processed;
2117                 }
2118
2119                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2120                         break;
2121                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2122                 rx_buff =
2123                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2124                                                           rx_comp.correlator);
2125                 /* do error checking */
2126                 if (next->rx_comp.rc) {
2127                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2128                                    be16_to_cpu(next->rx_comp.rc));
2129                         /* free the entry */
2130                         next->rx_comp.first = 0;
2131                         dev_kfree_skb_any(rx_buff->skb);
2132                         remove_buff_from_pool(adapter, rx_buff);
2133                         continue;
2134                 } else if (!rx_buff->skb) {
2135                         /* free the entry */
2136                         next->rx_comp.first = 0;
2137                         remove_buff_from_pool(adapter, rx_buff);
2138                         continue;
2139                 }
2140
2141                 length = be32_to_cpu(next->rx_comp.len);
2142                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2143                 flags = next->rx_comp.flags;
2144                 skb = rx_buff->skb;
2145                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2146                                         length);
2147
2148                 /* VLAN Header has been stripped by the system firmware and
2149                  * needs to be inserted by the driver
2150                  */
2151                 if (adapter->rx_vlan_header_insertion &&
2152                     (flags & IBMVNIC_VLAN_STRIPPED))
2153                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2154                                                ntohs(next->rx_comp.vlan_tci));
2155
2156                 /* free the entry */
2157                 next->rx_comp.first = 0;
2158                 remove_buff_from_pool(adapter, rx_buff);
2159
2160                 skb_put(skb, length);
2161                 skb->protocol = eth_type_trans(skb, netdev);
2162                 skb_record_rx_queue(skb, scrq_num);
2163
2164                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2165                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2166                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2167                 }
2168
2169                 length = skb->len;
2170                 napi_gro_receive(napi, skb); /* send it up */
2171                 netdev->stats.rx_packets++;
2172                 netdev->stats.rx_bytes += length;
2173                 adapter->rx_stats_buffers[scrq_num].packets++;
2174                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2175                 frames_processed++;
2176         }
2177
2178         if (adapter->state != VNIC_CLOSING)
2179                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2180
2181         if (frames_processed < budget) {
2182                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2183                 napi_complete_done(napi, frames_processed);
2184                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2185                     napi_reschedule(napi)) {
2186                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2187                         goto restart_poll;
2188                 }
2189         }
2190         return frames_processed;
2191 }
2192
2193 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2194 {
2195         int rc, ret;
2196
2197         adapter->fallback.mtu = adapter->req_mtu;
2198         adapter->fallback.rx_queues = adapter->req_rx_queues;
2199         adapter->fallback.tx_queues = adapter->req_tx_queues;
2200         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2201         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2202
2203         init_completion(&adapter->reset_done);
2204         adapter->wait_for_reset = true;
2205         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2206         if (rc)
2207                 return rc;
2208         wait_for_completion(&adapter->reset_done);
2209
2210         ret = 0;
2211         if (adapter->reset_done_rc) {
2212                 ret = -EIO;
2213                 adapter->desired.mtu = adapter->fallback.mtu;
2214                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2215                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2216                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2217                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2218
2219                 init_completion(&adapter->reset_done);
2220                 adapter->wait_for_reset = true;
2221                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2222                 if (rc)
2223                         return ret;
2224                 wait_for_completion(&adapter->reset_done);
2225         }
2226         adapter->wait_for_reset = false;
2227
2228         return ret;
2229 }
2230
2231 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2232 {
2233         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2234
2235         adapter->desired.mtu = new_mtu + ETH_HLEN;
2236
2237         return wait_for_reset(adapter);
2238 }
2239
2240 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2241                                                 struct net_device *dev,
2242                                                 netdev_features_t features)
2243 {
2244         /* Some backing hardware adapters can not
2245          * handle packets with a MSS less than 224
2246          * or with only one segment.
2247          */
2248         if (skb_is_gso(skb)) {
2249                 if (skb_shinfo(skb)->gso_size < 224 ||
2250                     skb_shinfo(skb)->gso_segs == 1)
2251                         features &= ~NETIF_F_GSO_MASK;
2252         }
2253
2254         return features;
2255 }
2256
2257 static const struct net_device_ops ibmvnic_netdev_ops = {
2258         .ndo_open               = ibmvnic_open,
2259         .ndo_stop               = ibmvnic_close,
2260         .ndo_start_xmit         = ibmvnic_xmit,
2261         .ndo_set_rx_mode        = ibmvnic_set_multi,
2262         .ndo_set_mac_address    = ibmvnic_set_mac,
2263         .ndo_validate_addr      = eth_validate_addr,
2264         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2265         .ndo_change_mtu         = ibmvnic_change_mtu,
2266         .ndo_features_check     = ibmvnic_features_check,
2267 };
2268
2269 /* ethtool functions */
2270
2271 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2272                                       struct ethtool_link_ksettings *cmd)
2273 {
2274         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2275         int rc;
2276
2277         rc = send_query_phys_parms(adapter);
2278         if (rc) {
2279                 adapter->speed = SPEED_UNKNOWN;
2280                 adapter->duplex = DUPLEX_UNKNOWN;
2281         }
2282         cmd->base.speed = adapter->speed;
2283         cmd->base.duplex = adapter->duplex;
2284         cmd->base.port = PORT_FIBRE;
2285         cmd->base.phy_address = 0;
2286         cmd->base.autoneg = AUTONEG_ENABLE;
2287
2288         return 0;
2289 }
2290
2291 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2292                                 struct ethtool_drvinfo *info)
2293 {
2294         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2295
2296         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2297         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2298         strlcpy(info->fw_version, adapter->fw_version,
2299                 sizeof(info->fw_version));
2300 }
2301
2302 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2303 {
2304         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2305
2306         return adapter->msg_enable;
2307 }
2308
2309 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2310 {
2311         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2312
2313         adapter->msg_enable = data;
2314 }
2315
2316 static u32 ibmvnic_get_link(struct net_device *netdev)
2317 {
2318         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2319
2320         /* Don't need to send a query because we request a logical link up at
2321          * init and then we wait for link state indications
2322          */
2323         return adapter->logical_link_state;
2324 }
2325
2326 static void ibmvnic_get_ringparam(struct net_device *netdev,
2327                                   struct ethtool_ringparam *ring)
2328 {
2329         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2330
2331         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2332                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2333                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2334         } else {
2335                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2336                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2337         }
2338         ring->rx_mini_max_pending = 0;
2339         ring->rx_jumbo_max_pending = 0;
2340         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2341         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2342         ring->rx_mini_pending = 0;
2343         ring->rx_jumbo_pending = 0;
2344 }
2345
2346 static int ibmvnic_set_ringparam(struct net_device *netdev,
2347                                  struct ethtool_ringparam *ring)
2348 {
2349         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2350         int ret;
2351
2352         ret = 0;
2353         adapter->desired.rx_entries = ring->rx_pending;
2354         adapter->desired.tx_entries = ring->tx_pending;
2355
2356         ret = wait_for_reset(adapter);
2357
2358         if (!ret &&
2359             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2360              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2361                 netdev_info(netdev,
2362                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2363                             ring->rx_pending, ring->tx_pending,
2364                             adapter->req_rx_add_entries_per_subcrq,
2365                             adapter->req_tx_entries_per_subcrq);
2366         return ret;
2367 }
2368
2369 static void ibmvnic_get_channels(struct net_device *netdev,
2370                                  struct ethtool_channels *channels)
2371 {
2372         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2373
2374         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2375                 channels->max_rx = adapter->max_rx_queues;
2376                 channels->max_tx = adapter->max_tx_queues;
2377         } else {
2378                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2379                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2380         }
2381
2382         channels->max_other = 0;
2383         channels->max_combined = 0;
2384         channels->rx_count = adapter->req_rx_queues;
2385         channels->tx_count = adapter->req_tx_queues;
2386         channels->other_count = 0;
2387         channels->combined_count = 0;
2388 }
2389
2390 static int ibmvnic_set_channels(struct net_device *netdev,
2391                                 struct ethtool_channels *channels)
2392 {
2393         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2394         int ret;
2395
2396         ret = 0;
2397         adapter->desired.rx_queues = channels->rx_count;
2398         adapter->desired.tx_queues = channels->tx_count;
2399
2400         ret = wait_for_reset(adapter);
2401
2402         if (!ret &&
2403             (adapter->req_rx_queues != channels->rx_count ||
2404              adapter->req_tx_queues != channels->tx_count))
2405                 netdev_info(netdev,
2406                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2407                             channels->rx_count, channels->tx_count,
2408                             adapter->req_rx_queues, adapter->req_tx_queues);
2409         return ret;
2410
2411 }
2412
2413 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2414 {
2415         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2416         int i;
2417
2418         switch (stringset) {
2419         case ETH_SS_STATS:
2420                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2421                                 i++, data += ETH_GSTRING_LEN)
2422                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2423
2424                 for (i = 0; i < adapter->req_tx_queues; i++) {
2425                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2426                         data += ETH_GSTRING_LEN;
2427
2428                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2429                         data += ETH_GSTRING_LEN;
2430
2431                         snprintf(data, ETH_GSTRING_LEN,
2432                                  "tx%d_dropped_packets", i);
2433                         data += ETH_GSTRING_LEN;
2434                 }
2435
2436                 for (i = 0; i < adapter->req_rx_queues; i++) {
2437                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2438                         data += ETH_GSTRING_LEN;
2439
2440                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2441                         data += ETH_GSTRING_LEN;
2442
2443                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2444                         data += ETH_GSTRING_LEN;
2445                 }
2446                 break;
2447
2448         case ETH_SS_PRIV_FLAGS:
2449                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2450                         strcpy(data + i * ETH_GSTRING_LEN,
2451                                ibmvnic_priv_flags[i]);
2452                 break;
2453         default:
2454                 return;
2455         }
2456 }
2457
2458 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2459 {
2460         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2461
2462         switch (sset) {
2463         case ETH_SS_STATS:
2464                 return ARRAY_SIZE(ibmvnic_stats) +
2465                        adapter->req_tx_queues * NUM_TX_STATS +
2466                        adapter->req_rx_queues * NUM_RX_STATS;
2467         case ETH_SS_PRIV_FLAGS:
2468                 return ARRAY_SIZE(ibmvnic_priv_flags);
2469         default:
2470                 return -EOPNOTSUPP;
2471         }
2472 }
2473
2474 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2475                                       struct ethtool_stats *stats, u64 *data)
2476 {
2477         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2478         union ibmvnic_crq crq;
2479         int i, j;
2480         int rc;
2481
2482         memset(&crq, 0, sizeof(crq));
2483         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2484         crq.request_statistics.cmd = REQUEST_STATISTICS;
2485         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2486         crq.request_statistics.len =
2487             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2488
2489         /* Wait for data to be written */
2490         init_completion(&adapter->stats_done);
2491         rc = ibmvnic_send_crq(adapter, &crq);
2492         if (rc)
2493                 return;
2494         wait_for_completion(&adapter->stats_done);
2495
2496         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2497                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2498                                                 ibmvnic_stats[i].offset));
2499
2500         for (j = 0; j < adapter->req_tx_queues; j++) {
2501                 data[i] = adapter->tx_stats_buffers[j].packets;
2502                 i++;
2503                 data[i] = adapter->tx_stats_buffers[j].bytes;
2504                 i++;
2505                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2506                 i++;
2507         }
2508
2509         for (j = 0; j < adapter->req_rx_queues; j++) {
2510                 data[i] = adapter->rx_stats_buffers[j].packets;
2511                 i++;
2512                 data[i] = adapter->rx_stats_buffers[j].bytes;
2513                 i++;
2514                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2515                 i++;
2516         }
2517 }
2518
2519 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2520 {
2521         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2522
2523         return adapter->priv_flags;
2524 }
2525
2526 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2527 {
2528         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2529         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2530
2531         if (which_maxes)
2532                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2533         else
2534                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2535
2536         return 0;
2537 }
2538 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2539         .get_drvinfo            = ibmvnic_get_drvinfo,
2540         .get_msglevel           = ibmvnic_get_msglevel,
2541         .set_msglevel           = ibmvnic_set_msglevel,
2542         .get_link               = ibmvnic_get_link,
2543         .get_ringparam          = ibmvnic_get_ringparam,
2544         .set_ringparam          = ibmvnic_set_ringparam,
2545         .get_channels           = ibmvnic_get_channels,
2546         .set_channels           = ibmvnic_set_channels,
2547         .get_strings            = ibmvnic_get_strings,
2548         .get_sset_count         = ibmvnic_get_sset_count,
2549         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2550         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2551         .get_priv_flags         = ibmvnic_get_priv_flags,
2552         .set_priv_flags         = ibmvnic_set_priv_flags,
2553 };
2554
2555 /* Routines for managing CRQs/sCRQs  */
2556
2557 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2558                                    struct ibmvnic_sub_crq_queue *scrq)
2559 {
2560         int rc;
2561
2562         if (scrq->irq) {
2563                 free_irq(scrq->irq, scrq);
2564                 irq_dispose_mapping(scrq->irq);
2565                 scrq->irq = 0;
2566         }
2567
2568         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2569         atomic_set(&scrq->used, 0);
2570         scrq->cur = 0;
2571
2572         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2573                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2574         return rc;
2575 }
2576
2577 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2578 {
2579         int i, rc;
2580
2581         for (i = 0; i < adapter->req_tx_queues; i++) {
2582                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2583                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2584                 if (rc)
2585                         return rc;
2586         }
2587
2588         for (i = 0; i < adapter->req_rx_queues; i++) {
2589                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2590                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2591                 if (rc)
2592                         return rc;
2593         }
2594
2595         return rc;
2596 }
2597
2598 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2599                                   struct ibmvnic_sub_crq_queue *scrq,
2600                                   bool do_h_free)
2601 {
2602         struct device *dev = &adapter->vdev->dev;
2603         long rc;
2604
2605         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2606
2607         if (do_h_free) {
2608                 /* Close the sub-crqs */
2609                 do {
2610                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2611                                                 adapter->vdev->unit_address,
2612                                                 scrq->crq_num);
2613                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2614
2615                 if (rc) {
2616                         netdev_err(adapter->netdev,
2617                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2618                                    scrq->crq_num, rc);
2619                 }
2620         }
2621
2622         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2623                          DMA_BIDIRECTIONAL);
2624         free_pages((unsigned long)scrq->msgs, 2);
2625         kfree(scrq);
2626 }
2627
2628 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2629                                                         *adapter)
2630 {
2631         struct device *dev = &adapter->vdev->dev;
2632         struct ibmvnic_sub_crq_queue *scrq;
2633         int rc;
2634
2635         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2636         if (!scrq)
2637                 return NULL;
2638
2639         scrq->msgs =
2640                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2641         if (!scrq->msgs) {
2642                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2643                 goto zero_page_failed;
2644         }
2645
2646         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2647                                          DMA_BIDIRECTIONAL);
2648         if (dma_mapping_error(dev, scrq->msg_token)) {
2649                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2650                 goto map_failed;
2651         }
2652
2653         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2654                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2655
2656         if (rc == H_RESOURCE)
2657                 rc = ibmvnic_reset_crq(adapter);
2658
2659         if (rc == H_CLOSED) {
2660                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2661         } else if (rc) {
2662                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2663                 goto reg_failed;
2664         }
2665
2666         scrq->adapter = adapter;
2667         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2668         spin_lock_init(&scrq->lock);
2669
2670         netdev_dbg(adapter->netdev,
2671                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2672                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2673
2674         return scrq;
2675
2676 reg_failed:
2677         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2678                          DMA_BIDIRECTIONAL);
2679 map_failed:
2680         free_pages((unsigned long)scrq->msgs, 2);
2681 zero_page_failed:
2682         kfree(scrq);
2683
2684         return NULL;
2685 }
2686
2687 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2688 {
2689         int i;
2690
2691         if (adapter->tx_scrq) {
2692                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2693                         if (!adapter->tx_scrq[i])
2694                                 continue;
2695
2696                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2697                                    i);
2698                         if (adapter->tx_scrq[i]->irq) {
2699                                 free_irq(adapter->tx_scrq[i]->irq,
2700                                          adapter->tx_scrq[i]);
2701                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2702                                 adapter->tx_scrq[i]->irq = 0;
2703                         }
2704
2705                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2706                                               do_h_free);
2707                 }
2708
2709                 kfree(adapter->tx_scrq);
2710                 adapter->tx_scrq = NULL;
2711                 adapter->num_active_tx_scrqs = 0;
2712         }
2713
2714         if (adapter->rx_scrq) {
2715                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2716                         if (!adapter->rx_scrq[i])
2717                                 continue;
2718
2719                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2720                                    i);
2721                         if (adapter->rx_scrq[i]->irq) {
2722                                 free_irq(adapter->rx_scrq[i]->irq,
2723                                          adapter->rx_scrq[i]);
2724                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2725                                 adapter->rx_scrq[i]->irq = 0;
2726                         }
2727
2728                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2729                                               do_h_free);
2730                 }
2731
2732                 kfree(adapter->rx_scrq);
2733                 adapter->rx_scrq = NULL;
2734                 adapter->num_active_rx_scrqs = 0;
2735         }
2736 }
2737
2738 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2739                             struct ibmvnic_sub_crq_queue *scrq)
2740 {
2741         struct device *dev = &adapter->vdev->dev;
2742         unsigned long rc;
2743
2744         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2745                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2746         if (rc)
2747                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2748                         scrq->hw_irq, rc);
2749         return rc;
2750 }
2751
2752 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2753                            struct ibmvnic_sub_crq_queue *scrq)
2754 {
2755         struct device *dev = &adapter->vdev->dev;
2756         unsigned long rc;
2757
2758         if (scrq->hw_irq > 0x100000000ULL) {
2759                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2760                 return 1;
2761         }
2762
2763         if (adapter->resetting &&
2764             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2765                 u64 val = (0xff000000) | scrq->hw_irq;
2766
2767                 rc = plpar_hcall_norets(H_EOI, val);
2768                 if (rc)
2769                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2770                                 val, rc);
2771         }
2772
2773         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2774                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2775         if (rc)
2776                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2777                         scrq->hw_irq, rc);
2778         return rc;
2779 }
2780
2781 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2782                                struct ibmvnic_sub_crq_queue *scrq)
2783 {
2784         struct device *dev = &adapter->vdev->dev;
2785         struct ibmvnic_tx_pool *tx_pool;
2786         struct ibmvnic_tx_buff *txbuff;
2787         union sub_crq *next;
2788         int index;
2789         int i, j;
2790         u8 *first;
2791
2792 restart_loop:
2793         while (pending_scrq(adapter, scrq)) {
2794                 unsigned int pool = scrq->pool_index;
2795                 int num_entries = 0;
2796
2797                 next = ibmvnic_next_scrq(adapter, scrq);
2798                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2799                         if (next->tx_comp.rcs[i]) {
2800                                 dev_err(dev, "tx error %x\n",
2801                                         next->tx_comp.rcs[i]);
2802                                 continue;
2803                         }
2804                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2805                         if (index & IBMVNIC_TSO_POOL_MASK) {
2806                                 tx_pool = &adapter->tso_pool[pool];
2807                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2808                         } else {
2809                                 tx_pool = &adapter->tx_pool[pool];
2810                         }
2811
2812                         txbuff = &tx_pool->tx_buff[index];
2813
2814                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2815                                 if (!txbuff->data_dma[j])
2816                                         continue;
2817
2818                                 txbuff->data_dma[j] = 0;
2819                         }
2820                         /* if sub_crq was sent indirectly */
2821                         first = &txbuff->indir_arr[0].generic.first;
2822                         if (*first == IBMVNIC_CRQ_CMD) {
2823                                 dma_unmap_single(dev, txbuff->indir_dma,
2824                                                  sizeof(txbuff->indir_arr),
2825                                                  DMA_TO_DEVICE);
2826                                 *first = 0;
2827                         }
2828
2829                         if (txbuff->last_frag) {
2830                                 dev_kfree_skb_any(txbuff->skb);
2831                                 txbuff->skb = NULL;
2832                         }
2833
2834                         num_entries += txbuff->num_entries;
2835
2836                         tx_pool->free_map[tx_pool->producer_index] = index;
2837                         tx_pool->producer_index =
2838                                 (tx_pool->producer_index + 1) %
2839                                         tx_pool->num_buffers;
2840                 }
2841                 /* remove tx_comp scrq*/
2842                 next->tx_comp.first = 0;
2843
2844                 if (atomic_sub_return(num_entries, &scrq->used) <=
2845                     (adapter->req_tx_entries_per_subcrq / 2) &&
2846                     __netif_subqueue_stopped(adapter->netdev,
2847                                              scrq->pool_index)) {
2848                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2849                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2850                                    scrq->pool_index);
2851                 }
2852         }
2853
2854         enable_scrq_irq(adapter, scrq);
2855
2856         if (pending_scrq(adapter, scrq)) {
2857                 disable_scrq_irq(adapter, scrq);
2858                 goto restart_loop;
2859         }
2860
2861         return 0;
2862 }
2863
2864 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2865 {
2866         struct ibmvnic_sub_crq_queue *scrq = instance;
2867         struct ibmvnic_adapter *adapter = scrq->adapter;
2868
2869         disable_scrq_irq(adapter, scrq);
2870         ibmvnic_complete_tx(adapter, scrq);
2871
2872         return IRQ_HANDLED;
2873 }
2874
2875 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2876 {
2877         struct ibmvnic_sub_crq_queue *scrq = instance;
2878         struct ibmvnic_adapter *adapter = scrq->adapter;
2879
2880         /* When booting a kdump kernel we can hit pending interrupts
2881          * prior to completing driver initialization.
2882          */
2883         if (unlikely(adapter->state != VNIC_OPEN))
2884                 return IRQ_NONE;
2885
2886         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2887
2888         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2889                 disable_scrq_irq(adapter, scrq);
2890                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2891         }
2892
2893         return IRQ_HANDLED;
2894 }
2895
2896 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2897 {
2898         struct device *dev = &adapter->vdev->dev;
2899         struct ibmvnic_sub_crq_queue *scrq;
2900         int i = 0, j = 0;
2901         int rc = 0;
2902
2903         for (i = 0; i < adapter->req_tx_queues; i++) {
2904                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2905                            i);
2906                 scrq = adapter->tx_scrq[i];
2907                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2908
2909                 if (!scrq->irq) {
2910                         rc = -EINVAL;
2911                         dev_err(dev, "Error mapping irq\n");
2912                         goto req_tx_irq_failed;
2913                 }
2914
2915                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
2916                          adapter->vdev->unit_address, i);
2917                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2918                                  0, scrq->name, scrq);
2919
2920                 if (rc) {
2921                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2922                                 scrq->irq, rc);
2923                         irq_dispose_mapping(scrq->irq);
2924                         goto req_tx_irq_failed;
2925                 }
2926         }
2927
2928         for (i = 0; i < adapter->req_rx_queues; i++) {
2929                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2930                            i);
2931                 scrq = adapter->rx_scrq[i];
2932                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2933                 if (!scrq->irq) {
2934                         rc = -EINVAL;
2935                         dev_err(dev, "Error mapping irq\n");
2936                         goto req_rx_irq_failed;
2937                 }
2938                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
2939                          adapter->vdev->unit_address, i);
2940                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2941                                  0, scrq->name, scrq);
2942                 if (rc) {
2943                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2944                                 scrq->irq, rc);
2945                         irq_dispose_mapping(scrq->irq);
2946                         goto req_rx_irq_failed;
2947                 }
2948         }
2949         return rc;
2950
2951 req_rx_irq_failed:
2952         for (j = 0; j < i; j++) {
2953                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2954                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2955         }
2956         i = adapter->req_tx_queues;
2957 req_tx_irq_failed:
2958         for (j = 0; j < i; j++) {
2959                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2960                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2961         }
2962         release_sub_crqs(adapter, 1);
2963         return rc;
2964 }
2965
2966 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2967 {
2968         struct device *dev = &adapter->vdev->dev;
2969         struct ibmvnic_sub_crq_queue **allqueues;
2970         int registered_queues = 0;
2971         int total_queues;
2972         int more = 0;
2973         int i;
2974
2975         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2976
2977         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2978         if (!allqueues)
2979                 return -1;
2980
2981         for (i = 0; i < total_queues; i++) {
2982                 allqueues[i] = init_sub_crq_queue(adapter);
2983                 if (!allqueues[i]) {
2984                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2985                         break;
2986                 }
2987                 registered_queues++;
2988         }
2989
2990         /* Make sure we were able to register the minimum number of queues */
2991         if (registered_queues <
2992             adapter->min_tx_queues + adapter->min_rx_queues) {
2993                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2994                 goto tx_failed;
2995         }
2996
2997         /* Distribute the failed allocated queues*/
2998         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2999                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3000                 switch (i % 3) {
3001                 case 0:
3002                         if (adapter->req_rx_queues > adapter->min_rx_queues)
3003                                 adapter->req_rx_queues--;
3004                         else
3005                                 more++;
3006                         break;
3007                 case 1:
3008                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3009                                 adapter->req_tx_queues--;
3010                         else
3011                                 more++;
3012                         break;
3013                 }
3014         }
3015
3016         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3017                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3018         if (!adapter->tx_scrq)
3019                 goto tx_failed;
3020
3021         for (i = 0; i < adapter->req_tx_queues; i++) {
3022                 adapter->tx_scrq[i] = allqueues[i];
3023                 adapter->tx_scrq[i]->pool_index = i;
3024                 adapter->num_active_tx_scrqs++;
3025         }
3026
3027         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3028                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3029         if (!adapter->rx_scrq)
3030                 goto rx_failed;
3031
3032         for (i = 0; i < adapter->req_rx_queues; i++) {
3033                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3034                 adapter->rx_scrq[i]->scrq_num = i;
3035                 adapter->num_active_rx_scrqs++;
3036         }
3037
3038         kfree(allqueues);
3039         return 0;
3040
3041 rx_failed:
3042         kfree(adapter->tx_scrq);
3043         adapter->tx_scrq = NULL;
3044 tx_failed:
3045         for (i = 0; i < registered_queues; i++)
3046                 release_sub_crq_queue(adapter, allqueues[i], 1);
3047         kfree(allqueues);
3048         return -1;
3049 }
3050
3051 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3052 {
3053         struct device *dev = &adapter->vdev->dev;
3054         union ibmvnic_crq crq;
3055         int max_entries;
3056
3057         if (!retry) {
3058                 /* Sub-CRQ entries are 32 byte long */
3059                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3060
3061                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3062                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3063                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3064                         return;
3065                 }
3066
3067                 if (adapter->desired.mtu)
3068                         adapter->req_mtu = adapter->desired.mtu;
3069                 else
3070                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3071
3072                 if (!adapter->desired.tx_entries)
3073                         adapter->desired.tx_entries =
3074                                         adapter->max_tx_entries_per_subcrq;
3075                 if (!adapter->desired.rx_entries)
3076                         adapter->desired.rx_entries =
3077                                         adapter->max_rx_add_entries_per_subcrq;
3078
3079                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3080                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3081
3082                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3083                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3084                         adapter->desired.tx_entries = max_entries;
3085                 }
3086
3087                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3088                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3089                         adapter->desired.rx_entries = max_entries;
3090                 }
3091
3092                 if (adapter->desired.tx_entries)
3093                         adapter->req_tx_entries_per_subcrq =
3094                                         adapter->desired.tx_entries;
3095                 else
3096                         adapter->req_tx_entries_per_subcrq =
3097                                         adapter->max_tx_entries_per_subcrq;
3098
3099                 if (adapter->desired.rx_entries)
3100                         adapter->req_rx_add_entries_per_subcrq =
3101                                         adapter->desired.rx_entries;
3102                 else
3103                         adapter->req_rx_add_entries_per_subcrq =
3104                                         adapter->max_rx_add_entries_per_subcrq;
3105
3106                 if (adapter->desired.tx_queues)
3107                         adapter->req_tx_queues =
3108                                         adapter->desired.tx_queues;
3109                 else
3110                         adapter->req_tx_queues =
3111                                         adapter->opt_tx_comp_sub_queues;
3112
3113                 if (adapter->desired.rx_queues)
3114                         adapter->req_rx_queues =
3115                                         adapter->desired.rx_queues;
3116                 else
3117                         adapter->req_rx_queues =
3118                                         adapter->opt_rx_comp_queues;
3119
3120                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3121         }
3122
3123         memset(&crq, 0, sizeof(crq));
3124         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3125         crq.request_capability.cmd = REQUEST_CAPABILITY;
3126
3127         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3128         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3129         atomic_inc(&adapter->running_cap_crqs);
3130         ibmvnic_send_crq(adapter, &crq);
3131
3132         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3133         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3134         atomic_inc(&adapter->running_cap_crqs);
3135         ibmvnic_send_crq(adapter, &crq);
3136
3137         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3138         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3139         atomic_inc(&adapter->running_cap_crqs);
3140         ibmvnic_send_crq(adapter, &crq);
3141
3142         crq.request_capability.capability =
3143             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3144         crq.request_capability.number =
3145             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3146         atomic_inc(&adapter->running_cap_crqs);
3147         ibmvnic_send_crq(adapter, &crq);
3148
3149         crq.request_capability.capability =
3150             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3151         crq.request_capability.number =
3152             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3153         atomic_inc(&adapter->running_cap_crqs);
3154         ibmvnic_send_crq(adapter, &crq);
3155
3156         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3157         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3158         atomic_inc(&adapter->running_cap_crqs);
3159         ibmvnic_send_crq(adapter, &crq);
3160
3161         if (adapter->netdev->flags & IFF_PROMISC) {
3162                 if (adapter->promisc_supported) {
3163                         crq.request_capability.capability =
3164                             cpu_to_be16(PROMISC_REQUESTED);
3165                         crq.request_capability.number = cpu_to_be64(1);
3166                         atomic_inc(&adapter->running_cap_crqs);
3167                         ibmvnic_send_crq(adapter, &crq);
3168                 }
3169         } else {
3170                 crq.request_capability.capability =
3171                     cpu_to_be16(PROMISC_REQUESTED);
3172                 crq.request_capability.number = cpu_to_be64(0);
3173                 atomic_inc(&adapter->running_cap_crqs);
3174                 ibmvnic_send_crq(adapter, &crq);
3175         }
3176 }
3177
3178 static int pending_scrq(struct ibmvnic_adapter *adapter,
3179                         struct ibmvnic_sub_crq_queue *scrq)
3180 {
3181         union sub_crq *entry = &scrq->msgs[scrq->cur];
3182
3183         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3184                 return 1;
3185         else
3186                 return 0;
3187 }
3188
3189 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3190                                         struct ibmvnic_sub_crq_queue *scrq)
3191 {
3192         union sub_crq *entry;
3193         unsigned long flags;
3194
3195         spin_lock_irqsave(&scrq->lock, flags);
3196         entry = &scrq->msgs[scrq->cur];
3197         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3198                 if (++scrq->cur == scrq->size)
3199                         scrq->cur = 0;
3200         } else {
3201                 entry = NULL;
3202         }
3203         spin_unlock_irqrestore(&scrq->lock, flags);
3204
3205         return entry;
3206 }
3207
3208 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3209 {
3210         struct ibmvnic_crq_queue *queue = &adapter->crq;
3211         union ibmvnic_crq *crq;
3212
3213         crq = &queue->msgs[queue->cur];
3214         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3215                 if (++queue->cur == queue->size)
3216                         queue->cur = 0;
3217         } else {
3218                 crq = NULL;
3219         }
3220
3221         return crq;
3222 }
3223
3224 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3225 {
3226         switch (rc) {
3227         case H_PARAMETER:
3228                 dev_warn_ratelimited(dev,
3229                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3230                                      func, rc);
3231                 break;
3232         case H_CLOSED:
3233                 dev_warn_ratelimited(dev,
3234                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3235                                      func, rc);
3236                 break;
3237         default:
3238                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3239                 break;
3240         }
3241 }
3242
3243 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3244                        union sub_crq *sub_crq)
3245 {
3246         unsigned int ua = adapter->vdev->unit_address;
3247         struct device *dev = &adapter->vdev->dev;
3248         u64 *u64_crq = (u64 *)sub_crq;
3249         int rc;
3250
3251         netdev_dbg(adapter->netdev,
3252                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3253                    (unsigned long int)cpu_to_be64(remote_handle),
3254                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3255                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3256                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3257                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3258
3259         /* Make sure the hypervisor sees the complete request */
3260         mb();
3261
3262         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3263                                 cpu_to_be64(remote_handle),
3264                                 cpu_to_be64(u64_crq[0]),
3265                                 cpu_to_be64(u64_crq[1]),
3266                                 cpu_to_be64(u64_crq[2]),
3267                                 cpu_to_be64(u64_crq[3]));
3268
3269         if (rc)
3270                 print_subcrq_error(dev, rc, __func__);
3271
3272         return rc;
3273 }
3274
3275 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3276                                 u64 remote_handle, u64 ioba, u64 num_entries)
3277 {
3278         unsigned int ua = adapter->vdev->unit_address;
3279         struct device *dev = &adapter->vdev->dev;
3280         int rc;
3281
3282         /* Make sure the hypervisor sees the complete request */
3283         mb();
3284         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3285                                 cpu_to_be64(remote_handle),
3286                                 ioba, num_entries);
3287
3288         if (rc)
3289                 print_subcrq_error(dev, rc, __func__);
3290
3291         return rc;
3292 }
3293
3294 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3295                             union ibmvnic_crq *crq)
3296 {
3297         unsigned int ua = adapter->vdev->unit_address;
3298         struct device *dev = &adapter->vdev->dev;
3299         u64 *u64_crq = (u64 *)crq;
3300         int rc;
3301
3302         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3303                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3304                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3305
3306         if (!adapter->crq.active &&
3307             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3308                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3309                 return -EINVAL;
3310         }
3311
3312         /* Make sure the hypervisor sees the complete request */
3313         mb();
3314
3315         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3316                                 cpu_to_be64(u64_crq[0]),
3317                                 cpu_to_be64(u64_crq[1]));
3318
3319         if (rc) {
3320                 if (rc == H_CLOSED) {
3321                         dev_warn(dev, "CRQ Queue closed\n");
3322                         if (adapter->resetting)
3323                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3324                 }
3325
3326                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3327         }
3328
3329         return rc;
3330 }
3331
3332 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3333 {
3334         union ibmvnic_crq crq;
3335
3336         memset(&crq, 0, sizeof(crq));
3337         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3338         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3339         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3340
3341         return ibmvnic_send_crq(adapter, &crq);
3342 }
3343
3344 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3345 {
3346         union ibmvnic_crq crq;
3347
3348         memset(&crq, 0, sizeof(crq));
3349         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3350         crq.version_exchange.cmd = VERSION_EXCHANGE;
3351         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3352
3353         return ibmvnic_send_crq(adapter, &crq);
3354 }
3355
3356 struct vnic_login_client_data {
3357         u8      type;
3358         __be16  len;
3359         char    name[];
3360 } __packed;
3361
3362 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3363 {
3364         int len;
3365
3366         /* Calculate the amount of buffer space needed for the
3367          * vnic client data in the login buffer. There are four entries,
3368          * OS name, LPAR name, device name, and a null last entry.
3369          */
3370         len = 4 * sizeof(struct vnic_login_client_data);
3371         len += 6; /* "Linux" plus NULL */
3372         len += strlen(utsname()->nodename) + 1;
3373         len += strlen(adapter->netdev->name) + 1;
3374
3375         return len;
3376 }
3377
3378 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3379                                  struct vnic_login_client_data *vlcd)
3380 {
3381         const char *os_name = "Linux";
3382         int len;
3383
3384         /* Type 1 - LPAR OS */
3385         vlcd->type = 1;
3386         len = strlen(os_name) + 1;
3387         vlcd->len = cpu_to_be16(len);
3388         strncpy(vlcd->name, os_name, len);
3389         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3390
3391         /* Type 2 - LPAR name */
3392         vlcd->type = 2;
3393         len = strlen(utsname()->nodename) + 1;
3394         vlcd->len = cpu_to_be16(len);
3395         strncpy(vlcd->name, utsname()->nodename, len);
3396         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3397
3398         /* Type 3 - device name */
3399         vlcd->type = 3;
3400         len = strlen(adapter->netdev->name) + 1;
3401         vlcd->len = cpu_to_be16(len);
3402         strncpy(vlcd->name, adapter->netdev->name, len);
3403 }
3404
3405 static int send_login(struct ibmvnic_adapter *adapter)
3406 {
3407         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3408         struct ibmvnic_login_buffer *login_buffer;
3409         struct device *dev = &adapter->vdev->dev;
3410         dma_addr_t rsp_buffer_token;
3411         dma_addr_t buffer_token;
3412         size_t rsp_buffer_size;
3413         union ibmvnic_crq crq;
3414         size_t buffer_size;
3415         __be64 *tx_list_p;
3416         __be64 *rx_list_p;
3417         int client_data_len;
3418         struct vnic_login_client_data *vlcd;
3419         int i;
3420
3421         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3422                 netdev_err(adapter->netdev,
3423                            "RX or TX queues are not allocated, device login failed\n");
3424                 return -1;
3425         }
3426
3427         release_login_rsp_buffer(adapter);
3428         client_data_len = vnic_client_data_len(adapter);
3429
3430         buffer_size =
3431             sizeof(struct ibmvnic_login_buffer) +
3432             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3433             client_data_len;
3434
3435         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3436         if (!login_buffer)
3437                 goto buf_alloc_failed;
3438
3439         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3440                                       DMA_TO_DEVICE);
3441         if (dma_mapping_error(dev, buffer_token)) {
3442                 dev_err(dev, "Couldn't map login buffer\n");
3443                 goto buf_map_failed;
3444         }
3445
3446         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3447                           sizeof(u64) * adapter->req_tx_queues +
3448                           sizeof(u64) * adapter->req_rx_queues +
3449                           sizeof(u64) * adapter->req_rx_queues +
3450                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3451
3452         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3453         if (!login_rsp_buffer)
3454                 goto buf_rsp_alloc_failed;
3455
3456         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3457                                           rsp_buffer_size, DMA_FROM_DEVICE);
3458         if (dma_mapping_error(dev, rsp_buffer_token)) {
3459                 dev_err(dev, "Couldn't map login rsp buffer\n");
3460                 goto buf_rsp_map_failed;
3461         }
3462
3463         adapter->login_buf = login_buffer;
3464         adapter->login_buf_token = buffer_token;
3465         adapter->login_buf_sz = buffer_size;
3466         adapter->login_rsp_buf = login_rsp_buffer;
3467         adapter->login_rsp_buf_token = rsp_buffer_token;
3468         adapter->login_rsp_buf_sz = rsp_buffer_size;
3469
3470         login_buffer->len = cpu_to_be32(buffer_size);
3471         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3472         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3473         login_buffer->off_txcomp_subcrqs =
3474             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3475         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3476         login_buffer->off_rxcomp_subcrqs =
3477             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3478                         sizeof(u64) * adapter->req_tx_queues);
3479         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3480         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3481
3482         tx_list_p = (__be64 *)((char *)login_buffer +
3483                                       sizeof(struct ibmvnic_login_buffer));
3484         rx_list_p = (__be64 *)((char *)login_buffer +
3485                                       sizeof(struct ibmvnic_login_buffer) +
3486                                       sizeof(u64) * adapter->req_tx_queues);
3487
3488         for (i = 0; i < adapter->req_tx_queues; i++) {
3489                 if (adapter->tx_scrq[i]) {
3490                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3491                                                    crq_num);
3492                 }
3493         }
3494
3495         for (i = 0; i < adapter->req_rx_queues; i++) {
3496                 if (adapter->rx_scrq[i]) {
3497                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3498                                                    crq_num);
3499                 }
3500         }
3501
3502         /* Insert vNIC login client data */
3503         vlcd = (struct vnic_login_client_data *)
3504                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3505         login_buffer->client_data_offset =
3506                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3507         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3508
3509         vnic_add_client_data(adapter, vlcd);
3510
3511         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3512         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3513                 netdev_dbg(adapter->netdev, "%016lx\n",
3514                            ((unsigned long int *)(adapter->login_buf))[i]);
3515         }
3516
3517         memset(&crq, 0, sizeof(crq));
3518         crq.login.first = IBMVNIC_CRQ_CMD;
3519         crq.login.cmd = LOGIN;
3520         crq.login.ioba = cpu_to_be32(buffer_token);
3521         crq.login.len = cpu_to_be32(buffer_size);
3522         ibmvnic_send_crq(adapter, &crq);
3523
3524         return 0;
3525
3526 buf_rsp_map_failed:
3527         kfree(login_rsp_buffer);
3528 buf_rsp_alloc_failed:
3529         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3530 buf_map_failed:
3531         kfree(login_buffer);
3532 buf_alloc_failed:
3533         return -1;
3534 }
3535
3536 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3537                             u32 len, u8 map_id)
3538 {
3539         union ibmvnic_crq crq;
3540
3541         memset(&crq, 0, sizeof(crq));
3542         crq.request_map.first = IBMVNIC_CRQ_CMD;
3543         crq.request_map.cmd = REQUEST_MAP;
3544         crq.request_map.map_id = map_id;
3545         crq.request_map.ioba = cpu_to_be32(addr);
3546         crq.request_map.len = cpu_to_be32(len);
3547         return ibmvnic_send_crq(adapter, &crq);
3548 }
3549
3550 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3551 {
3552         union ibmvnic_crq crq;
3553
3554         memset(&crq, 0, sizeof(crq));
3555         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3556         crq.request_unmap.cmd = REQUEST_UNMAP;
3557         crq.request_unmap.map_id = map_id;
3558         return ibmvnic_send_crq(adapter, &crq);
3559 }
3560
3561 static void send_map_query(struct ibmvnic_adapter *adapter)
3562 {
3563         union ibmvnic_crq crq;
3564
3565         memset(&crq, 0, sizeof(crq));
3566         crq.query_map.first = IBMVNIC_CRQ_CMD;
3567         crq.query_map.cmd = QUERY_MAP;
3568         ibmvnic_send_crq(adapter, &crq);
3569 }
3570
3571 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3572 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3573 {
3574         union ibmvnic_crq crq;
3575
3576         atomic_set(&adapter->running_cap_crqs, 0);
3577         memset(&crq, 0, sizeof(crq));
3578         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3579         crq.query_capability.cmd = QUERY_CAPABILITY;
3580
3581         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3582         atomic_inc(&adapter->running_cap_crqs);
3583         ibmvnic_send_crq(adapter, &crq);
3584
3585         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3586         atomic_inc(&adapter->running_cap_crqs);
3587         ibmvnic_send_crq(adapter, &crq);
3588
3589         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3590         atomic_inc(&adapter->running_cap_crqs);
3591         ibmvnic_send_crq(adapter, &crq);
3592
3593         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3594         atomic_inc(&adapter->running_cap_crqs);
3595         ibmvnic_send_crq(adapter, &crq);
3596
3597         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3598         atomic_inc(&adapter->running_cap_crqs);
3599         ibmvnic_send_crq(adapter, &crq);
3600
3601         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3602         atomic_inc(&adapter->running_cap_crqs);
3603         ibmvnic_send_crq(adapter, &crq);
3604
3605         crq.query_capability.capability =
3606             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3607         atomic_inc(&adapter->running_cap_crqs);
3608         ibmvnic_send_crq(adapter, &crq);
3609
3610         crq.query_capability.capability =
3611             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3612         atomic_inc(&adapter->running_cap_crqs);
3613         ibmvnic_send_crq(adapter, &crq);
3614
3615         crq.query_capability.capability =
3616             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3617         atomic_inc(&adapter->running_cap_crqs);
3618         ibmvnic_send_crq(adapter, &crq);
3619
3620         crq.query_capability.capability =
3621             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3622         atomic_inc(&adapter->running_cap_crqs);
3623         ibmvnic_send_crq(adapter, &crq);
3624
3625         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3626         atomic_inc(&adapter->running_cap_crqs);
3627         ibmvnic_send_crq(adapter, &crq);
3628
3629         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3630         atomic_inc(&adapter->running_cap_crqs);
3631         ibmvnic_send_crq(adapter, &crq);
3632
3633         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3634         atomic_inc(&adapter->running_cap_crqs);
3635         ibmvnic_send_crq(adapter, &crq);
3636
3637         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3638         atomic_inc(&adapter->running_cap_crqs);
3639         ibmvnic_send_crq(adapter, &crq);
3640
3641         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3642         atomic_inc(&adapter->running_cap_crqs);
3643         ibmvnic_send_crq(adapter, &crq);
3644
3645         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3646         atomic_inc(&adapter->running_cap_crqs);
3647         ibmvnic_send_crq(adapter, &crq);
3648
3649         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3650         atomic_inc(&adapter->running_cap_crqs);
3651         ibmvnic_send_crq(adapter, &crq);
3652
3653         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3654         atomic_inc(&adapter->running_cap_crqs);
3655         ibmvnic_send_crq(adapter, &crq);
3656
3657         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3658         atomic_inc(&adapter->running_cap_crqs);
3659         ibmvnic_send_crq(adapter, &crq);
3660
3661         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3662         atomic_inc(&adapter->running_cap_crqs);
3663         ibmvnic_send_crq(adapter, &crq);
3664
3665         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3666         atomic_inc(&adapter->running_cap_crqs);
3667         ibmvnic_send_crq(adapter, &crq);
3668
3669         crq.query_capability.capability =
3670                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3671         atomic_inc(&adapter->running_cap_crqs);
3672         ibmvnic_send_crq(adapter, &crq);
3673
3674         crq.query_capability.capability =
3675                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3676         atomic_inc(&adapter->running_cap_crqs);
3677         ibmvnic_send_crq(adapter, &crq);
3678
3679         crq.query_capability.capability =
3680                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3681         atomic_inc(&adapter->running_cap_crqs);
3682         ibmvnic_send_crq(adapter, &crq);
3683
3684         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3685         atomic_inc(&adapter->running_cap_crqs);
3686         ibmvnic_send_crq(adapter, &crq);
3687 }
3688
3689 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3690                                 struct ibmvnic_adapter *adapter)
3691 {
3692         struct device *dev = &adapter->vdev->dev;
3693
3694         if (crq->get_vpd_size_rsp.rc.code) {
3695                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3696                         crq->get_vpd_size_rsp.rc.code);
3697                 complete(&adapter->fw_done);
3698                 return;
3699         }
3700
3701         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3702         complete(&adapter->fw_done);
3703 }
3704
3705 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3706                            struct ibmvnic_adapter *adapter)
3707 {
3708         struct device *dev = &adapter->vdev->dev;
3709         unsigned char *substr = NULL;
3710         u8 fw_level_len = 0;
3711
3712         memset(adapter->fw_version, 0, 32);
3713
3714         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3715                          DMA_FROM_DEVICE);
3716
3717         if (crq->get_vpd_rsp.rc.code) {
3718                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3719                         crq->get_vpd_rsp.rc.code);
3720                 goto complete;
3721         }
3722
3723         /* get the position of the firmware version info
3724          * located after the ASCII 'RM' substring in the buffer
3725          */
3726         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3727         if (!substr) {
3728                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3729                 goto complete;
3730         }
3731
3732         /* get length of firmware level ASCII substring */
3733         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3734                 fw_level_len = *(substr + 2);
3735         } else {
3736                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3737                 goto complete;
3738         }
3739
3740         /* copy firmware version string from vpd into adapter */
3741         if ((substr + 3 + fw_level_len) <
3742             (adapter->vpd->buff + adapter->vpd->len)) {
3743                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3744         } else {
3745                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3746         }
3747
3748 complete:
3749         if (adapter->fw_version[0] == '\0')
3750                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3751         complete(&adapter->fw_done);
3752 }
3753
3754 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3755 {
3756         struct device *dev = &adapter->vdev->dev;
3757         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3758         netdev_features_t old_hw_features = 0;
3759         union ibmvnic_crq crq;
3760         int i;
3761
3762         dma_unmap_single(dev, adapter->ip_offload_tok,
3763                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3764
3765         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3766         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3767                 netdev_dbg(adapter->netdev, "%016lx\n",
3768                            ((unsigned long int *)(buf))[i]);
3769
3770         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3771         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3772         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3773                    buf->tcp_ipv4_chksum);
3774         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3775                    buf->tcp_ipv6_chksum);
3776         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3777                    buf->udp_ipv4_chksum);
3778         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3779                    buf->udp_ipv6_chksum);
3780         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3781                    buf->large_tx_ipv4);
3782         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3783                    buf->large_tx_ipv6);
3784         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3785                    buf->large_rx_ipv4);
3786         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3787                    buf->large_rx_ipv6);
3788         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3789                    buf->max_ipv4_header_size);
3790         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3791                    buf->max_ipv6_header_size);
3792         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3793                    buf->max_tcp_header_size);
3794         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3795                    buf->max_udp_header_size);
3796         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3797                    buf->max_large_tx_size);
3798         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3799                    buf->max_large_rx_size);
3800         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3801                    buf->ipv6_extension_header);
3802         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3803                    buf->tcp_pseudosum_req);
3804         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3805                    buf->num_ipv6_ext_headers);
3806         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3807                    buf->off_ipv6_ext_headers);
3808
3809         adapter->ip_offload_ctrl_tok =
3810             dma_map_single(dev, &adapter->ip_offload_ctrl,
3811                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3812
3813         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3814                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3815                 return;
3816         }
3817
3818         adapter->ip_offload_ctrl.len =
3819             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3820         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3821         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3822         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3823         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3824         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3825         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3826         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3827         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3828         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3829
3830         /* large_rx disabled for now, additional features needed */
3831         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3832         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3833
3834         if (adapter->state != VNIC_PROBING) {
3835                 old_hw_features = adapter->netdev->hw_features;
3836                 adapter->netdev->hw_features = 0;
3837         }
3838
3839         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
3840
3841         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3842                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
3843
3844         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3845                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
3846
3847         if ((adapter->netdev->features &
3848             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3849                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
3850
3851         if (buf->large_tx_ipv4)
3852                 adapter->netdev->hw_features |= NETIF_F_TSO;
3853         if (buf->large_tx_ipv6)
3854                 adapter->netdev->hw_features |= NETIF_F_TSO6;
3855
3856         if (adapter->state == VNIC_PROBING) {
3857                 adapter->netdev->features |= adapter->netdev->hw_features;
3858         } else if (old_hw_features != adapter->netdev->hw_features) {
3859                 netdev_features_t tmp = 0;
3860
3861                 /* disable features no longer supported */
3862                 adapter->netdev->features &= adapter->netdev->hw_features;
3863                 /* turn on features now supported if previously enabled */
3864                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
3865                         adapter->netdev->hw_features;
3866                 adapter->netdev->features |=
3867                                 tmp & adapter->netdev->wanted_features;
3868         }
3869
3870         memset(&crq, 0, sizeof(crq));
3871         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3872         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3873         crq.control_ip_offload.len =
3874             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3875         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3876         ibmvnic_send_crq(adapter, &crq);
3877 }
3878
3879 static const char *ibmvnic_fw_err_cause(u16 cause)
3880 {
3881         switch (cause) {
3882         case ADAPTER_PROBLEM:
3883                 return "adapter problem";
3884         case BUS_PROBLEM:
3885                 return "bus problem";
3886         case FW_PROBLEM:
3887                 return "firmware problem";
3888         case DD_PROBLEM:
3889                 return "device driver problem";
3890         case EEH_RECOVERY:
3891                 return "EEH recovery";
3892         case FW_UPDATED:
3893                 return "firmware updated";
3894         case LOW_MEMORY:
3895                 return "low Memory";
3896         default:
3897                 return "unknown";
3898         }
3899 }
3900
3901 static void handle_error_indication(union ibmvnic_crq *crq,
3902                                     struct ibmvnic_adapter *adapter)
3903 {
3904         struct device *dev = &adapter->vdev->dev;
3905         u16 cause;
3906
3907         cause = be16_to_cpu(crq->error_indication.error_cause);
3908
3909         dev_warn_ratelimited(dev,
3910                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3911                              crq->error_indication.flags
3912                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3913                              ibmvnic_fw_err_cause(cause));
3914
3915         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3916                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3917         else
3918                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3919 }
3920
3921 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3922                                  struct ibmvnic_adapter *adapter)
3923 {
3924         struct net_device *netdev = adapter->netdev;
3925         struct device *dev = &adapter->vdev->dev;
3926         long rc;
3927
3928         rc = crq->change_mac_addr_rsp.rc.code;
3929         if (rc) {
3930                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3931                 goto out;
3932         }
3933         ether_addr_copy(netdev->dev_addr,
3934                         &crq->change_mac_addr_rsp.mac_addr[0]);
3935 out:
3936         complete(&adapter->fw_done);
3937         return rc;
3938 }
3939
3940 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3941                                    struct ibmvnic_adapter *adapter)
3942 {
3943         struct device *dev = &adapter->vdev->dev;
3944         u64 *req_value;
3945         char *name;
3946
3947         atomic_dec(&adapter->running_cap_crqs);
3948         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3949         case REQ_TX_QUEUES:
3950                 req_value = &adapter->req_tx_queues;
3951                 name = "tx";
3952                 break;
3953         case REQ_RX_QUEUES:
3954                 req_value = &adapter->req_rx_queues;
3955                 name = "rx";
3956                 break;
3957         case REQ_RX_ADD_QUEUES:
3958                 req_value = &adapter->req_rx_add_queues;
3959                 name = "rx_add";
3960                 break;
3961         case REQ_TX_ENTRIES_PER_SUBCRQ:
3962                 req_value = &adapter->req_tx_entries_per_subcrq;
3963                 name = "tx_entries_per_subcrq";
3964                 break;
3965         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3966                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3967                 name = "rx_add_entries_per_subcrq";
3968                 break;
3969         case REQ_MTU:
3970                 req_value = &adapter->req_mtu;
3971                 name = "mtu";
3972                 break;
3973         case PROMISC_REQUESTED:
3974                 req_value = &adapter->promisc;
3975                 name = "promisc";
3976                 break;
3977         default:
3978                 dev_err(dev, "Got invalid cap request rsp %d\n",
3979                         crq->request_capability.capability);
3980                 return;
3981         }
3982
3983         switch (crq->request_capability_rsp.rc.code) {
3984         case SUCCESS:
3985                 break;
3986         case PARTIALSUCCESS:
3987                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3988                          *req_value,
3989                          (long int)be64_to_cpu(crq->request_capability_rsp.
3990                                                number), name);
3991
3992                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3993                     REQ_MTU) {
3994                         pr_err("mtu of %llu is not supported. Reverting.\n",
3995                                *req_value);
3996                         *req_value = adapter->fallback.mtu;
3997                 } else {
3998                         *req_value =
3999                                 be64_to_cpu(crq->request_capability_rsp.number);
4000                 }
4001
4002                 ibmvnic_send_req_caps(adapter, 1);
4003                 return;
4004         default:
4005                 dev_err(dev, "Error %d in request cap rsp\n",
4006                         crq->request_capability_rsp.rc.code);
4007                 return;
4008         }
4009
4010         /* Done receiving requested capabilities, query IP offload support */
4011         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4012                 union ibmvnic_crq newcrq;
4013                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4014                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4015                     &adapter->ip_offload_buf;
4016
4017                 adapter->wait_capability = false;
4018                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4019                                                          buf_sz,
4020                                                          DMA_FROM_DEVICE);
4021
4022                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4023                         if (!firmware_has_feature(FW_FEATURE_CMO))
4024                                 dev_err(dev, "Couldn't map offload buffer\n");
4025                         return;
4026                 }
4027
4028                 memset(&newcrq, 0, sizeof(newcrq));
4029                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4030                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4031                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4032                 newcrq.query_ip_offload.ioba =
4033                     cpu_to_be32(adapter->ip_offload_tok);
4034
4035                 ibmvnic_send_crq(adapter, &newcrq);
4036         }
4037 }
4038
4039 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4040                             struct ibmvnic_adapter *adapter)
4041 {
4042         struct device *dev = &adapter->vdev->dev;
4043         struct net_device *netdev = adapter->netdev;
4044         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4045         struct ibmvnic_login_buffer *login = adapter->login_buf;
4046         int i;
4047
4048         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4049                          DMA_TO_DEVICE);
4050         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4051                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4052
4053         /* If the number of queues requested can't be allocated by the
4054          * server, the login response will return with code 1. We will need
4055          * to resend the login buffer with fewer queues requested.
4056          */
4057         if (login_rsp_crq->generic.rc.code) {
4058                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4059                 complete(&adapter->init_done);
4060                 return 0;
4061         }
4062
4063         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4064
4065         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4066         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4067                 netdev_dbg(adapter->netdev, "%016lx\n",
4068                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4069         }
4070
4071         /* Sanity checks */
4072         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4073             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4074              adapter->req_rx_add_queues !=
4075              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4076                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4077                 ibmvnic_remove(adapter->vdev);
4078                 return -EIO;
4079         }
4080         release_login_buffer(adapter);
4081         complete(&adapter->init_done);
4082
4083         return 0;
4084 }
4085
4086 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4087                                      struct ibmvnic_adapter *adapter)
4088 {
4089         struct device *dev = &adapter->vdev->dev;
4090         long rc;
4091
4092         rc = crq->request_unmap_rsp.rc.code;
4093         if (rc)
4094                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4095 }
4096
4097 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4098                                  struct ibmvnic_adapter *adapter)
4099 {
4100         struct net_device *netdev = adapter->netdev;
4101         struct device *dev = &adapter->vdev->dev;
4102         long rc;
4103
4104         rc = crq->query_map_rsp.rc.code;
4105         if (rc) {
4106                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4107                 return;
4108         }
4109         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4110                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4111                    crq->query_map_rsp.free_pages);
4112 }
4113
4114 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4115                                  struct ibmvnic_adapter *adapter)
4116 {
4117         struct net_device *netdev = adapter->netdev;
4118         struct device *dev = &adapter->vdev->dev;
4119         long rc;
4120
4121         atomic_dec(&adapter->running_cap_crqs);
4122         netdev_dbg(netdev, "Outstanding queries: %d\n",
4123                    atomic_read(&adapter->running_cap_crqs));
4124         rc = crq->query_capability.rc.code;
4125         if (rc) {
4126                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4127                 goto out;
4128         }
4129
4130         switch (be16_to_cpu(crq->query_capability.capability)) {
4131         case MIN_TX_QUEUES:
4132                 adapter->min_tx_queues =
4133                     be64_to_cpu(crq->query_capability.number);
4134                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4135                            adapter->min_tx_queues);
4136                 break;
4137         case MIN_RX_QUEUES:
4138                 adapter->min_rx_queues =
4139                     be64_to_cpu(crq->query_capability.number);
4140                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4141                            adapter->min_rx_queues);
4142                 break;
4143         case MIN_RX_ADD_QUEUES:
4144                 adapter->min_rx_add_queues =
4145                     be64_to_cpu(crq->query_capability.number);
4146                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4147                            adapter->min_rx_add_queues);
4148                 break;
4149         case MAX_TX_QUEUES:
4150                 adapter->max_tx_queues =
4151                     be64_to_cpu(crq->query_capability.number);
4152                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4153                            adapter->max_tx_queues);
4154                 break;
4155         case MAX_RX_QUEUES:
4156                 adapter->max_rx_queues =
4157                     be64_to_cpu(crq->query_capability.number);
4158                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4159                            adapter->max_rx_queues);
4160                 break;
4161         case MAX_RX_ADD_QUEUES:
4162                 adapter->max_rx_add_queues =
4163                     be64_to_cpu(crq->query_capability.number);
4164                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4165                            adapter->max_rx_add_queues);
4166                 break;
4167         case MIN_TX_ENTRIES_PER_SUBCRQ:
4168                 adapter->min_tx_entries_per_subcrq =
4169                     be64_to_cpu(crq->query_capability.number);
4170                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4171                            adapter->min_tx_entries_per_subcrq);
4172                 break;
4173         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4174                 adapter->min_rx_add_entries_per_subcrq =
4175                     be64_to_cpu(crq->query_capability.number);
4176                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4177                            adapter->min_rx_add_entries_per_subcrq);
4178                 break;
4179         case MAX_TX_ENTRIES_PER_SUBCRQ:
4180                 adapter->max_tx_entries_per_subcrq =
4181                     be64_to_cpu(crq->query_capability.number);
4182                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4183                            adapter->max_tx_entries_per_subcrq);
4184                 break;
4185         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4186                 adapter->max_rx_add_entries_per_subcrq =
4187                     be64_to_cpu(crq->query_capability.number);
4188                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4189                            adapter->max_rx_add_entries_per_subcrq);
4190                 break;
4191         case TCP_IP_OFFLOAD:
4192                 adapter->tcp_ip_offload =
4193                     be64_to_cpu(crq->query_capability.number);
4194                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4195                            adapter->tcp_ip_offload);
4196                 break;
4197         case PROMISC_SUPPORTED:
4198                 adapter->promisc_supported =
4199                     be64_to_cpu(crq->query_capability.number);
4200                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4201                            adapter->promisc_supported);
4202                 break;
4203         case MIN_MTU:
4204                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4205                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4206                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4207                 break;
4208         case MAX_MTU:
4209                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4210                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4211                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4212                 break;
4213         case MAX_MULTICAST_FILTERS:
4214                 adapter->max_multicast_filters =
4215                     be64_to_cpu(crq->query_capability.number);
4216                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4217                            adapter->max_multicast_filters);
4218                 break;
4219         case VLAN_HEADER_INSERTION:
4220                 adapter->vlan_header_insertion =
4221                     be64_to_cpu(crq->query_capability.number);
4222                 if (adapter->vlan_header_insertion)
4223                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4224                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4225                            adapter->vlan_header_insertion);
4226                 break;
4227         case RX_VLAN_HEADER_INSERTION:
4228                 adapter->rx_vlan_header_insertion =
4229                     be64_to_cpu(crq->query_capability.number);
4230                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4231                            adapter->rx_vlan_header_insertion);
4232                 break;
4233         case MAX_TX_SG_ENTRIES:
4234                 adapter->max_tx_sg_entries =
4235                     be64_to_cpu(crq->query_capability.number);
4236                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4237                            adapter->max_tx_sg_entries);
4238                 break;
4239         case RX_SG_SUPPORTED:
4240                 adapter->rx_sg_supported =
4241                     be64_to_cpu(crq->query_capability.number);
4242                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4243                            adapter->rx_sg_supported);
4244                 break;
4245         case OPT_TX_COMP_SUB_QUEUES:
4246                 adapter->opt_tx_comp_sub_queues =
4247                     be64_to_cpu(crq->query_capability.number);
4248                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4249                            adapter->opt_tx_comp_sub_queues);
4250                 break;
4251         case OPT_RX_COMP_QUEUES:
4252                 adapter->opt_rx_comp_queues =
4253                     be64_to_cpu(crq->query_capability.number);
4254                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4255                            adapter->opt_rx_comp_queues);
4256                 break;
4257         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4258                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4259                     be64_to_cpu(crq->query_capability.number);
4260                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4261                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4262                 break;
4263         case OPT_TX_ENTRIES_PER_SUBCRQ:
4264                 adapter->opt_tx_entries_per_subcrq =
4265                     be64_to_cpu(crq->query_capability.number);
4266                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4267                            adapter->opt_tx_entries_per_subcrq);
4268                 break;
4269         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4270                 adapter->opt_rxba_entries_per_subcrq =
4271                     be64_to_cpu(crq->query_capability.number);
4272                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4273                            adapter->opt_rxba_entries_per_subcrq);
4274                 break;
4275         case TX_RX_DESC_REQ:
4276                 adapter->tx_rx_desc_req = crq->query_capability.number;
4277                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4278                            adapter->tx_rx_desc_req);
4279                 break;
4280
4281         default:
4282                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4283                            crq->query_capability.capability);
4284         }
4285
4286 out:
4287         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4288                 adapter->wait_capability = false;
4289                 ibmvnic_send_req_caps(adapter, 0);
4290         }
4291 }
4292
4293 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4294 {
4295         union ibmvnic_crq crq;
4296         int rc;
4297
4298         memset(&crq, 0, sizeof(crq));
4299         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4300         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4301         init_completion(&adapter->fw_done);
4302         rc = ibmvnic_send_crq(adapter, &crq);
4303         if (rc)
4304                 return rc;
4305         wait_for_completion(&adapter->fw_done);
4306         return adapter->fw_done_rc ? -EIO : 0;
4307 }
4308
4309 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4310                                        struct ibmvnic_adapter *adapter)
4311 {
4312         struct net_device *netdev = adapter->netdev;
4313         int rc;
4314
4315         rc = crq->query_phys_parms_rsp.rc.code;
4316         if (rc) {
4317                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4318                 return rc;
4319         }
4320         switch (cpu_to_be32(crq->query_phys_parms_rsp.speed)) {
4321         case IBMVNIC_10MBPS:
4322                 adapter->speed = SPEED_10;
4323                 break;
4324         case IBMVNIC_100MBPS:
4325                 adapter->speed = SPEED_100;
4326                 break;
4327         case IBMVNIC_1GBPS:
4328                 adapter->speed = SPEED_1000;
4329                 break;
4330         case IBMVNIC_10GBP:
4331                 adapter->speed = SPEED_10000;
4332                 break;
4333         case IBMVNIC_25GBPS:
4334                 adapter->speed = SPEED_25000;
4335                 break;
4336         case IBMVNIC_40GBPS:
4337                 adapter->speed = SPEED_40000;
4338                 break;
4339         case IBMVNIC_50GBPS:
4340                 adapter->speed = SPEED_50000;
4341                 break;
4342         case IBMVNIC_100GBPS:
4343                 adapter->speed = SPEED_100000;
4344                 break;
4345         default:
4346                 netdev_warn(netdev, "Unknown speed 0x%08x\n",
4347                             cpu_to_be32(crq->query_phys_parms_rsp.speed));
4348                 adapter->speed = SPEED_UNKNOWN;
4349         }
4350         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4351                 adapter->duplex = DUPLEX_FULL;
4352         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4353                 adapter->duplex = DUPLEX_HALF;
4354         else
4355                 adapter->duplex = DUPLEX_UNKNOWN;
4356
4357         return rc;
4358 }
4359
4360 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4361                                struct ibmvnic_adapter *adapter)
4362 {
4363         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4364         struct net_device *netdev = adapter->netdev;
4365         struct device *dev = &adapter->vdev->dev;
4366         u64 *u64_crq = (u64 *)crq;
4367         long rc;
4368
4369         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4370                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4371                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4372         switch (gen_crq->first) {
4373         case IBMVNIC_CRQ_INIT_RSP:
4374                 switch (gen_crq->cmd) {
4375                 case IBMVNIC_CRQ_INIT:
4376                         dev_info(dev, "Partner initialized\n");
4377                         adapter->from_passive_init = true;
4378                         adapter->failover_pending = false;
4379                         if (!completion_done(&adapter->init_done)) {
4380                                 complete(&adapter->init_done);
4381                                 adapter->init_done_rc = -EIO;
4382                         }
4383                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4384                         break;
4385                 case IBMVNIC_CRQ_INIT_COMPLETE:
4386                         dev_info(dev, "Partner initialization complete\n");
4387                         adapter->crq.active = true;
4388                         send_version_xchg(adapter);
4389                         break;
4390                 default:
4391                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4392                 }
4393                 return;
4394         case IBMVNIC_CRQ_XPORT_EVENT:
4395                 netif_carrier_off(netdev);
4396                 adapter->crq.active = false;
4397                 if (adapter->resetting)
4398                         adapter->force_reset_recovery = true;
4399                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4400                         dev_info(dev, "Migrated, re-enabling adapter\n");
4401                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4402                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4403                         dev_info(dev, "Backing device failover detected\n");
4404                         adapter->failover_pending = true;
4405                 } else {
4406                         /* The adapter lost the connection */
4407                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4408                                 gen_crq->cmd);
4409                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4410                 }
4411                 return;
4412         case IBMVNIC_CRQ_CMD_RSP:
4413                 break;
4414         default:
4415                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4416                         gen_crq->first);
4417                 return;
4418         }
4419
4420         switch (gen_crq->cmd) {
4421         case VERSION_EXCHANGE_RSP:
4422                 rc = crq->version_exchange_rsp.rc.code;
4423                 if (rc) {
4424                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4425                         break;
4426                 }
4427                 dev_info(dev, "Partner protocol version is %d\n",
4428                          crq->version_exchange_rsp.version);
4429                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4430                     ibmvnic_version)
4431                         ibmvnic_version =
4432                             be16_to_cpu(crq->version_exchange_rsp.version);
4433                 send_cap_queries(adapter);
4434                 break;
4435         case QUERY_CAPABILITY_RSP:
4436                 handle_query_cap_rsp(crq, adapter);
4437                 break;
4438         case QUERY_MAP_RSP:
4439                 handle_query_map_rsp(crq, adapter);
4440                 break;
4441         case REQUEST_MAP_RSP:
4442                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4443                 complete(&adapter->fw_done);
4444                 break;
4445         case REQUEST_UNMAP_RSP:
4446                 handle_request_unmap_rsp(crq, adapter);
4447                 break;
4448         case REQUEST_CAPABILITY_RSP:
4449                 handle_request_cap_rsp(crq, adapter);
4450                 break;
4451         case LOGIN_RSP:
4452                 netdev_dbg(netdev, "Got Login Response\n");
4453                 handle_login_rsp(crq, adapter);
4454                 break;
4455         case LOGICAL_LINK_STATE_RSP:
4456                 netdev_dbg(netdev,
4457                            "Got Logical Link State Response, state: %d rc: %d\n",
4458                            crq->logical_link_state_rsp.link_state,
4459                            crq->logical_link_state_rsp.rc.code);
4460                 adapter->logical_link_state =
4461                     crq->logical_link_state_rsp.link_state;
4462                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4463                 complete(&adapter->init_done);
4464                 break;
4465         case LINK_STATE_INDICATION:
4466                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4467                 adapter->phys_link_state =
4468                     crq->link_state_indication.phys_link_state;
4469                 adapter->logical_link_state =
4470                     crq->link_state_indication.logical_link_state;
4471                 if (adapter->phys_link_state && adapter->logical_link_state)
4472                         netif_carrier_on(netdev);
4473                 else
4474                         netif_carrier_off(netdev);
4475                 break;
4476         case CHANGE_MAC_ADDR_RSP:
4477                 netdev_dbg(netdev, "Got MAC address change Response\n");
4478                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4479                 break;
4480         case ERROR_INDICATION:
4481                 netdev_dbg(netdev, "Got Error Indication\n");
4482                 handle_error_indication(crq, adapter);
4483                 break;
4484         case REQUEST_STATISTICS_RSP:
4485                 netdev_dbg(netdev, "Got Statistics Response\n");
4486                 complete(&adapter->stats_done);
4487                 break;
4488         case QUERY_IP_OFFLOAD_RSP:
4489                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4490                 handle_query_ip_offload_rsp(adapter);
4491                 break;
4492         case MULTICAST_CTRL_RSP:
4493                 netdev_dbg(netdev, "Got multicast control Response\n");
4494                 break;
4495         case CONTROL_IP_OFFLOAD_RSP:
4496                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4497                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4498                                  sizeof(adapter->ip_offload_ctrl),
4499                                  DMA_TO_DEVICE);
4500                 complete(&adapter->init_done);
4501                 break;
4502         case COLLECT_FW_TRACE_RSP:
4503                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4504                 complete(&adapter->fw_done);
4505                 break;
4506         case GET_VPD_SIZE_RSP:
4507                 handle_vpd_size_rsp(crq, adapter);
4508                 break;
4509         case GET_VPD_RSP:
4510                 handle_vpd_rsp(crq, adapter);
4511                 break;
4512         case QUERY_PHYS_PARMS_RSP:
4513                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4514                 complete(&adapter->fw_done);
4515                 break;
4516         default:
4517                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4518                            gen_crq->cmd);
4519         }
4520 }
4521
4522 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4523 {
4524         struct ibmvnic_adapter *adapter = instance;
4525
4526         tasklet_schedule(&adapter->tasklet);
4527         return IRQ_HANDLED;
4528 }
4529
4530 static void ibmvnic_tasklet(void *data)
4531 {
4532         struct ibmvnic_adapter *adapter = data;
4533         struct ibmvnic_crq_queue *queue = &adapter->crq;
4534         union ibmvnic_crq *crq;
4535         unsigned long flags;
4536         bool done = false;
4537
4538         spin_lock_irqsave(&queue->lock, flags);
4539         while (!done) {
4540                 /* Pull all the valid messages off the CRQ */
4541                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4542                         ibmvnic_handle_crq(crq, adapter);
4543                         crq->generic.first = 0;
4544                 }
4545
4546                 /* remain in tasklet until all
4547                  * capabilities responses are received
4548                  */
4549                 if (!adapter->wait_capability)
4550                         done = true;
4551         }
4552         /* if capabilities CRQ's were sent in this tasklet, the following
4553          * tasklet must wait until all responses are received
4554          */
4555         if (atomic_read(&adapter->running_cap_crqs) != 0)
4556                 adapter->wait_capability = true;
4557         spin_unlock_irqrestore(&queue->lock, flags);
4558 }
4559
4560 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4561 {
4562         struct vio_dev *vdev = adapter->vdev;
4563         int rc;
4564
4565         do {
4566                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4567         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4568
4569         if (rc)
4570                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4571
4572         return rc;
4573 }
4574
4575 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4576 {
4577         struct ibmvnic_crq_queue *crq = &adapter->crq;
4578         struct device *dev = &adapter->vdev->dev;
4579         struct vio_dev *vdev = adapter->vdev;
4580         int rc;
4581
4582         /* Close the CRQ */
4583         do {
4584                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4585         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4586
4587         /* Clean out the queue */
4588         memset(crq->msgs, 0, PAGE_SIZE);
4589         crq->cur = 0;
4590         crq->active = false;
4591
4592         /* And re-open it again */
4593         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4594                                 crq->msg_token, PAGE_SIZE);
4595
4596         if (rc == H_CLOSED)
4597                 /* Adapter is good, but other end is not ready */
4598                 dev_warn(dev, "Partner adapter not ready\n");
4599         else if (rc != 0)
4600                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4601
4602         return rc;
4603 }
4604
4605 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4606 {
4607         struct ibmvnic_crq_queue *crq = &adapter->crq;
4608         struct vio_dev *vdev = adapter->vdev;
4609         long rc;
4610
4611         if (!crq->msgs)
4612                 return;
4613
4614         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4615         free_irq(vdev->irq, adapter);
4616         tasklet_kill(&adapter->tasklet);
4617         do {
4618                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4619         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4620
4621         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4622                          DMA_BIDIRECTIONAL);
4623         free_page((unsigned long)crq->msgs);
4624         crq->msgs = NULL;
4625         crq->active = false;
4626 }
4627
4628 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4629 {
4630         struct ibmvnic_crq_queue *crq = &adapter->crq;
4631         struct device *dev = &adapter->vdev->dev;
4632         struct vio_dev *vdev = adapter->vdev;
4633         int rc, retrc = -ENOMEM;
4634
4635         if (crq->msgs)
4636                 return 0;
4637
4638         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4639         /* Should we allocate more than one page? */
4640
4641         if (!crq->msgs)
4642                 return -ENOMEM;
4643
4644         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4645         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4646                                         DMA_BIDIRECTIONAL);
4647         if (dma_mapping_error(dev, crq->msg_token))
4648                 goto map_failed;
4649
4650         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4651                                 crq->msg_token, PAGE_SIZE);
4652
4653         if (rc == H_RESOURCE)
4654                 /* maybe kexecing and resource is busy. try a reset */
4655                 rc = ibmvnic_reset_crq(adapter);
4656         retrc = rc;
4657
4658         if (rc == H_CLOSED) {
4659                 dev_warn(dev, "Partner adapter not ready\n");
4660         } else if (rc) {
4661                 dev_warn(dev, "Error %d opening adapter\n", rc);
4662                 goto reg_crq_failed;
4663         }
4664
4665         retrc = 0;
4666
4667         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4668                      (unsigned long)adapter);
4669
4670         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4671         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4672                  adapter->vdev->unit_address);
4673         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4674         if (rc) {
4675                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4676                         vdev->irq, rc);
4677                 goto req_irq_failed;
4678         }
4679
4680         rc = vio_enable_interrupts(vdev);
4681         if (rc) {
4682                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4683                 goto req_irq_failed;
4684         }
4685
4686         crq->cur = 0;
4687         spin_lock_init(&crq->lock);
4688
4689         return retrc;
4690
4691 req_irq_failed:
4692         tasklet_kill(&adapter->tasklet);
4693         do {
4694                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4695         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4696 reg_crq_failed:
4697         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4698 map_failed:
4699         free_page((unsigned long)crq->msgs);
4700         crq->msgs = NULL;
4701         return retrc;
4702 }
4703
4704 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4705 {
4706         struct device *dev = &adapter->vdev->dev;
4707         unsigned long timeout = msecs_to_jiffies(30000);
4708         u64 old_num_rx_queues, old_num_tx_queues;
4709         int rc;
4710
4711         adapter->from_passive_init = false;
4712
4713         old_num_rx_queues = adapter->req_rx_queues;
4714         old_num_tx_queues = adapter->req_tx_queues;
4715
4716         reinit_completion(&adapter->init_done);
4717         adapter->init_done_rc = 0;
4718         ibmvnic_send_crq_init(adapter);
4719         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4720                 dev_err(dev, "Initialization sequence timed out\n");
4721                 return -1;
4722         }
4723
4724         if (adapter->init_done_rc) {
4725                 release_crq_queue(adapter);
4726                 return adapter->init_done_rc;
4727         }
4728
4729         if (adapter->from_passive_init) {
4730                 adapter->state = VNIC_OPEN;
4731                 adapter->from_passive_init = false;
4732                 return -1;
4733         }
4734
4735         if (adapter->resetting && !adapter->wait_for_reset &&
4736             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4737                 if (adapter->req_rx_queues != old_num_rx_queues ||
4738                     adapter->req_tx_queues != old_num_tx_queues) {
4739                         release_sub_crqs(adapter, 0);
4740                         rc = init_sub_crqs(adapter);
4741                 } else {
4742                         rc = reset_sub_crq_queues(adapter);
4743                 }
4744         } else {
4745                 rc = init_sub_crqs(adapter);
4746         }
4747
4748         if (rc) {
4749                 dev_err(dev, "Initialization of sub crqs failed\n");
4750                 release_crq_queue(adapter);
4751                 return rc;
4752         }
4753
4754         rc = init_sub_crq_irqs(adapter);
4755         if (rc) {
4756                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4757                 release_crq_queue(adapter);
4758         }
4759
4760         return rc;
4761 }
4762
4763 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4764 {
4765         struct device *dev = &adapter->vdev->dev;
4766         unsigned long timeout = msecs_to_jiffies(30000);
4767         int rc;
4768
4769         adapter->from_passive_init = false;
4770
4771         adapter->init_done_rc = 0;
4772         ibmvnic_send_crq_init(adapter);
4773         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4774                 dev_err(dev, "Initialization sequence timed out\n");
4775                 return -1;
4776         }
4777
4778         if (adapter->init_done_rc) {
4779                 release_crq_queue(adapter);
4780                 return adapter->init_done_rc;
4781         }
4782
4783         if (adapter->from_passive_init) {
4784                 adapter->state = VNIC_OPEN;
4785                 adapter->from_passive_init = false;
4786                 return -1;
4787         }
4788
4789         rc = init_sub_crqs(adapter);
4790         if (rc) {
4791                 dev_err(dev, "Initialization of sub crqs failed\n");
4792                 release_crq_queue(adapter);
4793                 return rc;
4794         }
4795
4796         rc = init_sub_crq_irqs(adapter);
4797         if (rc) {
4798                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4799                 release_crq_queue(adapter);
4800         }
4801
4802         return rc;
4803 }
4804
4805 static struct device_attribute dev_attr_failover;
4806
4807 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4808 {
4809         struct ibmvnic_adapter *adapter;
4810         struct net_device *netdev;
4811         unsigned char *mac_addr_p;
4812         int rc;
4813
4814         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4815                 dev->unit_address);
4816
4817         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4818                                                         VETH_MAC_ADDR, NULL);
4819         if (!mac_addr_p) {
4820                 dev_err(&dev->dev,
4821                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4822                         __FILE__, __LINE__);
4823                 return 0;
4824         }
4825
4826         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4827                                    IBMVNIC_MAX_QUEUES);
4828         if (!netdev)
4829                 return -ENOMEM;
4830
4831         adapter = netdev_priv(netdev);
4832         adapter->state = VNIC_PROBING;
4833         dev_set_drvdata(&dev->dev, netdev);
4834         adapter->vdev = dev;
4835         adapter->netdev = netdev;
4836
4837         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4838         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4839         netdev->irq = dev->irq;
4840         netdev->netdev_ops = &ibmvnic_netdev_ops;
4841         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4842         SET_NETDEV_DEV(netdev, &dev->dev);
4843
4844         spin_lock_init(&adapter->stats_lock);
4845
4846         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4847         INIT_LIST_HEAD(&adapter->rwi_list);
4848         spin_lock_init(&adapter->rwi_lock);
4849         init_completion(&adapter->init_done);
4850         adapter->resetting = false;
4851
4852         do {
4853                 rc = init_crq_queue(adapter);
4854                 if (rc) {
4855                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4856                                 rc);
4857                         goto ibmvnic_init_fail;
4858                 }
4859
4860                 rc = ibmvnic_init(adapter);
4861                 if (rc && rc != EAGAIN)
4862                         goto ibmvnic_init_fail;
4863         } while (rc == EAGAIN);
4864
4865         rc = init_stats_buffers(adapter);
4866         if (rc)
4867                 goto ibmvnic_init_fail;
4868
4869         rc = init_stats_token(adapter);
4870         if (rc)
4871                 goto ibmvnic_stats_fail;
4872
4873         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4874         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4875         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4876
4877         rc = device_create_file(&dev->dev, &dev_attr_failover);
4878         if (rc)
4879                 goto ibmvnic_dev_file_err;
4880
4881         netif_carrier_off(netdev);
4882         rc = register_netdev(netdev);
4883         if (rc) {
4884                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4885                 goto ibmvnic_register_fail;
4886         }
4887         dev_info(&dev->dev, "ibmvnic registered\n");
4888
4889         adapter->state = VNIC_PROBED;
4890
4891         adapter->wait_for_reset = false;
4892
4893         return 0;
4894
4895 ibmvnic_register_fail:
4896         device_remove_file(&dev->dev, &dev_attr_failover);
4897
4898 ibmvnic_dev_file_err:
4899         release_stats_token(adapter);
4900
4901 ibmvnic_stats_fail:
4902         release_stats_buffers(adapter);
4903
4904 ibmvnic_init_fail:
4905         release_sub_crqs(adapter, 1);
4906         release_crq_queue(adapter);
4907         free_netdev(netdev);
4908
4909         return rc;
4910 }
4911
4912 static int ibmvnic_remove(struct vio_dev *dev)
4913 {
4914         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4915         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4916
4917         adapter->state = VNIC_REMOVING;
4918         rtnl_lock();
4919         unregister_netdevice(netdev);
4920
4921         release_resources(adapter);
4922         release_sub_crqs(adapter, 1);
4923         release_crq_queue(adapter);
4924
4925         release_stats_token(adapter);
4926         release_stats_buffers(adapter);
4927
4928         adapter->state = VNIC_REMOVED;
4929
4930         rtnl_unlock();
4931         device_remove_file(&dev->dev, &dev_attr_failover);
4932         free_netdev(netdev);
4933         dev_set_drvdata(&dev->dev, NULL);
4934
4935         return 0;
4936 }
4937
4938 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4939                               const char *buf, size_t count)
4940 {
4941         struct net_device *netdev = dev_get_drvdata(dev);
4942         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4943         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4944         __be64 session_token;
4945         long rc;
4946
4947         if (!sysfs_streq(buf, "1"))
4948                 return -EINVAL;
4949
4950         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4951                          H_GET_SESSION_TOKEN, 0, 0, 0);
4952         if (rc) {
4953                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4954                            rc);
4955                 return -EINVAL;
4956         }
4957
4958         session_token = (__be64)retbuf[0];
4959         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4960                    be64_to_cpu(session_token));
4961         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4962                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4963         if (rc) {
4964                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4965                            rc);
4966                 return -EINVAL;
4967         }
4968
4969         return count;
4970 }
4971
4972 static DEVICE_ATTR_WO(failover);
4973
4974 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4975 {
4976         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4977         struct ibmvnic_adapter *adapter;
4978         struct iommu_table *tbl;
4979         unsigned long ret = 0;
4980         int i;
4981
4982         tbl = get_iommu_table_base(&vdev->dev);
4983
4984         /* netdev inits at probe time along with the structures we need below*/
4985         if (!netdev)
4986                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4987
4988         adapter = netdev_priv(netdev);
4989
4990         ret += PAGE_SIZE; /* the crq message queue */
4991         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4992
4993         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4994                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4995
4996         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4997              i++)
4998                 ret += adapter->rx_pool[i].size *
4999                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5000
5001         return ret;
5002 }
5003
5004 static int ibmvnic_resume(struct device *dev)
5005 {
5006         struct net_device *netdev = dev_get_drvdata(dev);
5007         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5008
5009         if (adapter->state != VNIC_OPEN)
5010                 return 0;
5011
5012         tasklet_schedule(&adapter->tasklet);
5013
5014         return 0;
5015 }
5016
5017 static const struct vio_device_id ibmvnic_device_table[] = {
5018         {"network", "IBM,vnic"},
5019         {"", "" }
5020 };
5021 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5022
5023 static const struct dev_pm_ops ibmvnic_pm_ops = {
5024         .resume = ibmvnic_resume
5025 };
5026
5027 static struct vio_driver ibmvnic_driver = {
5028         .id_table       = ibmvnic_device_table,
5029         .probe          = ibmvnic_probe,
5030         .remove         = ibmvnic_remove,
5031         .get_desired_dma = ibmvnic_get_desired_dma,
5032         .name           = ibmvnic_driver_name,
5033         .pm             = &ibmvnic_pm_ops,
5034 };
5035
5036 /* module functions */
5037 static int __init ibmvnic_module_init(void)
5038 {
5039         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5040                 IBMVNIC_DRIVER_VERSION);
5041
5042         return vio_register_driver(&ibmvnic_driver);
5043 }
5044
5045 static void __exit ibmvnic_module_exit(void)
5046 {
5047         vio_unregister_driver(&ibmvnic_driver);
5048 }
5049
5050 module_init(ibmvnic_module_init);
5051 module_exit(ibmvnic_module_exit);