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 (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1749             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1750                 rc = __ibmvnic_close(netdev);
1751                 if (rc)
1752                         return rc;
1753         }
1754
1755         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1756             adapter->wait_for_reset) {
1757                 release_resources(adapter);
1758                 release_sub_crqs(adapter, 1);
1759                 release_crq_queue(adapter);
1760         }
1761
1762         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1763                 /* remove the closed state so when we call open it appears
1764                  * we are coming from the probed state.
1765                  */
1766                 adapter->state = VNIC_PROBED;
1767
1768                 if (adapter->wait_for_reset) {
1769                         rc = init_crq_queue(adapter);
1770                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1771                         rc = ibmvnic_reenable_crq_queue(adapter);
1772                         release_sub_crqs(adapter, 1);
1773                 } else {
1774                         rc = ibmvnic_reset_crq(adapter);
1775                         if (!rc)
1776                                 rc = vio_enable_interrupts(adapter->vdev);
1777                 }
1778
1779                 if (rc) {
1780                         netdev_err(adapter->netdev,
1781                                    "Couldn't initialize crq. rc=%d\n", rc);
1782                         return rc;
1783                 }
1784
1785                 rc = ibmvnic_reset_init(adapter);
1786                 if (rc)
1787                         return IBMVNIC_INIT_FAILED;
1788
1789                 /* If the adapter was in PROBE state prior to the reset,
1790                  * exit here.
1791                  */
1792                 if (reset_state == VNIC_PROBED)
1793                         return 0;
1794
1795                 rc = ibmvnic_login(netdev);
1796                 if (rc) {
1797                         adapter->state = reset_state;
1798                         return rc;
1799                 }
1800
1801                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1802                     adapter->wait_for_reset) {
1803                         rc = init_resources(adapter);
1804                         if (rc)
1805                                 return rc;
1806                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1807                            adapter->req_tx_queues != old_num_tx_queues ||
1808                            adapter->req_rx_add_entries_per_subcrq !=
1809                                                         old_num_rx_slots ||
1810                            adapter->req_tx_entries_per_subcrq !=
1811                                                         old_num_tx_slots) {
1812                         release_rx_pools(adapter);
1813                         release_tx_pools(adapter);
1814                         release_napi(adapter);
1815                         release_vpd_data(adapter);
1816
1817                         rc = init_resources(adapter);
1818                         if (rc)
1819                                 return rc;
1820
1821                 } else {
1822                         rc = reset_tx_pools(adapter);
1823                         if (rc)
1824                                 return rc;
1825
1826                         rc = reset_rx_pools(adapter);
1827                         if (rc)
1828                                 return rc;
1829                 }
1830                 ibmvnic_disable_irqs(adapter);
1831         }
1832         adapter->state = VNIC_CLOSED;
1833
1834         if (reset_state == VNIC_CLOSED)
1835                 return 0;
1836
1837         rc = __ibmvnic_open(netdev);
1838         if (rc) {
1839                 if (list_empty(&adapter->rwi_list))
1840                         adapter->state = VNIC_CLOSED;
1841                 else
1842                         adapter->state = reset_state;
1843
1844                 return 0;
1845         }
1846
1847         /* kick napi */
1848         for (i = 0; i < adapter->req_rx_queues; i++)
1849                 napi_schedule(&adapter->napi[i]);
1850
1851         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1852             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1853                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1854
1855         return 0;
1856 }
1857
1858 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1859                          struct ibmvnic_rwi *rwi, u32 reset_state)
1860 {
1861         struct net_device *netdev = adapter->netdev;
1862         int rc;
1863
1864         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1865                    rwi->reset_reason);
1866
1867         netif_carrier_off(netdev);
1868         adapter->reset_reason = rwi->reset_reason;
1869
1870         ibmvnic_cleanup(netdev);
1871         release_resources(adapter);
1872         release_sub_crqs(adapter, 0);
1873         release_crq_queue(adapter);
1874
1875         /* remove the closed state so when we call open it appears
1876          * we are coming from the probed state.
1877          */
1878         adapter->state = VNIC_PROBED;
1879
1880         reinit_completion(&adapter->init_done);
1881         rc = init_crq_queue(adapter);
1882         if (rc) {
1883                 netdev_err(adapter->netdev,
1884                            "Couldn't initialize crq. rc=%d\n", rc);
1885                 return rc;
1886         }
1887
1888         rc = ibmvnic_init(adapter);
1889         if (rc)
1890                 return rc;
1891
1892         /* If the adapter was in PROBE state prior to the reset,
1893          * exit here.
1894          */
1895         if (reset_state == VNIC_PROBED)
1896                 return 0;
1897
1898         rc = ibmvnic_login(netdev);
1899         if (rc) {
1900                 adapter->state = VNIC_PROBED;
1901                 return 0;
1902         }
1903
1904         rc = init_resources(adapter);
1905         if (rc)
1906                 return rc;
1907
1908         ibmvnic_disable_irqs(adapter);
1909         adapter->state = VNIC_CLOSED;
1910
1911         if (reset_state == VNIC_CLOSED)
1912                 return 0;
1913
1914         rc = __ibmvnic_open(netdev);
1915         if (rc) {
1916                 if (list_empty(&adapter->rwi_list))
1917                         adapter->state = VNIC_CLOSED;
1918                 else
1919                         adapter->state = reset_state;
1920
1921                 return 0;
1922         }
1923
1924         return 0;
1925 }
1926
1927 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1928 {
1929         struct ibmvnic_rwi *rwi;
1930         unsigned long flags;
1931
1932         spin_lock_irqsave(&adapter->rwi_lock, flags);
1933
1934         if (!list_empty(&adapter->rwi_list)) {
1935                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1936                                        list);
1937                 list_del(&rwi->list);
1938         } else {
1939                 rwi = NULL;
1940         }
1941
1942         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1943         return rwi;
1944 }
1945
1946 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1947 {
1948         struct ibmvnic_rwi *rwi;
1949
1950         rwi = get_next_rwi(adapter);
1951         while (rwi) {
1952                 kfree(rwi);
1953                 rwi = get_next_rwi(adapter);
1954         }
1955 }
1956
1957 static void __ibmvnic_reset(struct work_struct *work)
1958 {
1959         struct ibmvnic_rwi *rwi;
1960         struct ibmvnic_adapter *adapter;
1961         bool we_lock_rtnl = false;
1962         u32 reset_state;
1963         int rc = 0;
1964
1965         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1966
1967         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1968          * unless wait_for_reset is set, in which case the rtnl lock
1969          * has already been taken before initializing the reset
1970          */
1971         if (!adapter->wait_for_reset) {
1972                 rtnl_lock();
1973                 we_lock_rtnl = true;
1974         }
1975         reset_state = adapter->state;
1976
1977         rwi = get_next_rwi(adapter);
1978         while (rwi) {
1979                 if (adapter->force_reset_recovery) {
1980                         adapter->force_reset_recovery = false;
1981                         rc = do_hard_reset(adapter, rwi, reset_state);
1982                 } else {
1983                         rc = do_reset(adapter, rwi, reset_state);
1984                 }
1985                 kfree(rwi);
1986                 if (rc && rc != IBMVNIC_INIT_FAILED &&
1987                     !adapter->force_reset_recovery)
1988                         break;
1989
1990                 rwi = get_next_rwi(adapter);
1991         }
1992
1993         if (adapter->wait_for_reset) {
1994                 adapter->wait_for_reset = false;
1995                 adapter->reset_done_rc = rc;
1996                 complete(&adapter->reset_done);
1997         }
1998
1999         if (rc) {
2000                 netdev_dbg(adapter->netdev, "Reset failed\n");
2001                 free_all_rwi(adapter);
2002         }
2003
2004         adapter->resetting = false;
2005         if (we_lock_rtnl)
2006                 rtnl_unlock();
2007 }
2008
2009 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2010                          enum ibmvnic_reset_reason reason)
2011 {
2012         struct list_head *entry, *tmp_entry;
2013         struct ibmvnic_rwi *rwi, *tmp;
2014         struct net_device *netdev = adapter->netdev;
2015         unsigned long flags;
2016         int ret;
2017
2018         if (adapter->state == VNIC_REMOVING ||
2019             adapter->state == VNIC_REMOVED ||
2020             adapter->failover_pending) {
2021                 ret = EBUSY;
2022                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2023                 goto err;
2024         }
2025
2026         if (adapter->state == VNIC_PROBING) {
2027                 netdev_warn(netdev, "Adapter reset during probe\n");
2028                 ret = adapter->init_done_rc = EAGAIN;
2029                 goto err;
2030         }
2031
2032         spin_lock_irqsave(&adapter->rwi_lock, flags);
2033
2034         list_for_each(entry, &adapter->rwi_list) {
2035                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2036                 if (tmp->reset_reason == reason) {
2037                         netdev_dbg(netdev, "Skipping matching reset\n");
2038                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2039                         ret = EBUSY;
2040                         goto err;
2041                 }
2042         }
2043
2044         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2045         if (!rwi) {
2046                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2047                 ibmvnic_close(netdev);
2048                 ret = ENOMEM;
2049                 goto err;
2050         }
2051         /* if we just received a transport event,
2052          * flush reset queue and process this reset
2053          */
2054         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2055                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2056                         list_del(entry);
2057         }
2058         rwi->reset_reason = reason;
2059         list_add_tail(&rwi->list, &adapter->rwi_list);
2060         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2061         adapter->resetting = true;
2062         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2063         schedule_work(&adapter->ibmvnic_reset);
2064
2065         return 0;
2066 err:
2067         if (adapter->wait_for_reset)
2068                 adapter->wait_for_reset = false;
2069         return -ret;
2070 }
2071
2072 static void ibmvnic_tx_timeout(struct net_device *dev)
2073 {
2074         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2075
2076         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2077 }
2078
2079 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2080                                   struct ibmvnic_rx_buff *rx_buff)
2081 {
2082         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2083
2084         rx_buff->skb = NULL;
2085
2086         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2087         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2088
2089         atomic_dec(&pool->available);
2090 }
2091
2092 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2093 {
2094         struct net_device *netdev = napi->dev;
2095         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2096         int scrq_num = (int)(napi - adapter->napi);
2097         int frames_processed = 0;
2098
2099 restart_poll:
2100         while (frames_processed < budget) {
2101                 struct sk_buff *skb;
2102                 struct ibmvnic_rx_buff *rx_buff;
2103                 union sub_crq *next;
2104                 u32 length;
2105                 u16 offset;
2106                 u8 flags = 0;
2107
2108                 if (unlikely(adapter->resetting &&
2109                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2110                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2111                         napi_complete_done(napi, frames_processed);
2112                         return frames_processed;
2113                 }
2114
2115                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2116                         break;
2117                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2118                 rx_buff =
2119                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2120                                                           rx_comp.correlator);
2121                 /* do error checking */
2122                 if (next->rx_comp.rc) {
2123                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2124                                    be16_to_cpu(next->rx_comp.rc));
2125                         /* free the entry */
2126                         next->rx_comp.first = 0;
2127                         dev_kfree_skb_any(rx_buff->skb);
2128                         remove_buff_from_pool(adapter, rx_buff);
2129                         continue;
2130                 } else if (!rx_buff->skb) {
2131                         /* free the entry */
2132                         next->rx_comp.first = 0;
2133                         remove_buff_from_pool(adapter, rx_buff);
2134                         continue;
2135                 }
2136
2137                 length = be32_to_cpu(next->rx_comp.len);
2138                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2139                 flags = next->rx_comp.flags;
2140                 skb = rx_buff->skb;
2141                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2142                                         length);
2143
2144                 /* VLAN Header has been stripped by the system firmware and
2145                  * needs to be inserted by the driver
2146                  */
2147                 if (adapter->rx_vlan_header_insertion &&
2148                     (flags & IBMVNIC_VLAN_STRIPPED))
2149                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2150                                                ntohs(next->rx_comp.vlan_tci));
2151
2152                 /* free the entry */
2153                 next->rx_comp.first = 0;
2154                 remove_buff_from_pool(adapter, rx_buff);
2155
2156                 skb_put(skb, length);
2157                 skb->protocol = eth_type_trans(skb, netdev);
2158                 skb_record_rx_queue(skb, scrq_num);
2159
2160                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2161                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2162                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2163                 }
2164
2165                 length = skb->len;
2166                 napi_gro_receive(napi, skb); /* send it up */
2167                 netdev->stats.rx_packets++;
2168                 netdev->stats.rx_bytes += length;
2169                 adapter->rx_stats_buffers[scrq_num].packets++;
2170                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2171                 frames_processed++;
2172         }
2173
2174         if (adapter->state != VNIC_CLOSING)
2175                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2176
2177         if (frames_processed < budget) {
2178                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2179                 napi_complete_done(napi, frames_processed);
2180                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2181                     napi_reschedule(napi)) {
2182                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2183                         goto restart_poll;
2184                 }
2185         }
2186         return frames_processed;
2187 }
2188
2189 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2190 {
2191         int rc, ret;
2192
2193         adapter->fallback.mtu = adapter->req_mtu;
2194         adapter->fallback.rx_queues = adapter->req_rx_queues;
2195         adapter->fallback.tx_queues = adapter->req_tx_queues;
2196         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2197         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2198
2199         init_completion(&adapter->reset_done);
2200         adapter->wait_for_reset = true;
2201         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2202         if (rc)
2203                 return rc;
2204         wait_for_completion(&adapter->reset_done);
2205
2206         ret = 0;
2207         if (adapter->reset_done_rc) {
2208                 ret = -EIO;
2209                 adapter->desired.mtu = adapter->fallback.mtu;
2210                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2211                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2212                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2213                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2214
2215                 init_completion(&adapter->reset_done);
2216                 adapter->wait_for_reset = true;
2217                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2218                 if (rc)
2219                         return ret;
2220                 wait_for_completion(&adapter->reset_done);
2221         }
2222         adapter->wait_for_reset = false;
2223
2224         return ret;
2225 }
2226
2227 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2228 {
2229         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2230
2231         adapter->desired.mtu = new_mtu + ETH_HLEN;
2232
2233         return wait_for_reset(adapter);
2234 }
2235
2236 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2237                                                 struct net_device *dev,
2238                                                 netdev_features_t features)
2239 {
2240         /* Some backing hardware adapters can not
2241          * handle packets with a MSS less than 224
2242          * or with only one segment.
2243          */
2244         if (skb_is_gso(skb)) {
2245                 if (skb_shinfo(skb)->gso_size < 224 ||
2246                     skb_shinfo(skb)->gso_segs == 1)
2247                         features &= ~NETIF_F_GSO_MASK;
2248         }
2249
2250         return features;
2251 }
2252
2253 static const struct net_device_ops ibmvnic_netdev_ops = {
2254         .ndo_open               = ibmvnic_open,
2255         .ndo_stop               = ibmvnic_close,
2256         .ndo_start_xmit         = ibmvnic_xmit,
2257         .ndo_set_rx_mode        = ibmvnic_set_multi,
2258         .ndo_set_mac_address    = ibmvnic_set_mac,
2259         .ndo_validate_addr      = eth_validate_addr,
2260         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2261         .ndo_change_mtu         = ibmvnic_change_mtu,
2262         .ndo_features_check     = ibmvnic_features_check,
2263 };
2264
2265 /* ethtool functions */
2266
2267 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2268                                       struct ethtool_link_ksettings *cmd)
2269 {
2270         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2271         int rc;
2272
2273         rc = send_query_phys_parms(adapter);
2274         if (rc) {
2275                 adapter->speed = SPEED_UNKNOWN;
2276                 adapter->duplex = DUPLEX_UNKNOWN;
2277         }
2278         cmd->base.speed = adapter->speed;
2279         cmd->base.duplex = adapter->duplex;
2280         cmd->base.port = PORT_FIBRE;
2281         cmd->base.phy_address = 0;
2282         cmd->base.autoneg = AUTONEG_ENABLE;
2283
2284         return 0;
2285 }
2286
2287 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2288                                 struct ethtool_drvinfo *info)
2289 {
2290         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2291
2292         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2293         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2294         strlcpy(info->fw_version, adapter->fw_version,
2295                 sizeof(info->fw_version));
2296 }
2297
2298 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2299 {
2300         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2301
2302         return adapter->msg_enable;
2303 }
2304
2305 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2306 {
2307         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2308
2309         adapter->msg_enable = data;
2310 }
2311
2312 static u32 ibmvnic_get_link(struct net_device *netdev)
2313 {
2314         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2315
2316         /* Don't need to send a query because we request a logical link up at
2317          * init and then we wait for link state indications
2318          */
2319         return adapter->logical_link_state;
2320 }
2321
2322 static void ibmvnic_get_ringparam(struct net_device *netdev,
2323                                   struct ethtool_ringparam *ring)
2324 {
2325         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2326
2327         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2328                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2329                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2330         } else {
2331                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2332                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2333         }
2334         ring->rx_mini_max_pending = 0;
2335         ring->rx_jumbo_max_pending = 0;
2336         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2337         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2338         ring->rx_mini_pending = 0;
2339         ring->rx_jumbo_pending = 0;
2340 }
2341
2342 static int ibmvnic_set_ringparam(struct net_device *netdev,
2343                                  struct ethtool_ringparam *ring)
2344 {
2345         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2346         int ret;
2347
2348         ret = 0;
2349         adapter->desired.rx_entries = ring->rx_pending;
2350         adapter->desired.tx_entries = ring->tx_pending;
2351
2352         ret = wait_for_reset(adapter);
2353
2354         if (!ret &&
2355             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2356              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2357                 netdev_info(netdev,
2358                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2359                             ring->rx_pending, ring->tx_pending,
2360                             adapter->req_rx_add_entries_per_subcrq,
2361                             adapter->req_tx_entries_per_subcrq);
2362         return ret;
2363 }
2364
2365 static void ibmvnic_get_channels(struct net_device *netdev,
2366                                  struct ethtool_channels *channels)
2367 {
2368         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2369
2370         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2371                 channels->max_rx = adapter->max_rx_queues;
2372                 channels->max_tx = adapter->max_tx_queues;
2373         } else {
2374                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2375                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2376         }
2377
2378         channels->max_other = 0;
2379         channels->max_combined = 0;
2380         channels->rx_count = adapter->req_rx_queues;
2381         channels->tx_count = adapter->req_tx_queues;
2382         channels->other_count = 0;
2383         channels->combined_count = 0;
2384 }
2385
2386 static int ibmvnic_set_channels(struct net_device *netdev,
2387                                 struct ethtool_channels *channels)
2388 {
2389         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2390         int ret;
2391
2392         ret = 0;
2393         adapter->desired.rx_queues = channels->rx_count;
2394         adapter->desired.tx_queues = channels->tx_count;
2395
2396         ret = wait_for_reset(adapter);
2397
2398         if (!ret &&
2399             (adapter->req_rx_queues != channels->rx_count ||
2400              adapter->req_tx_queues != channels->tx_count))
2401                 netdev_info(netdev,
2402                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2403                             channels->rx_count, channels->tx_count,
2404                             adapter->req_rx_queues, adapter->req_tx_queues);
2405         return ret;
2406
2407 }
2408
2409 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2410 {
2411         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2412         int i;
2413
2414         switch (stringset) {
2415         case ETH_SS_STATS:
2416                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2417                                 i++, data += ETH_GSTRING_LEN)
2418                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2419
2420                 for (i = 0; i < adapter->req_tx_queues; i++) {
2421                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2422                         data += ETH_GSTRING_LEN;
2423
2424                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2425                         data += ETH_GSTRING_LEN;
2426
2427                         snprintf(data, ETH_GSTRING_LEN,
2428                                  "tx%d_dropped_packets", i);
2429                         data += ETH_GSTRING_LEN;
2430                 }
2431
2432                 for (i = 0; i < adapter->req_rx_queues; i++) {
2433                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2434                         data += ETH_GSTRING_LEN;
2435
2436                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2437                         data += ETH_GSTRING_LEN;
2438
2439                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2440                         data += ETH_GSTRING_LEN;
2441                 }
2442                 break;
2443
2444         case ETH_SS_PRIV_FLAGS:
2445                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2446                         strcpy(data + i * ETH_GSTRING_LEN,
2447                                ibmvnic_priv_flags[i]);
2448                 break;
2449         default:
2450                 return;
2451         }
2452 }
2453
2454 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2455 {
2456         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2457
2458         switch (sset) {
2459         case ETH_SS_STATS:
2460                 return ARRAY_SIZE(ibmvnic_stats) +
2461                        adapter->req_tx_queues * NUM_TX_STATS +
2462                        adapter->req_rx_queues * NUM_RX_STATS;
2463         case ETH_SS_PRIV_FLAGS:
2464                 return ARRAY_SIZE(ibmvnic_priv_flags);
2465         default:
2466                 return -EOPNOTSUPP;
2467         }
2468 }
2469
2470 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2471                                       struct ethtool_stats *stats, u64 *data)
2472 {
2473         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2474         union ibmvnic_crq crq;
2475         int i, j;
2476         int rc;
2477
2478         memset(&crq, 0, sizeof(crq));
2479         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2480         crq.request_statistics.cmd = REQUEST_STATISTICS;
2481         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2482         crq.request_statistics.len =
2483             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2484
2485         /* Wait for data to be written */
2486         init_completion(&adapter->stats_done);
2487         rc = ibmvnic_send_crq(adapter, &crq);
2488         if (rc)
2489                 return;
2490         wait_for_completion(&adapter->stats_done);
2491
2492         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2493                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2494                                                 ibmvnic_stats[i].offset));
2495
2496         for (j = 0; j < adapter->req_tx_queues; j++) {
2497                 data[i] = adapter->tx_stats_buffers[j].packets;
2498                 i++;
2499                 data[i] = adapter->tx_stats_buffers[j].bytes;
2500                 i++;
2501                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2502                 i++;
2503         }
2504
2505         for (j = 0; j < adapter->req_rx_queues; j++) {
2506                 data[i] = adapter->rx_stats_buffers[j].packets;
2507                 i++;
2508                 data[i] = adapter->rx_stats_buffers[j].bytes;
2509                 i++;
2510                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2511                 i++;
2512         }
2513 }
2514
2515 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2516 {
2517         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2518
2519         return adapter->priv_flags;
2520 }
2521
2522 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2523 {
2524         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2525         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2526
2527         if (which_maxes)
2528                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2529         else
2530                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2531
2532         return 0;
2533 }
2534 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2535         .get_drvinfo            = ibmvnic_get_drvinfo,
2536         .get_msglevel           = ibmvnic_get_msglevel,
2537         .set_msglevel           = ibmvnic_set_msglevel,
2538         .get_link               = ibmvnic_get_link,
2539         .get_ringparam          = ibmvnic_get_ringparam,
2540         .set_ringparam          = ibmvnic_set_ringparam,
2541         .get_channels           = ibmvnic_get_channels,
2542         .set_channels           = ibmvnic_set_channels,
2543         .get_strings            = ibmvnic_get_strings,
2544         .get_sset_count         = ibmvnic_get_sset_count,
2545         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2546         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2547         .get_priv_flags         = ibmvnic_get_priv_flags,
2548         .set_priv_flags         = ibmvnic_set_priv_flags,
2549 };
2550
2551 /* Routines for managing CRQs/sCRQs  */
2552
2553 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2554                                    struct ibmvnic_sub_crq_queue *scrq)
2555 {
2556         int rc;
2557
2558         if (scrq->irq) {
2559                 free_irq(scrq->irq, scrq);
2560                 irq_dispose_mapping(scrq->irq);
2561                 scrq->irq = 0;
2562         }
2563
2564         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2565         atomic_set(&scrq->used, 0);
2566         scrq->cur = 0;
2567
2568         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2569                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2570         return rc;
2571 }
2572
2573 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2574 {
2575         int i, rc;
2576
2577         for (i = 0; i < adapter->req_tx_queues; i++) {
2578                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2579                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2580                 if (rc)
2581                         return rc;
2582         }
2583
2584         for (i = 0; i < adapter->req_rx_queues; i++) {
2585                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2586                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2587                 if (rc)
2588                         return rc;
2589         }
2590
2591         return rc;
2592 }
2593
2594 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2595                                   struct ibmvnic_sub_crq_queue *scrq,
2596                                   bool do_h_free)
2597 {
2598         struct device *dev = &adapter->vdev->dev;
2599         long rc;
2600
2601         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2602
2603         if (do_h_free) {
2604                 /* Close the sub-crqs */
2605                 do {
2606                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2607                                                 adapter->vdev->unit_address,
2608                                                 scrq->crq_num);
2609                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2610
2611                 if (rc) {
2612                         netdev_err(adapter->netdev,
2613                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2614                                    scrq->crq_num, rc);
2615                 }
2616         }
2617
2618         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2619                          DMA_BIDIRECTIONAL);
2620         free_pages((unsigned long)scrq->msgs, 2);
2621         kfree(scrq);
2622 }
2623
2624 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2625                                                         *adapter)
2626 {
2627         struct device *dev = &adapter->vdev->dev;
2628         struct ibmvnic_sub_crq_queue *scrq;
2629         int rc;
2630
2631         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2632         if (!scrq)
2633                 return NULL;
2634
2635         scrq->msgs =
2636                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2637         if (!scrq->msgs) {
2638                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2639                 goto zero_page_failed;
2640         }
2641
2642         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2643                                          DMA_BIDIRECTIONAL);
2644         if (dma_mapping_error(dev, scrq->msg_token)) {
2645                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2646                 goto map_failed;
2647         }
2648
2649         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2650                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2651
2652         if (rc == H_RESOURCE)
2653                 rc = ibmvnic_reset_crq(adapter);
2654
2655         if (rc == H_CLOSED) {
2656                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2657         } else if (rc) {
2658                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2659                 goto reg_failed;
2660         }
2661
2662         scrq->adapter = adapter;
2663         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2664         spin_lock_init(&scrq->lock);
2665
2666         netdev_dbg(adapter->netdev,
2667                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2668                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2669
2670         return scrq;
2671
2672 reg_failed:
2673         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2674                          DMA_BIDIRECTIONAL);
2675 map_failed:
2676         free_pages((unsigned long)scrq->msgs, 2);
2677 zero_page_failed:
2678         kfree(scrq);
2679
2680         return NULL;
2681 }
2682
2683 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2684 {
2685         int i;
2686
2687         if (adapter->tx_scrq) {
2688                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2689                         if (!adapter->tx_scrq[i])
2690                                 continue;
2691
2692                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2693                                    i);
2694                         if (adapter->tx_scrq[i]->irq) {
2695                                 free_irq(adapter->tx_scrq[i]->irq,
2696                                          adapter->tx_scrq[i]);
2697                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2698                                 adapter->tx_scrq[i]->irq = 0;
2699                         }
2700
2701                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2702                                               do_h_free);
2703                 }
2704
2705                 kfree(adapter->tx_scrq);
2706                 adapter->tx_scrq = NULL;
2707                 adapter->num_active_tx_scrqs = 0;
2708         }
2709
2710         if (adapter->rx_scrq) {
2711                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2712                         if (!adapter->rx_scrq[i])
2713                                 continue;
2714
2715                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2716                                    i);
2717                         if (adapter->rx_scrq[i]->irq) {
2718                                 free_irq(adapter->rx_scrq[i]->irq,
2719                                          adapter->rx_scrq[i]);
2720                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2721                                 adapter->rx_scrq[i]->irq = 0;
2722                         }
2723
2724                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2725                                               do_h_free);
2726                 }
2727
2728                 kfree(adapter->rx_scrq);
2729                 adapter->rx_scrq = NULL;
2730                 adapter->num_active_rx_scrqs = 0;
2731         }
2732 }
2733
2734 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2735                             struct ibmvnic_sub_crq_queue *scrq)
2736 {
2737         struct device *dev = &adapter->vdev->dev;
2738         unsigned long rc;
2739
2740         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2741                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2742         if (rc)
2743                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2744                         scrq->hw_irq, rc);
2745         return rc;
2746 }
2747
2748 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2749                            struct ibmvnic_sub_crq_queue *scrq)
2750 {
2751         struct device *dev = &adapter->vdev->dev;
2752         unsigned long rc;
2753
2754         if (scrq->hw_irq > 0x100000000ULL) {
2755                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2756                 return 1;
2757         }
2758
2759         if (adapter->resetting &&
2760             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2761                 u64 val = (0xff000000) | scrq->hw_irq;
2762
2763                 rc = plpar_hcall_norets(H_EOI, val);
2764                 if (rc)
2765                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2766                                 val, rc);
2767         }
2768
2769         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2770                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2771         if (rc)
2772                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2773                         scrq->hw_irq, rc);
2774         return rc;
2775 }
2776
2777 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2778                                struct ibmvnic_sub_crq_queue *scrq)
2779 {
2780         struct device *dev = &adapter->vdev->dev;
2781         struct ibmvnic_tx_pool *tx_pool;
2782         struct ibmvnic_tx_buff *txbuff;
2783         union sub_crq *next;
2784         int index;
2785         int i, j;
2786         u8 *first;
2787
2788 restart_loop:
2789         while (pending_scrq(adapter, scrq)) {
2790                 unsigned int pool = scrq->pool_index;
2791                 int num_entries = 0;
2792
2793                 next = ibmvnic_next_scrq(adapter, scrq);
2794                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2795                         if (next->tx_comp.rcs[i]) {
2796                                 dev_err(dev, "tx error %x\n",
2797                                         next->tx_comp.rcs[i]);
2798                                 continue;
2799                         }
2800                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2801                         if (index & IBMVNIC_TSO_POOL_MASK) {
2802                                 tx_pool = &adapter->tso_pool[pool];
2803                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2804                         } else {
2805                                 tx_pool = &adapter->tx_pool[pool];
2806                         }
2807
2808                         txbuff = &tx_pool->tx_buff[index];
2809
2810                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2811                                 if (!txbuff->data_dma[j])
2812                                         continue;
2813
2814                                 txbuff->data_dma[j] = 0;
2815                         }
2816                         /* if sub_crq was sent indirectly */
2817                         first = &txbuff->indir_arr[0].generic.first;
2818                         if (*first == IBMVNIC_CRQ_CMD) {
2819                                 dma_unmap_single(dev, txbuff->indir_dma,
2820                                                  sizeof(txbuff->indir_arr),
2821                                                  DMA_TO_DEVICE);
2822                                 *first = 0;
2823                         }
2824
2825                         if (txbuff->last_frag) {
2826                                 dev_kfree_skb_any(txbuff->skb);
2827                                 txbuff->skb = NULL;
2828                         }
2829
2830                         num_entries += txbuff->num_entries;
2831
2832                         tx_pool->free_map[tx_pool->producer_index] = index;
2833                         tx_pool->producer_index =
2834                                 (tx_pool->producer_index + 1) %
2835                                         tx_pool->num_buffers;
2836                 }
2837                 /* remove tx_comp scrq*/
2838                 next->tx_comp.first = 0;
2839
2840                 if (atomic_sub_return(num_entries, &scrq->used) <=
2841                     (adapter->req_tx_entries_per_subcrq / 2) &&
2842                     __netif_subqueue_stopped(adapter->netdev,
2843                                              scrq->pool_index)) {
2844                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2845                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2846                                    scrq->pool_index);
2847                 }
2848         }
2849
2850         enable_scrq_irq(adapter, scrq);
2851
2852         if (pending_scrq(adapter, scrq)) {
2853                 disable_scrq_irq(adapter, scrq);
2854                 goto restart_loop;
2855         }
2856
2857         return 0;
2858 }
2859
2860 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2861 {
2862         struct ibmvnic_sub_crq_queue *scrq = instance;
2863         struct ibmvnic_adapter *adapter = scrq->adapter;
2864
2865         disable_scrq_irq(adapter, scrq);
2866         ibmvnic_complete_tx(adapter, scrq);
2867
2868         return IRQ_HANDLED;
2869 }
2870
2871 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2872 {
2873         struct ibmvnic_sub_crq_queue *scrq = instance;
2874         struct ibmvnic_adapter *adapter = scrq->adapter;
2875
2876         /* When booting a kdump kernel we can hit pending interrupts
2877          * prior to completing driver initialization.
2878          */
2879         if (unlikely(adapter->state != VNIC_OPEN))
2880                 return IRQ_NONE;
2881
2882         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2883
2884         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2885                 disable_scrq_irq(adapter, scrq);
2886                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2887         }
2888
2889         return IRQ_HANDLED;
2890 }
2891
2892 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2893 {
2894         struct device *dev = &adapter->vdev->dev;
2895         struct ibmvnic_sub_crq_queue *scrq;
2896         int i = 0, j = 0;
2897         int rc = 0;
2898
2899         for (i = 0; i < adapter->req_tx_queues; i++) {
2900                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2901                            i);
2902                 scrq = adapter->tx_scrq[i];
2903                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2904
2905                 if (!scrq->irq) {
2906                         rc = -EINVAL;
2907                         dev_err(dev, "Error mapping irq\n");
2908                         goto req_tx_irq_failed;
2909                 }
2910
2911                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
2912                          adapter->vdev->unit_address, i);
2913                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2914                                  0, scrq->name, scrq);
2915
2916                 if (rc) {
2917                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2918                                 scrq->irq, rc);
2919                         irq_dispose_mapping(scrq->irq);
2920                         goto req_tx_irq_failed;
2921                 }
2922         }
2923
2924         for (i = 0; i < adapter->req_rx_queues; i++) {
2925                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2926                            i);
2927                 scrq = adapter->rx_scrq[i];
2928                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2929                 if (!scrq->irq) {
2930                         rc = -EINVAL;
2931                         dev_err(dev, "Error mapping irq\n");
2932                         goto req_rx_irq_failed;
2933                 }
2934                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
2935                          adapter->vdev->unit_address, i);
2936                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2937                                  0, scrq->name, scrq);
2938                 if (rc) {
2939                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2940                                 scrq->irq, rc);
2941                         irq_dispose_mapping(scrq->irq);
2942                         goto req_rx_irq_failed;
2943                 }
2944         }
2945         return rc;
2946
2947 req_rx_irq_failed:
2948         for (j = 0; j < i; j++) {
2949                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2950                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2951         }
2952         i = adapter->req_tx_queues;
2953 req_tx_irq_failed:
2954         for (j = 0; j < i; j++) {
2955                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2956                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2957         }
2958         release_sub_crqs(adapter, 1);
2959         return rc;
2960 }
2961
2962 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2963 {
2964         struct device *dev = &adapter->vdev->dev;
2965         struct ibmvnic_sub_crq_queue **allqueues;
2966         int registered_queues = 0;
2967         int total_queues;
2968         int more = 0;
2969         int i;
2970
2971         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2972
2973         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2974         if (!allqueues)
2975                 return -1;
2976
2977         for (i = 0; i < total_queues; i++) {
2978                 allqueues[i] = init_sub_crq_queue(adapter);
2979                 if (!allqueues[i]) {
2980                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2981                         break;
2982                 }
2983                 registered_queues++;
2984         }
2985
2986         /* Make sure we were able to register the minimum number of queues */
2987         if (registered_queues <
2988             adapter->min_tx_queues + adapter->min_rx_queues) {
2989                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2990                 goto tx_failed;
2991         }
2992
2993         /* Distribute the failed allocated queues*/
2994         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2995                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2996                 switch (i % 3) {
2997                 case 0:
2998                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2999                                 adapter->req_rx_queues--;
3000                         else
3001                                 more++;
3002                         break;
3003                 case 1:
3004                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3005                                 adapter->req_tx_queues--;
3006                         else
3007                                 more++;
3008                         break;
3009                 }
3010         }
3011
3012         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3013                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3014         if (!adapter->tx_scrq)
3015                 goto tx_failed;
3016
3017         for (i = 0; i < adapter->req_tx_queues; i++) {
3018                 adapter->tx_scrq[i] = allqueues[i];
3019                 adapter->tx_scrq[i]->pool_index = i;
3020                 adapter->num_active_tx_scrqs++;
3021         }
3022
3023         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3024                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3025         if (!adapter->rx_scrq)
3026                 goto rx_failed;
3027
3028         for (i = 0; i < adapter->req_rx_queues; i++) {
3029                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3030                 adapter->rx_scrq[i]->scrq_num = i;
3031                 adapter->num_active_rx_scrqs++;
3032         }
3033
3034         kfree(allqueues);
3035         return 0;
3036
3037 rx_failed:
3038         kfree(adapter->tx_scrq);
3039         adapter->tx_scrq = NULL;
3040 tx_failed:
3041         for (i = 0; i < registered_queues; i++)
3042                 release_sub_crq_queue(adapter, allqueues[i], 1);
3043         kfree(allqueues);
3044         return -1;
3045 }
3046
3047 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3048 {
3049         struct device *dev = &adapter->vdev->dev;
3050         union ibmvnic_crq crq;
3051         int max_entries;
3052
3053         if (!retry) {
3054                 /* Sub-CRQ entries are 32 byte long */
3055                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3056
3057                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3058                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3059                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3060                         return;
3061                 }
3062
3063                 if (adapter->desired.mtu)
3064                         adapter->req_mtu = adapter->desired.mtu;
3065                 else
3066                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3067
3068                 if (!adapter->desired.tx_entries)
3069                         adapter->desired.tx_entries =
3070                                         adapter->max_tx_entries_per_subcrq;
3071                 if (!adapter->desired.rx_entries)
3072                         adapter->desired.rx_entries =
3073                                         adapter->max_rx_add_entries_per_subcrq;
3074
3075                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3076                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3077
3078                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3079                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3080                         adapter->desired.tx_entries = max_entries;
3081                 }
3082
3083                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3084                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3085                         adapter->desired.rx_entries = max_entries;
3086                 }
3087
3088                 if (adapter->desired.tx_entries)
3089                         adapter->req_tx_entries_per_subcrq =
3090                                         adapter->desired.tx_entries;
3091                 else
3092                         adapter->req_tx_entries_per_subcrq =
3093                                         adapter->max_tx_entries_per_subcrq;
3094
3095                 if (adapter->desired.rx_entries)
3096                         adapter->req_rx_add_entries_per_subcrq =
3097                                         adapter->desired.rx_entries;
3098                 else
3099                         adapter->req_rx_add_entries_per_subcrq =
3100                                         adapter->max_rx_add_entries_per_subcrq;
3101
3102                 if (adapter->desired.tx_queues)
3103                         adapter->req_tx_queues =
3104                                         adapter->desired.tx_queues;
3105                 else
3106                         adapter->req_tx_queues =
3107                                         adapter->opt_tx_comp_sub_queues;
3108
3109                 if (adapter->desired.rx_queues)
3110                         adapter->req_rx_queues =
3111                                         adapter->desired.rx_queues;
3112                 else
3113                         adapter->req_rx_queues =
3114                                         adapter->opt_rx_comp_queues;
3115
3116                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3117         }
3118
3119         memset(&crq, 0, sizeof(crq));
3120         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3121         crq.request_capability.cmd = REQUEST_CAPABILITY;
3122
3123         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3124         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3125         atomic_inc(&adapter->running_cap_crqs);
3126         ibmvnic_send_crq(adapter, &crq);
3127
3128         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3129         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3130         atomic_inc(&adapter->running_cap_crqs);
3131         ibmvnic_send_crq(adapter, &crq);
3132
3133         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3134         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3135         atomic_inc(&adapter->running_cap_crqs);
3136         ibmvnic_send_crq(adapter, &crq);
3137
3138         crq.request_capability.capability =
3139             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3140         crq.request_capability.number =
3141             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3142         atomic_inc(&adapter->running_cap_crqs);
3143         ibmvnic_send_crq(adapter, &crq);
3144
3145         crq.request_capability.capability =
3146             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3147         crq.request_capability.number =
3148             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3149         atomic_inc(&adapter->running_cap_crqs);
3150         ibmvnic_send_crq(adapter, &crq);
3151
3152         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3153         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3154         atomic_inc(&adapter->running_cap_crqs);
3155         ibmvnic_send_crq(adapter, &crq);
3156
3157         if (adapter->netdev->flags & IFF_PROMISC) {
3158                 if (adapter->promisc_supported) {
3159                         crq.request_capability.capability =
3160                             cpu_to_be16(PROMISC_REQUESTED);
3161                         crq.request_capability.number = cpu_to_be64(1);
3162                         atomic_inc(&adapter->running_cap_crqs);
3163                         ibmvnic_send_crq(adapter, &crq);
3164                 }
3165         } else {
3166                 crq.request_capability.capability =
3167                     cpu_to_be16(PROMISC_REQUESTED);
3168                 crq.request_capability.number = cpu_to_be64(0);
3169                 atomic_inc(&adapter->running_cap_crqs);
3170                 ibmvnic_send_crq(adapter, &crq);
3171         }
3172 }
3173
3174 static int pending_scrq(struct ibmvnic_adapter *adapter,
3175                         struct ibmvnic_sub_crq_queue *scrq)
3176 {
3177         union sub_crq *entry = &scrq->msgs[scrq->cur];
3178
3179         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3180                 return 1;
3181         else
3182                 return 0;
3183 }
3184
3185 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3186                                         struct ibmvnic_sub_crq_queue *scrq)
3187 {
3188         union sub_crq *entry;
3189         unsigned long flags;
3190
3191         spin_lock_irqsave(&scrq->lock, flags);
3192         entry = &scrq->msgs[scrq->cur];
3193         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3194                 if (++scrq->cur == scrq->size)
3195                         scrq->cur = 0;
3196         } else {
3197                 entry = NULL;
3198         }
3199         spin_unlock_irqrestore(&scrq->lock, flags);
3200
3201         return entry;
3202 }
3203
3204 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3205 {
3206         struct ibmvnic_crq_queue *queue = &adapter->crq;
3207         union ibmvnic_crq *crq;
3208
3209         crq = &queue->msgs[queue->cur];
3210         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3211                 if (++queue->cur == queue->size)
3212                         queue->cur = 0;
3213         } else {
3214                 crq = NULL;
3215         }
3216
3217         return crq;
3218 }
3219
3220 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3221 {
3222         switch (rc) {
3223         case H_PARAMETER:
3224                 dev_warn_ratelimited(dev,
3225                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3226                                      func, rc);
3227                 break;
3228         case H_CLOSED:
3229                 dev_warn_ratelimited(dev,
3230                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3231                                      func, rc);
3232                 break;
3233         default:
3234                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3235                 break;
3236         }
3237 }
3238
3239 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3240                        union sub_crq *sub_crq)
3241 {
3242         unsigned int ua = adapter->vdev->unit_address;
3243         struct device *dev = &adapter->vdev->dev;
3244         u64 *u64_crq = (u64 *)sub_crq;
3245         int rc;
3246
3247         netdev_dbg(adapter->netdev,
3248                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3249                    (unsigned long int)cpu_to_be64(remote_handle),
3250                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3251                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3252                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3253                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3254
3255         /* Make sure the hypervisor sees the complete request */
3256         mb();
3257
3258         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3259                                 cpu_to_be64(remote_handle),
3260                                 cpu_to_be64(u64_crq[0]),
3261                                 cpu_to_be64(u64_crq[1]),
3262                                 cpu_to_be64(u64_crq[2]),
3263                                 cpu_to_be64(u64_crq[3]));
3264
3265         if (rc)
3266                 print_subcrq_error(dev, rc, __func__);
3267
3268         return rc;
3269 }
3270
3271 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3272                                 u64 remote_handle, u64 ioba, u64 num_entries)
3273 {
3274         unsigned int ua = adapter->vdev->unit_address;
3275         struct device *dev = &adapter->vdev->dev;
3276         int rc;
3277
3278         /* Make sure the hypervisor sees the complete request */
3279         mb();
3280         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3281                                 cpu_to_be64(remote_handle),
3282                                 ioba, num_entries);
3283
3284         if (rc)
3285                 print_subcrq_error(dev, rc, __func__);
3286
3287         return rc;
3288 }
3289
3290 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3291                             union ibmvnic_crq *crq)
3292 {
3293         unsigned int ua = adapter->vdev->unit_address;
3294         struct device *dev = &adapter->vdev->dev;
3295         u64 *u64_crq = (u64 *)crq;
3296         int rc;
3297
3298         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3299                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3300                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3301
3302         if (!adapter->crq.active &&
3303             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3304                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3305                 return -EINVAL;
3306         }
3307
3308         /* Make sure the hypervisor sees the complete request */
3309         mb();
3310
3311         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3312                                 cpu_to_be64(u64_crq[0]),
3313                                 cpu_to_be64(u64_crq[1]));
3314
3315         if (rc) {
3316                 if (rc == H_CLOSED) {
3317                         dev_warn(dev, "CRQ Queue closed\n");
3318                         if (adapter->resetting)
3319                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3320                 }
3321
3322                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3323         }
3324
3325         return rc;
3326 }
3327
3328 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3329 {
3330         union ibmvnic_crq crq;
3331
3332         memset(&crq, 0, sizeof(crq));
3333         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3334         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3335         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3336
3337         return ibmvnic_send_crq(adapter, &crq);
3338 }
3339
3340 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3341 {
3342         union ibmvnic_crq crq;
3343
3344         memset(&crq, 0, sizeof(crq));
3345         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3346         crq.version_exchange.cmd = VERSION_EXCHANGE;
3347         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3348
3349         return ibmvnic_send_crq(adapter, &crq);
3350 }
3351
3352 struct vnic_login_client_data {
3353         u8      type;
3354         __be16  len;
3355         char    name[];
3356 } __packed;
3357
3358 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3359 {
3360         int len;
3361
3362         /* Calculate the amount of buffer space needed for the
3363          * vnic client data in the login buffer. There are four entries,
3364          * OS name, LPAR name, device name, and a null last entry.
3365          */
3366         len = 4 * sizeof(struct vnic_login_client_data);
3367         len += 6; /* "Linux" plus NULL */
3368         len += strlen(utsname()->nodename) + 1;
3369         len += strlen(adapter->netdev->name) + 1;
3370
3371         return len;
3372 }
3373
3374 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3375                                  struct vnic_login_client_data *vlcd)
3376 {
3377         const char *os_name = "Linux";
3378         int len;
3379
3380         /* Type 1 - LPAR OS */
3381         vlcd->type = 1;
3382         len = strlen(os_name) + 1;
3383         vlcd->len = cpu_to_be16(len);
3384         strncpy(vlcd->name, os_name, len);
3385         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3386
3387         /* Type 2 - LPAR name */
3388         vlcd->type = 2;
3389         len = strlen(utsname()->nodename) + 1;
3390         vlcd->len = cpu_to_be16(len);
3391         strncpy(vlcd->name, utsname()->nodename, len);
3392         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3393
3394         /* Type 3 - device name */
3395         vlcd->type = 3;
3396         len = strlen(adapter->netdev->name) + 1;
3397         vlcd->len = cpu_to_be16(len);
3398         strncpy(vlcd->name, adapter->netdev->name, len);
3399 }
3400
3401 static int send_login(struct ibmvnic_adapter *adapter)
3402 {
3403         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3404         struct ibmvnic_login_buffer *login_buffer;
3405         struct device *dev = &adapter->vdev->dev;
3406         dma_addr_t rsp_buffer_token;
3407         dma_addr_t buffer_token;
3408         size_t rsp_buffer_size;
3409         union ibmvnic_crq crq;
3410         size_t buffer_size;
3411         __be64 *tx_list_p;
3412         __be64 *rx_list_p;
3413         int client_data_len;
3414         struct vnic_login_client_data *vlcd;
3415         int i;
3416
3417         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3418                 netdev_err(adapter->netdev,
3419                            "RX or TX queues are not allocated, device login failed\n");
3420                 return -1;
3421         }
3422
3423         release_login_rsp_buffer(adapter);
3424         client_data_len = vnic_client_data_len(adapter);
3425
3426         buffer_size =
3427             sizeof(struct ibmvnic_login_buffer) +
3428             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3429             client_data_len;
3430
3431         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3432         if (!login_buffer)
3433                 goto buf_alloc_failed;
3434
3435         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3436                                       DMA_TO_DEVICE);
3437         if (dma_mapping_error(dev, buffer_token)) {
3438                 dev_err(dev, "Couldn't map login buffer\n");
3439                 goto buf_map_failed;
3440         }
3441
3442         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3443                           sizeof(u64) * adapter->req_tx_queues +
3444                           sizeof(u64) * adapter->req_rx_queues +
3445                           sizeof(u64) * adapter->req_rx_queues +
3446                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3447
3448         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3449         if (!login_rsp_buffer)
3450                 goto buf_rsp_alloc_failed;
3451
3452         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3453                                           rsp_buffer_size, DMA_FROM_DEVICE);
3454         if (dma_mapping_error(dev, rsp_buffer_token)) {
3455                 dev_err(dev, "Couldn't map login rsp buffer\n");
3456                 goto buf_rsp_map_failed;
3457         }
3458
3459         adapter->login_buf = login_buffer;
3460         adapter->login_buf_token = buffer_token;
3461         adapter->login_buf_sz = buffer_size;
3462         adapter->login_rsp_buf = login_rsp_buffer;
3463         adapter->login_rsp_buf_token = rsp_buffer_token;
3464         adapter->login_rsp_buf_sz = rsp_buffer_size;
3465
3466         login_buffer->len = cpu_to_be32(buffer_size);
3467         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3468         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3469         login_buffer->off_txcomp_subcrqs =
3470             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3471         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3472         login_buffer->off_rxcomp_subcrqs =
3473             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3474                         sizeof(u64) * adapter->req_tx_queues);
3475         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3476         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3477
3478         tx_list_p = (__be64 *)((char *)login_buffer +
3479                                       sizeof(struct ibmvnic_login_buffer));
3480         rx_list_p = (__be64 *)((char *)login_buffer +
3481                                       sizeof(struct ibmvnic_login_buffer) +
3482                                       sizeof(u64) * adapter->req_tx_queues);
3483
3484         for (i = 0; i < adapter->req_tx_queues; i++) {
3485                 if (adapter->tx_scrq[i]) {
3486                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3487                                                    crq_num);
3488                 }
3489         }
3490
3491         for (i = 0; i < adapter->req_rx_queues; i++) {
3492                 if (adapter->rx_scrq[i]) {
3493                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3494                                                    crq_num);
3495                 }
3496         }
3497
3498         /* Insert vNIC login client data */
3499         vlcd = (struct vnic_login_client_data *)
3500                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3501         login_buffer->client_data_offset =
3502                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3503         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3504
3505         vnic_add_client_data(adapter, vlcd);
3506
3507         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3508         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3509                 netdev_dbg(adapter->netdev, "%016lx\n",
3510                            ((unsigned long int *)(adapter->login_buf))[i]);
3511         }
3512
3513         memset(&crq, 0, sizeof(crq));
3514         crq.login.first = IBMVNIC_CRQ_CMD;
3515         crq.login.cmd = LOGIN;
3516         crq.login.ioba = cpu_to_be32(buffer_token);
3517         crq.login.len = cpu_to_be32(buffer_size);
3518         ibmvnic_send_crq(adapter, &crq);
3519
3520         return 0;
3521
3522 buf_rsp_map_failed:
3523         kfree(login_rsp_buffer);
3524 buf_rsp_alloc_failed:
3525         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3526 buf_map_failed:
3527         kfree(login_buffer);
3528 buf_alloc_failed:
3529         return -1;
3530 }
3531
3532 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3533                             u32 len, u8 map_id)
3534 {
3535         union ibmvnic_crq crq;
3536
3537         memset(&crq, 0, sizeof(crq));
3538         crq.request_map.first = IBMVNIC_CRQ_CMD;
3539         crq.request_map.cmd = REQUEST_MAP;
3540         crq.request_map.map_id = map_id;
3541         crq.request_map.ioba = cpu_to_be32(addr);
3542         crq.request_map.len = cpu_to_be32(len);
3543         return ibmvnic_send_crq(adapter, &crq);
3544 }
3545
3546 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3547 {
3548         union ibmvnic_crq crq;
3549
3550         memset(&crq, 0, sizeof(crq));
3551         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3552         crq.request_unmap.cmd = REQUEST_UNMAP;
3553         crq.request_unmap.map_id = map_id;
3554         return ibmvnic_send_crq(adapter, &crq);
3555 }
3556
3557 static void send_map_query(struct ibmvnic_adapter *adapter)
3558 {
3559         union ibmvnic_crq crq;
3560
3561         memset(&crq, 0, sizeof(crq));
3562         crq.query_map.first = IBMVNIC_CRQ_CMD;
3563         crq.query_map.cmd = QUERY_MAP;
3564         ibmvnic_send_crq(adapter, &crq);
3565 }
3566
3567 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3568 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3569 {
3570         union ibmvnic_crq crq;
3571
3572         atomic_set(&adapter->running_cap_crqs, 0);
3573         memset(&crq, 0, sizeof(crq));
3574         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3575         crq.query_capability.cmd = QUERY_CAPABILITY;
3576
3577         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3578         atomic_inc(&adapter->running_cap_crqs);
3579         ibmvnic_send_crq(adapter, &crq);
3580
3581         crq.query_capability.capability = cpu_to_be16(MIN_RX_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_ADD_QUEUES);
3586         atomic_inc(&adapter->running_cap_crqs);
3587         ibmvnic_send_crq(adapter, &crq);
3588
3589         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3590         atomic_inc(&adapter->running_cap_crqs);
3591         ibmvnic_send_crq(adapter, &crq);
3592
3593         crq.query_capability.capability = cpu_to_be16(MAX_RX_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_ADD_QUEUES);
3598         atomic_inc(&adapter->running_cap_crqs);
3599         ibmvnic_send_crq(adapter, &crq);
3600
3601         crq.query_capability.capability =
3602             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3603         atomic_inc(&adapter->running_cap_crqs);
3604         ibmvnic_send_crq(adapter, &crq);
3605
3606         crq.query_capability.capability =
3607             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3608         atomic_inc(&adapter->running_cap_crqs);
3609         ibmvnic_send_crq(adapter, &crq);
3610
3611         crq.query_capability.capability =
3612             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3613         atomic_inc(&adapter->running_cap_crqs);
3614         ibmvnic_send_crq(adapter, &crq);
3615
3616         crq.query_capability.capability =
3617             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3618         atomic_inc(&adapter->running_cap_crqs);
3619         ibmvnic_send_crq(adapter, &crq);
3620
3621         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3622         atomic_inc(&adapter->running_cap_crqs);
3623         ibmvnic_send_crq(adapter, &crq);
3624
3625         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3626         atomic_inc(&adapter->running_cap_crqs);
3627         ibmvnic_send_crq(adapter, &crq);
3628
3629         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3630         atomic_inc(&adapter->running_cap_crqs);
3631         ibmvnic_send_crq(adapter, &crq);
3632
3633         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3634         atomic_inc(&adapter->running_cap_crqs);
3635         ibmvnic_send_crq(adapter, &crq);
3636
3637         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3638         atomic_inc(&adapter->running_cap_crqs);
3639         ibmvnic_send_crq(adapter, &crq);
3640
3641         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3642         atomic_inc(&adapter->running_cap_crqs);
3643         ibmvnic_send_crq(adapter, &crq);
3644
3645         crq.query_capability.capability = cpu_to_be16(RX_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(MAX_TX_SG_ENTRIES);
3650         atomic_inc(&adapter->running_cap_crqs);
3651         ibmvnic_send_crq(adapter, &crq);
3652
3653         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3654         atomic_inc(&adapter->running_cap_crqs);
3655         ibmvnic_send_crq(adapter, &crq);
3656
3657         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3658         atomic_inc(&adapter->running_cap_crqs);
3659         ibmvnic_send_crq(adapter, &crq);
3660
3661         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3662         atomic_inc(&adapter->running_cap_crqs);
3663         ibmvnic_send_crq(adapter, &crq);
3664
3665         crq.query_capability.capability =
3666                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3667         atomic_inc(&adapter->running_cap_crqs);
3668         ibmvnic_send_crq(adapter, &crq);
3669
3670         crq.query_capability.capability =
3671                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3672         atomic_inc(&adapter->running_cap_crqs);
3673         ibmvnic_send_crq(adapter, &crq);
3674
3675         crq.query_capability.capability =
3676                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3677         atomic_inc(&adapter->running_cap_crqs);
3678         ibmvnic_send_crq(adapter, &crq);
3679
3680         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3681         atomic_inc(&adapter->running_cap_crqs);
3682         ibmvnic_send_crq(adapter, &crq);
3683 }
3684
3685 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3686                                 struct ibmvnic_adapter *adapter)
3687 {
3688         struct device *dev = &adapter->vdev->dev;
3689
3690         if (crq->get_vpd_size_rsp.rc.code) {
3691                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3692                         crq->get_vpd_size_rsp.rc.code);
3693                 complete(&adapter->fw_done);
3694                 return;
3695         }
3696
3697         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3698         complete(&adapter->fw_done);
3699 }
3700
3701 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3702                            struct ibmvnic_adapter *adapter)
3703 {
3704         struct device *dev = &adapter->vdev->dev;
3705         unsigned char *substr = NULL;
3706         u8 fw_level_len = 0;
3707
3708         memset(adapter->fw_version, 0, 32);
3709
3710         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3711                          DMA_FROM_DEVICE);
3712
3713         if (crq->get_vpd_rsp.rc.code) {
3714                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3715                         crq->get_vpd_rsp.rc.code);
3716                 goto complete;
3717         }
3718
3719         /* get the position of the firmware version info
3720          * located after the ASCII 'RM' substring in the buffer
3721          */
3722         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3723         if (!substr) {
3724                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3725                 goto complete;
3726         }
3727
3728         /* get length of firmware level ASCII substring */
3729         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3730                 fw_level_len = *(substr + 2);
3731         } else {
3732                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3733                 goto complete;
3734         }
3735
3736         /* copy firmware version string from vpd into adapter */
3737         if ((substr + 3 + fw_level_len) <
3738             (adapter->vpd->buff + adapter->vpd->len)) {
3739                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3740         } else {
3741                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3742         }
3743
3744 complete:
3745         if (adapter->fw_version[0] == '\0')
3746                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3747         complete(&adapter->fw_done);
3748 }
3749
3750 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3751 {
3752         struct device *dev = &adapter->vdev->dev;
3753         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3754         netdev_features_t old_hw_features = 0;
3755         union ibmvnic_crq crq;
3756         int i;
3757
3758         dma_unmap_single(dev, adapter->ip_offload_tok,
3759                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3760
3761         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3762         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3763                 netdev_dbg(adapter->netdev, "%016lx\n",
3764                            ((unsigned long int *)(buf))[i]);
3765
3766         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3767         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3768         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3769                    buf->tcp_ipv4_chksum);
3770         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3771                    buf->tcp_ipv6_chksum);
3772         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3773                    buf->udp_ipv4_chksum);
3774         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3775                    buf->udp_ipv6_chksum);
3776         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3777                    buf->large_tx_ipv4);
3778         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3779                    buf->large_tx_ipv6);
3780         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3781                    buf->large_rx_ipv4);
3782         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3783                    buf->large_rx_ipv6);
3784         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3785                    buf->max_ipv4_header_size);
3786         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3787                    buf->max_ipv6_header_size);
3788         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3789                    buf->max_tcp_header_size);
3790         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3791                    buf->max_udp_header_size);
3792         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3793                    buf->max_large_tx_size);
3794         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3795                    buf->max_large_rx_size);
3796         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3797                    buf->ipv6_extension_header);
3798         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3799                    buf->tcp_pseudosum_req);
3800         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3801                    buf->num_ipv6_ext_headers);
3802         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3803                    buf->off_ipv6_ext_headers);
3804
3805         adapter->ip_offload_ctrl_tok =
3806             dma_map_single(dev, &adapter->ip_offload_ctrl,
3807                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3808
3809         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3810                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3811                 return;
3812         }
3813
3814         adapter->ip_offload_ctrl.len =
3815             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3816         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3817         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3818         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3819         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3820         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3821         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3822         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3823         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3824         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3825
3826         /* large_rx disabled for now, additional features needed */
3827         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3828         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3829
3830         if (adapter->state != VNIC_PROBING) {
3831                 old_hw_features = adapter->netdev->hw_features;
3832                 adapter->netdev->hw_features = 0;
3833         }
3834
3835         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
3836
3837         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3838                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
3839
3840         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3841                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
3842
3843         if ((adapter->netdev->features &
3844             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3845                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
3846
3847         if (buf->large_tx_ipv4)
3848                 adapter->netdev->hw_features |= NETIF_F_TSO;
3849         if (buf->large_tx_ipv6)
3850                 adapter->netdev->hw_features |= NETIF_F_TSO6;
3851
3852         if (adapter->state == VNIC_PROBING) {
3853                 adapter->netdev->features |= adapter->netdev->hw_features;
3854         } else if (old_hw_features != adapter->netdev->hw_features) {
3855                 netdev_features_t tmp = 0;
3856
3857                 /* disable features no longer supported */
3858                 adapter->netdev->features &= adapter->netdev->hw_features;
3859                 /* turn on features now supported if previously enabled */
3860                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
3861                         adapter->netdev->hw_features;
3862                 adapter->netdev->features |=
3863                                 tmp & adapter->netdev->wanted_features;
3864         }
3865
3866         memset(&crq, 0, sizeof(crq));
3867         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3868         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3869         crq.control_ip_offload.len =
3870             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3871         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3872         ibmvnic_send_crq(adapter, &crq);
3873 }
3874
3875 static const char *ibmvnic_fw_err_cause(u16 cause)
3876 {
3877         switch (cause) {
3878         case ADAPTER_PROBLEM:
3879                 return "adapter problem";
3880         case BUS_PROBLEM:
3881                 return "bus problem";
3882         case FW_PROBLEM:
3883                 return "firmware problem";
3884         case DD_PROBLEM:
3885                 return "device driver problem";
3886         case EEH_RECOVERY:
3887                 return "EEH recovery";
3888         case FW_UPDATED:
3889                 return "firmware updated";
3890         case LOW_MEMORY:
3891                 return "low Memory";
3892         default:
3893                 return "unknown";
3894         }
3895 }
3896
3897 static void handle_error_indication(union ibmvnic_crq *crq,
3898                                     struct ibmvnic_adapter *adapter)
3899 {
3900         struct device *dev = &adapter->vdev->dev;
3901         u16 cause;
3902
3903         cause = be16_to_cpu(crq->error_indication.error_cause);
3904
3905         dev_warn_ratelimited(dev,
3906                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3907                              crq->error_indication.flags
3908                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3909                              ibmvnic_fw_err_cause(cause));
3910
3911         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3912                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3913         else
3914                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3915 }
3916
3917 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3918                                  struct ibmvnic_adapter *adapter)
3919 {
3920         struct net_device *netdev = adapter->netdev;
3921         struct device *dev = &adapter->vdev->dev;
3922         long rc;
3923
3924         rc = crq->change_mac_addr_rsp.rc.code;
3925         if (rc) {
3926                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3927                 goto out;
3928         }
3929         ether_addr_copy(netdev->dev_addr,
3930                         &crq->change_mac_addr_rsp.mac_addr[0]);
3931 out:
3932         complete(&adapter->fw_done);
3933         return rc;
3934 }
3935
3936 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3937                                    struct ibmvnic_adapter *adapter)
3938 {
3939         struct device *dev = &adapter->vdev->dev;
3940         u64 *req_value;
3941         char *name;
3942
3943         atomic_dec(&adapter->running_cap_crqs);
3944         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3945         case REQ_TX_QUEUES:
3946                 req_value = &adapter->req_tx_queues;
3947                 name = "tx";
3948                 break;
3949         case REQ_RX_QUEUES:
3950                 req_value = &adapter->req_rx_queues;
3951                 name = "rx";
3952                 break;
3953         case REQ_RX_ADD_QUEUES:
3954                 req_value = &adapter->req_rx_add_queues;
3955                 name = "rx_add";
3956                 break;
3957         case REQ_TX_ENTRIES_PER_SUBCRQ:
3958                 req_value = &adapter->req_tx_entries_per_subcrq;
3959                 name = "tx_entries_per_subcrq";
3960                 break;
3961         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3962                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3963                 name = "rx_add_entries_per_subcrq";
3964                 break;
3965         case REQ_MTU:
3966                 req_value = &adapter->req_mtu;
3967                 name = "mtu";
3968                 break;
3969         case PROMISC_REQUESTED:
3970                 req_value = &adapter->promisc;
3971                 name = "promisc";
3972                 break;
3973         default:
3974                 dev_err(dev, "Got invalid cap request rsp %d\n",
3975                         crq->request_capability.capability);
3976                 return;
3977         }
3978
3979         switch (crq->request_capability_rsp.rc.code) {
3980         case SUCCESS:
3981                 break;
3982         case PARTIALSUCCESS:
3983                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3984                          *req_value,
3985                          (long int)be64_to_cpu(crq->request_capability_rsp.
3986                                                number), name);
3987
3988                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3989                     REQ_MTU) {
3990                         pr_err("mtu of %llu is not supported. Reverting.\n",
3991                                *req_value);
3992                         *req_value = adapter->fallback.mtu;
3993                 } else {
3994                         *req_value =
3995                                 be64_to_cpu(crq->request_capability_rsp.number);
3996                 }
3997
3998                 ibmvnic_send_req_caps(adapter, 1);
3999                 return;
4000         default:
4001                 dev_err(dev, "Error %d in request cap rsp\n",
4002                         crq->request_capability_rsp.rc.code);
4003                 return;
4004         }
4005
4006         /* Done receiving requested capabilities, query IP offload support */
4007         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4008                 union ibmvnic_crq newcrq;
4009                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4010                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4011                     &adapter->ip_offload_buf;
4012
4013                 adapter->wait_capability = false;
4014                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4015                                                          buf_sz,
4016                                                          DMA_FROM_DEVICE);
4017
4018                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4019                         if (!firmware_has_feature(FW_FEATURE_CMO))
4020                                 dev_err(dev, "Couldn't map offload buffer\n");
4021                         return;
4022                 }
4023
4024                 memset(&newcrq, 0, sizeof(newcrq));
4025                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4026                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4027                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4028                 newcrq.query_ip_offload.ioba =
4029                     cpu_to_be32(adapter->ip_offload_tok);
4030
4031                 ibmvnic_send_crq(adapter, &newcrq);
4032         }
4033 }
4034
4035 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4036                             struct ibmvnic_adapter *adapter)
4037 {
4038         struct device *dev = &adapter->vdev->dev;
4039         struct net_device *netdev = adapter->netdev;
4040         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4041         struct ibmvnic_login_buffer *login = adapter->login_buf;
4042         int i;
4043
4044         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4045                          DMA_TO_DEVICE);
4046         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4047                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4048
4049         /* If the number of queues requested can't be allocated by the
4050          * server, the login response will return with code 1. We will need
4051          * to resend the login buffer with fewer queues requested.
4052          */
4053         if (login_rsp_crq->generic.rc.code) {
4054                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4055                 complete(&adapter->init_done);
4056                 return 0;
4057         }
4058
4059         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4060
4061         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4062         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4063                 netdev_dbg(adapter->netdev, "%016lx\n",
4064                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4065         }
4066
4067         /* Sanity checks */
4068         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4069             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4070              adapter->req_rx_add_queues !=
4071              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4072                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4073                 ibmvnic_remove(adapter->vdev);
4074                 return -EIO;
4075         }
4076         release_login_buffer(adapter);
4077         complete(&adapter->init_done);
4078
4079         return 0;
4080 }
4081
4082 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4083                                      struct ibmvnic_adapter *adapter)
4084 {
4085         struct device *dev = &adapter->vdev->dev;
4086         long rc;
4087
4088         rc = crq->request_unmap_rsp.rc.code;
4089         if (rc)
4090                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4091 }
4092
4093 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4094                                  struct ibmvnic_adapter *adapter)
4095 {
4096         struct net_device *netdev = adapter->netdev;
4097         struct device *dev = &adapter->vdev->dev;
4098         long rc;
4099
4100         rc = crq->query_map_rsp.rc.code;
4101         if (rc) {
4102                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4103                 return;
4104         }
4105         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4106                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4107                    crq->query_map_rsp.free_pages);
4108 }
4109
4110 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4111                                  struct ibmvnic_adapter *adapter)
4112 {
4113         struct net_device *netdev = adapter->netdev;
4114         struct device *dev = &adapter->vdev->dev;
4115         long rc;
4116
4117         atomic_dec(&adapter->running_cap_crqs);
4118         netdev_dbg(netdev, "Outstanding queries: %d\n",
4119                    atomic_read(&adapter->running_cap_crqs));
4120         rc = crq->query_capability.rc.code;
4121         if (rc) {
4122                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4123                 goto out;
4124         }
4125
4126         switch (be16_to_cpu(crq->query_capability.capability)) {
4127         case MIN_TX_QUEUES:
4128                 adapter->min_tx_queues =
4129                     be64_to_cpu(crq->query_capability.number);
4130                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4131                            adapter->min_tx_queues);
4132                 break;
4133         case MIN_RX_QUEUES:
4134                 adapter->min_rx_queues =
4135                     be64_to_cpu(crq->query_capability.number);
4136                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4137                            adapter->min_rx_queues);
4138                 break;
4139         case MIN_RX_ADD_QUEUES:
4140                 adapter->min_rx_add_queues =
4141                     be64_to_cpu(crq->query_capability.number);
4142                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4143                            adapter->min_rx_add_queues);
4144                 break;
4145         case MAX_TX_QUEUES:
4146                 adapter->max_tx_queues =
4147                     be64_to_cpu(crq->query_capability.number);
4148                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4149                            adapter->max_tx_queues);
4150                 break;
4151         case MAX_RX_QUEUES:
4152                 adapter->max_rx_queues =
4153                     be64_to_cpu(crq->query_capability.number);
4154                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4155                            adapter->max_rx_queues);
4156                 break;
4157         case MAX_RX_ADD_QUEUES:
4158                 adapter->max_rx_add_queues =
4159                     be64_to_cpu(crq->query_capability.number);
4160                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4161                            adapter->max_rx_add_queues);
4162                 break;
4163         case MIN_TX_ENTRIES_PER_SUBCRQ:
4164                 adapter->min_tx_entries_per_subcrq =
4165                     be64_to_cpu(crq->query_capability.number);
4166                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4167                            adapter->min_tx_entries_per_subcrq);
4168                 break;
4169         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4170                 adapter->min_rx_add_entries_per_subcrq =
4171                     be64_to_cpu(crq->query_capability.number);
4172                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4173                            adapter->min_rx_add_entries_per_subcrq);
4174                 break;
4175         case MAX_TX_ENTRIES_PER_SUBCRQ:
4176                 adapter->max_tx_entries_per_subcrq =
4177                     be64_to_cpu(crq->query_capability.number);
4178                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4179                            adapter->max_tx_entries_per_subcrq);
4180                 break;
4181         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4182                 adapter->max_rx_add_entries_per_subcrq =
4183                     be64_to_cpu(crq->query_capability.number);
4184                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4185                            adapter->max_rx_add_entries_per_subcrq);
4186                 break;
4187         case TCP_IP_OFFLOAD:
4188                 adapter->tcp_ip_offload =
4189                     be64_to_cpu(crq->query_capability.number);
4190                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4191                            adapter->tcp_ip_offload);
4192                 break;
4193         case PROMISC_SUPPORTED:
4194                 adapter->promisc_supported =
4195                     be64_to_cpu(crq->query_capability.number);
4196                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4197                            adapter->promisc_supported);
4198                 break;
4199         case MIN_MTU:
4200                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4201                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4202                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4203                 break;
4204         case MAX_MTU:
4205                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4206                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4207                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4208                 break;
4209         case MAX_MULTICAST_FILTERS:
4210                 adapter->max_multicast_filters =
4211                     be64_to_cpu(crq->query_capability.number);
4212                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4213                            adapter->max_multicast_filters);
4214                 break;
4215         case VLAN_HEADER_INSERTION:
4216                 adapter->vlan_header_insertion =
4217                     be64_to_cpu(crq->query_capability.number);
4218                 if (adapter->vlan_header_insertion)
4219                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4220                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4221                            adapter->vlan_header_insertion);
4222                 break;
4223         case RX_VLAN_HEADER_INSERTION:
4224                 adapter->rx_vlan_header_insertion =
4225                     be64_to_cpu(crq->query_capability.number);
4226                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4227                            adapter->rx_vlan_header_insertion);
4228                 break;
4229         case MAX_TX_SG_ENTRIES:
4230                 adapter->max_tx_sg_entries =
4231                     be64_to_cpu(crq->query_capability.number);
4232                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4233                            adapter->max_tx_sg_entries);
4234                 break;
4235         case RX_SG_SUPPORTED:
4236                 adapter->rx_sg_supported =
4237                     be64_to_cpu(crq->query_capability.number);
4238                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4239                            adapter->rx_sg_supported);
4240                 break;
4241         case OPT_TX_COMP_SUB_QUEUES:
4242                 adapter->opt_tx_comp_sub_queues =
4243                     be64_to_cpu(crq->query_capability.number);
4244                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4245                            adapter->opt_tx_comp_sub_queues);
4246                 break;
4247         case OPT_RX_COMP_QUEUES:
4248                 adapter->opt_rx_comp_queues =
4249                     be64_to_cpu(crq->query_capability.number);
4250                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4251                            adapter->opt_rx_comp_queues);
4252                 break;
4253         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4254                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4255                     be64_to_cpu(crq->query_capability.number);
4256                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4257                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4258                 break;
4259         case OPT_TX_ENTRIES_PER_SUBCRQ:
4260                 adapter->opt_tx_entries_per_subcrq =
4261                     be64_to_cpu(crq->query_capability.number);
4262                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4263                            adapter->opt_tx_entries_per_subcrq);
4264                 break;
4265         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4266                 adapter->opt_rxba_entries_per_subcrq =
4267                     be64_to_cpu(crq->query_capability.number);
4268                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4269                            adapter->opt_rxba_entries_per_subcrq);
4270                 break;
4271         case TX_RX_DESC_REQ:
4272                 adapter->tx_rx_desc_req = crq->query_capability.number;
4273                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4274                            adapter->tx_rx_desc_req);
4275                 break;
4276
4277         default:
4278                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4279                            crq->query_capability.capability);
4280         }
4281
4282 out:
4283         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4284                 adapter->wait_capability = false;
4285                 ibmvnic_send_req_caps(adapter, 0);
4286         }
4287 }
4288
4289 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4290 {
4291         union ibmvnic_crq crq;
4292         int rc;
4293
4294         memset(&crq, 0, sizeof(crq));
4295         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4296         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4297         init_completion(&adapter->fw_done);
4298         rc = ibmvnic_send_crq(adapter, &crq);
4299         if (rc)
4300                 return rc;
4301         wait_for_completion(&adapter->fw_done);
4302         return adapter->fw_done_rc ? -EIO : 0;
4303 }
4304
4305 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4306                                        struct ibmvnic_adapter *adapter)
4307 {
4308         struct net_device *netdev = adapter->netdev;
4309         int rc;
4310
4311         rc = crq->query_phys_parms_rsp.rc.code;
4312         if (rc) {
4313                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4314                 return rc;
4315         }
4316         switch (cpu_to_be32(crq->query_phys_parms_rsp.speed)) {
4317         case IBMVNIC_10MBPS:
4318                 adapter->speed = SPEED_10;
4319                 break;
4320         case IBMVNIC_100MBPS:
4321                 adapter->speed = SPEED_100;
4322                 break;
4323         case IBMVNIC_1GBPS:
4324                 adapter->speed = SPEED_1000;
4325                 break;
4326         case IBMVNIC_10GBP:
4327                 adapter->speed = SPEED_10000;
4328                 break;
4329         case IBMVNIC_25GBPS:
4330                 adapter->speed = SPEED_25000;
4331                 break;
4332         case IBMVNIC_40GBPS:
4333                 adapter->speed = SPEED_40000;
4334                 break;
4335         case IBMVNIC_50GBPS:
4336                 adapter->speed = SPEED_50000;
4337                 break;
4338         case IBMVNIC_100GBPS:
4339                 adapter->speed = SPEED_100000;
4340                 break;
4341         default:
4342                 netdev_warn(netdev, "Unknown speed 0x%08x\n",
4343                             cpu_to_be32(crq->query_phys_parms_rsp.speed));
4344                 adapter->speed = SPEED_UNKNOWN;
4345         }
4346         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4347                 adapter->duplex = DUPLEX_FULL;
4348         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4349                 adapter->duplex = DUPLEX_HALF;
4350         else
4351                 adapter->duplex = DUPLEX_UNKNOWN;
4352
4353         return rc;
4354 }
4355
4356 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4357                                struct ibmvnic_adapter *adapter)
4358 {
4359         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4360         struct net_device *netdev = adapter->netdev;
4361         struct device *dev = &adapter->vdev->dev;
4362         u64 *u64_crq = (u64 *)crq;
4363         long rc;
4364
4365         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4366                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4367                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4368         switch (gen_crq->first) {
4369         case IBMVNIC_CRQ_INIT_RSP:
4370                 switch (gen_crq->cmd) {
4371                 case IBMVNIC_CRQ_INIT:
4372                         dev_info(dev, "Partner initialized\n");
4373                         adapter->from_passive_init = true;
4374                         adapter->failover_pending = false;
4375                         if (!completion_done(&adapter->init_done)) {
4376                                 complete(&adapter->init_done);
4377                                 adapter->init_done_rc = -EIO;
4378                         }
4379                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4380                         break;
4381                 case IBMVNIC_CRQ_INIT_COMPLETE:
4382                         dev_info(dev, "Partner initialization complete\n");
4383                         adapter->crq.active = true;
4384                         send_version_xchg(adapter);
4385                         break;
4386                 default:
4387                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4388                 }
4389                 return;
4390         case IBMVNIC_CRQ_XPORT_EVENT:
4391                 netif_carrier_off(netdev);
4392                 adapter->crq.active = false;
4393                 if (adapter->resetting)
4394                         adapter->force_reset_recovery = true;
4395                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4396                         dev_info(dev, "Migrated, re-enabling adapter\n");
4397                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4398                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4399                         dev_info(dev, "Backing device failover detected\n");
4400                         adapter->failover_pending = true;
4401                 } else {
4402                         /* The adapter lost the connection */
4403                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4404                                 gen_crq->cmd);
4405                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4406                 }
4407                 return;
4408         case IBMVNIC_CRQ_CMD_RSP:
4409                 break;
4410         default:
4411                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4412                         gen_crq->first);
4413                 return;
4414         }
4415
4416         switch (gen_crq->cmd) {
4417         case VERSION_EXCHANGE_RSP:
4418                 rc = crq->version_exchange_rsp.rc.code;
4419                 if (rc) {
4420                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4421                         break;
4422                 }
4423                 dev_info(dev, "Partner protocol version is %d\n",
4424                          crq->version_exchange_rsp.version);
4425                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4426                     ibmvnic_version)
4427                         ibmvnic_version =
4428                             be16_to_cpu(crq->version_exchange_rsp.version);
4429                 send_cap_queries(adapter);
4430                 break;
4431         case QUERY_CAPABILITY_RSP:
4432                 handle_query_cap_rsp(crq, adapter);
4433                 break;
4434         case QUERY_MAP_RSP:
4435                 handle_query_map_rsp(crq, adapter);
4436                 break;
4437         case REQUEST_MAP_RSP:
4438                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4439                 complete(&adapter->fw_done);
4440                 break;
4441         case REQUEST_UNMAP_RSP:
4442                 handle_request_unmap_rsp(crq, adapter);
4443                 break;
4444         case REQUEST_CAPABILITY_RSP:
4445                 handle_request_cap_rsp(crq, adapter);
4446                 break;
4447         case LOGIN_RSP:
4448                 netdev_dbg(netdev, "Got Login Response\n");
4449                 handle_login_rsp(crq, adapter);
4450                 break;
4451         case LOGICAL_LINK_STATE_RSP:
4452                 netdev_dbg(netdev,
4453                            "Got Logical Link State Response, state: %d rc: %d\n",
4454                            crq->logical_link_state_rsp.link_state,
4455                            crq->logical_link_state_rsp.rc.code);
4456                 adapter->logical_link_state =
4457                     crq->logical_link_state_rsp.link_state;
4458                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4459                 complete(&adapter->init_done);
4460                 break;
4461         case LINK_STATE_INDICATION:
4462                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4463                 adapter->phys_link_state =
4464                     crq->link_state_indication.phys_link_state;
4465                 adapter->logical_link_state =
4466                     crq->link_state_indication.logical_link_state;
4467                 if (adapter->phys_link_state && adapter->logical_link_state)
4468                         netif_carrier_on(netdev);
4469                 else
4470                         netif_carrier_off(netdev);
4471                 break;
4472         case CHANGE_MAC_ADDR_RSP:
4473                 netdev_dbg(netdev, "Got MAC address change Response\n");
4474                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4475                 break;
4476         case ERROR_INDICATION:
4477                 netdev_dbg(netdev, "Got Error Indication\n");
4478                 handle_error_indication(crq, adapter);
4479                 break;
4480         case REQUEST_STATISTICS_RSP:
4481                 netdev_dbg(netdev, "Got Statistics Response\n");
4482                 complete(&adapter->stats_done);
4483                 break;
4484         case QUERY_IP_OFFLOAD_RSP:
4485                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4486                 handle_query_ip_offload_rsp(adapter);
4487                 break;
4488         case MULTICAST_CTRL_RSP:
4489                 netdev_dbg(netdev, "Got multicast control Response\n");
4490                 break;
4491         case CONTROL_IP_OFFLOAD_RSP:
4492                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4493                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4494                                  sizeof(adapter->ip_offload_ctrl),
4495                                  DMA_TO_DEVICE);
4496                 complete(&adapter->init_done);
4497                 break;
4498         case COLLECT_FW_TRACE_RSP:
4499                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4500                 complete(&adapter->fw_done);
4501                 break;
4502         case GET_VPD_SIZE_RSP:
4503                 handle_vpd_size_rsp(crq, adapter);
4504                 break;
4505         case GET_VPD_RSP:
4506                 handle_vpd_rsp(crq, adapter);
4507                 break;
4508         case QUERY_PHYS_PARMS_RSP:
4509                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4510                 complete(&adapter->fw_done);
4511                 break;
4512         default:
4513                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4514                            gen_crq->cmd);
4515         }
4516 }
4517
4518 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4519 {
4520         struct ibmvnic_adapter *adapter = instance;
4521
4522         tasklet_schedule(&adapter->tasklet);
4523         return IRQ_HANDLED;
4524 }
4525
4526 static void ibmvnic_tasklet(void *data)
4527 {
4528         struct ibmvnic_adapter *adapter = data;
4529         struct ibmvnic_crq_queue *queue = &adapter->crq;
4530         union ibmvnic_crq *crq;
4531         unsigned long flags;
4532         bool done = false;
4533
4534         spin_lock_irqsave(&queue->lock, flags);
4535         while (!done) {
4536                 /* Pull all the valid messages off the CRQ */
4537                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4538                         ibmvnic_handle_crq(crq, adapter);
4539                         crq->generic.first = 0;
4540                 }
4541
4542                 /* remain in tasklet until all
4543                  * capabilities responses are received
4544                  */
4545                 if (!adapter->wait_capability)
4546                         done = true;
4547         }
4548         /* if capabilities CRQ's were sent in this tasklet, the following
4549          * tasklet must wait until all responses are received
4550          */
4551         if (atomic_read(&adapter->running_cap_crqs) != 0)
4552                 adapter->wait_capability = true;
4553         spin_unlock_irqrestore(&queue->lock, flags);
4554 }
4555
4556 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4557 {
4558         struct vio_dev *vdev = adapter->vdev;
4559         int rc;
4560
4561         do {
4562                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4563         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4564
4565         if (rc)
4566                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4567
4568         return rc;
4569 }
4570
4571 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4572 {
4573         struct ibmvnic_crq_queue *crq = &adapter->crq;
4574         struct device *dev = &adapter->vdev->dev;
4575         struct vio_dev *vdev = adapter->vdev;
4576         int rc;
4577
4578         /* Close the CRQ */
4579         do {
4580                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4581         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4582
4583         /* Clean out the queue */
4584         memset(crq->msgs, 0, PAGE_SIZE);
4585         crq->cur = 0;
4586         crq->active = false;
4587
4588         /* And re-open it again */
4589         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4590                                 crq->msg_token, PAGE_SIZE);
4591
4592         if (rc == H_CLOSED)
4593                 /* Adapter is good, but other end is not ready */
4594                 dev_warn(dev, "Partner adapter not ready\n");
4595         else if (rc != 0)
4596                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4597
4598         return rc;
4599 }
4600
4601 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4602 {
4603         struct ibmvnic_crq_queue *crq = &adapter->crq;
4604         struct vio_dev *vdev = adapter->vdev;
4605         long rc;
4606
4607         if (!crq->msgs)
4608                 return;
4609
4610         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4611         free_irq(vdev->irq, adapter);
4612         tasklet_kill(&adapter->tasklet);
4613         do {
4614                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4615         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4616
4617         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4618                          DMA_BIDIRECTIONAL);
4619         free_page((unsigned long)crq->msgs);
4620         crq->msgs = NULL;
4621         crq->active = false;
4622 }
4623
4624 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4625 {
4626         struct ibmvnic_crq_queue *crq = &adapter->crq;
4627         struct device *dev = &adapter->vdev->dev;
4628         struct vio_dev *vdev = adapter->vdev;
4629         int rc, retrc = -ENOMEM;
4630
4631         if (crq->msgs)
4632                 return 0;
4633
4634         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4635         /* Should we allocate more than one page? */
4636
4637         if (!crq->msgs)
4638                 return -ENOMEM;
4639
4640         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4641         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4642                                         DMA_BIDIRECTIONAL);
4643         if (dma_mapping_error(dev, crq->msg_token))
4644                 goto map_failed;
4645
4646         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4647                                 crq->msg_token, PAGE_SIZE);
4648
4649         if (rc == H_RESOURCE)
4650                 /* maybe kexecing and resource is busy. try a reset */
4651                 rc = ibmvnic_reset_crq(adapter);
4652         retrc = rc;
4653
4654         if (rc == H_CLOSED) {
4655                 dev_warn(dev, "Partner adapter not ready\n");
4656         } else if (rc) {
4657                 dev_warn(dev, "Error %d opening adapter\n", rc);
4658                 goto reg_crq_failed;
4659         }
4660
4661         retrc = 0;
4662
4663         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4664                      (unsigned long)adapter);
4665
4666         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4667         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4668                  adapter->vdev->unit_address);
4669         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4670         if (rc) {
4671                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4672                         vdev->irq, rc);
4673                 goto req_irq_failed;
4674         }
4675
4676         rc = vio_enable_interrupts(vdev);
4677         if (rc) {
4678                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4679                 goto req_irq_failed;
4680         }
4681
4682         crq->cur = 0;
4683         spin_lock_init(&crq->lock);
4684
4685         return retrc;
4686
4687 req_irq_failed:
4688         tasklet_kill(&adapter->tasklet);
4689         do {
4690                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4691         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4692 reg_crq_failed:
4693         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4694 map_failed:
4695         free_page((unsigned long)crq->msgs);
4696         crq->msgs = NULL;
4697         return retrc;
4698 }
4699
4700 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4701 {
4702         struct device *dev = &adapter->vdev->dev;
4703         unsigned long timeout = msecs_to_jiffies(30000);
4704         u64 old_num_rx_queues, old_num_tx_queues;
4705         int rc;
4706
4707         adapter->from_passive_init = false;
4708
4709         old_num_rx_queues = adapter->req_rx_queues;
4710         old_num_tx_queues = adapter->req_tx_queues;
4711
4712         reinit_completion(&adapter->init_done);
4713         adapter->init_done_rc = 0;
4714         ibmvnic_send_crq_init(adapter);
4715         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4716                 dev_err(dev, "Initialization sequence timed out\n");
4717                 return -1;
4718         }
4719
4720         if (adapter->init_done_rc) {
4721                 release_crq_queue(adapter);
4722                 return adapter->init_done_rc;
4723         }
4724
4725         if (adapter->from_passive_init) {
4726                 adapter->state = VNIC_OPEN;
4727                 adapter->from_passive_init = false;
4728                 return -1;
4729         }
4730
4731         if (adapter->resetting && !adapter->wait_for_reset &&
4732             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4733                 if (adapter->req_rx_queues != old_num_rx_queues ||
4734                     adapter->req_tx_queues != old_num_tx_queues) {
4735                         release_sub_crqs(adapter, 0);
4736                         rc = init_sub_crqs(adapter);
4737                 } else {
4738                         rc = reset_sub_crq_queues(adapter);
4739                 }
4740         } else {
4741                 rc = init_sub_crqs(adapter);
4742         }
4743
4744         if (rc) {
4745                 dev_err(dev, "Initialization of sub crqs failed\n");
4746                 release_crq_queue(adapter);
4747                 return rc;
4748         }
4749
4750         rc = init_sub_crq_irqs(adapter);
4751         if (rc) {
4752                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4753                 release_crq_queue(adapter);
4754         }
4755
4756         return rc;
4757 }
4758
4759 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4760 {
4761         struct device *dev = &adapter->vdev->dev;
4762         unsigned long timeout = msecs_to_jiffies(30000);
4763         int rc;
4764
4765         adapter->from_passive_init = false;
4766
4767         adapter->init_done_rc = 0;
4768         ibmvnic_send_crq_init(adapter);
4769         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4770                 dev_err(dev, "Initialization sequence timed out\n");
4771                 return -1;
4772         }
4773
4774         if (adapter->init_done_rc) {
4775                 release_crq_queue(adapter);
4776                 return adapter->init_done_rc;
4777         }
4778
4779         if (adapter->from_passive_init) {
4780                 adapter->state = VNIC_OPEN;
4781                 adapter->from_passive_init = false;
4782                 return -1;
4783         }
4784
4785         rc = init_sub_crqs(adapter);
4786         if (rc) {
4787                 dev_err(dev, "Initialization of sub crqs failed\n");
4788                 release_crq_queue(adapter);
4789                 return rc;
4790         }
4791
4792         rc = init_sub_crq_irqs(adapter);
4793         if (rc) {
4794                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4795                 release_crq_queue(adapter);
4796         }
4797
4798         return rc;
4799 }
4800
4801 static struct device_attribute dev_attr_failover;
4802
4803 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4804 {
4805         struct ibmvnic_adapter *adapter;
4806         struct net_device *netdev;
4807         unsigned char *mac_addr_p;
4808         int rc;
4809
4810         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4811                 dev->unit_address);
4812
4813         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4814                                                         VETH_MAC_ADDR, NULL);
4815         if (!mac_addr_p) {
4816                 dev_err(&dev->dev,
4817                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4818                         __FILE__, __LINE__);
4819                 return 0;
4820         }
4821
4822         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4823                                    IBMVNIC_MAX_QUEUES);
4824         if (!netdev)
4825                 return -ENOMEM;
4826
4827         adapter = netdev_priv(netdev);
4828         adapter->state = VNIC_PROBING;
4829         dev_set_drvdata(&dev->dev, netdev);
4830         adapter->vdev = dev;
4831         adapter->netdev = netdev;
4832
4833         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4834         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4835         netdev->irq = dev->irq;
4836         netdev->netdev_ops = &ibmvnic_netdev_ops;
4837         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4838         SET_NETDEV_DEV(netdev, &dev->dev);
4839
4840         spin_lock_init(&adapter->stats_lock);
4841
4842         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4843         INIT_LIST_HEAD(&adapter->rwi_list);
4844         spin_lock_init(&adapter->rwi_lock);
4845         init_completion(&adapter->init_done);
4846         adapter->resetting = false;
4847
4848         do {
4849                 rc = init_crq_queue(adapter);
4850                 if (rc) {
4851                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4852                                 rc);
4853                         goto ibmvnic_init_fail;
4854                 }
4855
4856                 rc = ibmvnic_init(adapter);
4857                 if (rc && rc != EAGAIN)
4858                         goto ibmvnic_init_fail;
4859         } while (rc == EAGAIN);
4860
4861         rc = init_stats_buffers(adapter);
4862         if (rc)
4863                 goto ibmvnic_init_fail;
4864
4865         rc = init_stats_token(adapter);
4866         if (rc)
4867                 goto ibmvnic_stats_fail;
4868
4869         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4870         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4871         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4872
4873         rc = device_create_file(&dev->dev, &dev_attr_failover);
4874         if (rc)
4875                 goto ibmvnic_dev_file_err;
4876
4877         netif_carrier_off(netdev);
4878         rc = register_netdev(netdev);
4879         if (rc) {
4880                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4881                 goto ibmvnic_register_fail;
4882         }
4883         dev_info(&dev->dev, "ibmvnic registered\n");
4884
4885         adapter->state = VNIC_PROBED;
4886
4887         adapter->wait_for_reset = false;
4888
4889         return 0;
4890
4891 ibmvnic_register_fail:
4892         device_remove_file(&dev->dev, &dev_attr_failover);
4893
4894 ibmvnic_dev_file_err:
4895         release_stats_token(adapter);
4896
4897 ibmvnic_stats_fail:
4898         release_stats_buffers(adapter);
4899
4900 ibmvnic_init_fail:
4901         release_sub_crqs(adapter, 1);
4902         release_crq_queue(adapter);
4903         free_netdev(netdev);
4904
4905         return rc;
4906 }
4907
4908 static int ibmvnic_remove(struct vio_dev *dev)
4909 {
4910         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4911         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4912
4913         adapter->state = VNIC_REMOVING;
4914         rtnl_lock();
4915         unregister_netdevice(netdev);
4916
4917         release_resources(adapter);
4918         release_sub_crqs(adapter, 1);
4919         release_crq_queue(adapter);
4920
4921         release_stats_token(adapter);
4922         release_stats_buffers(adapter);
4923
4924         adapter->state = VNIC_REMOVED;
4925
4926         rtnl_unlock();
4927         device_remove_file(&dev->dev, &dev_attr_failover);
4928         free_netdev(netdev);
4929         dev_set_drvdata(&dev->dev, NULL);
4930
4931         return 0;
4932 }
4933
4934 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4935                               const char *buf, size_t count)
4936 {
4937         struct net_device *netdev = dev_get_drvdata(dev);
4938         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4939         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4940         __be64 session_token;
4941         long rc;
4942
4943         if (!sysfs_streq(buf, "1"))
4944                 return -EINVAL;
4945
4946         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4947                          H_GET_SESSION_TOKEN, 0, 0, 0);
4948         if (rc) {
4949                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4950                            rc);
4951                 return -EINVAL;
4952         }
4953
4954         session_token = (__be64)retbuf[0];
4955         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4956                    be64_to_cpu(session_token));
4957         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4958                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4959         if (rc) {
4960                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4961                            rc);
4962                 return -EINVAL;
4963         }
4964
4965         return count;
4966 }
4967
4968 static DEVICE_ATTR_WO(failover);
4969
4970 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4971 {
4972         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4973         struct ibmvnic_adapter *adapter;
4974         struct iommu_table *tbl;
4975         unsigned long ret = 0;
4976         int i;
4977
4978         tbl = get_iommu_table_base(&vdev->dev);
4979
4980         /* netdev inits at probe time along with the structures we need below*/
4981         if (!netdev)
4982                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4983
4984         adapter = netdev_priv(netdev);
4985
4986         ret += PAGE_SIZE; /* the crq message queue */
4987         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4988
4989         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4990                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4991
4992         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4993              i++)
4994                 ret += adapter->rx_pool[i].size *
4995                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4996
4997         return ret;
4998 }
4999
5000 static int ibmvnic_resume(struct device *dev)
5001 {
5002         struct net_device *netdev = dev_get_drvdata(dev);
5003         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5004
5005         if (adapter->state != VNIC_OPEN)
5006                 return 0;
5007
5008         tasklet_schedule(&adapter->tasklet);
5009
5010         return 0;
5011 }
5012
5013 static const struct vio_device_id ibmvnic_device_table[] = {
5014         {"network", "IBM,vnic"},
5015         {"", "" }
5016 };
5017 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5018
5019 static const struct dev_pm_ops ibmvnic_pm_ops = {
5020         .resume = ibmvnic_resume
5021 };
5022
5023 static struct vio_driver ibmvnic_driver = {
5024         .id_table       = ibmvnic_device_table,
5025         .probe          = ibmvnic_probe,
5026         .remove         = ibmvnic_remove,
5027         .get_desired_dma = ibmvnic_get_desired_dma,
5028         .name           = ibmvnic_driver_name,
5029         .pm             = &ibmvnic_pm_ops,
5030 };
5031
5032 /* module functions */
5033 static int __init ibmvnic_module_init(void)
5034 {
5035         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5036                 IBMVNIC_DRIVER_VERSION);
5037
5038         return vio_register_driver(&ibmvnic_driver);
5039 }
5040
5041 static void __exit ibmvnic_module_exit(void)
5042 {
5043         vio_unregister_driver(&ibmvnic_driver);
5044 }
5045
5046 module_init(ibmvnic_module_init);
5047 module_exit(ibmvnic_module_exit);