drivers/net/ethernet/ibm/ibmvnic.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /**************************************************************************/
   3 /*                                                                        */
   4 /*  IBM System i and System p Virtual NIC Device Driver                   */
   5 /*  Copyright (C) 2014 IBM Corp.                                          */
   6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   9 /*                                                                        */
  10 /*                                                                        */
  11 /* This module contains the implementation of a virtual ethernet device   */
  12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  15 /*                                                                         */
  16 /* Messages are passed between the VNIC driver and the VNIC server using  */
  17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  18 /* issue and receive commands that initiate communication with the server */
  19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  20 /* are used by the driver to notify the server that a packet is           */
  21 /* ready for transmission or that a buffer has been added to receive a    */
  22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  23 /* driver that a packet transmission has been completed or that a packet  */
  24 /* has been received and placed in a waiting buffer.                      */
  25 /*                                                                        */
  26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  28 /* or receive has been completed, the VNIC driver is required to use      */
  29 /* "long term mapping". This entails that large, continuous DMA mapped    */
  30 /* buffers are allocated on driver initialization and these buffers are   */
  31 /* then continuously reused to pass skbs to and from the VNIC server.     */
  32 /*                                                                        */
  33 /**************************************************************************/
  34
  35 #include <linux/module.h>
  36 #include <linux/moduleparam.h>
  37 #include <linux/types.h>
  38 #include <linux/errno.h>
  39 #include <linux/completion.h>
  40 #include <linux/ioport.h>
  41 #include <linux/dma-mapping.h>
  42 #include <linux/kernel.h>
  43 #include <linux/netdevice.h>
  44 #include <linux/etherdevice.h>
  45 #include <linux/skbuff.h>
  46 #include <linux/init.h>
  47 #include <linux/delay.h>
  48 #include <linux/mm.h>
  49 #include <linux/ethtool.h>
  50 #include <linux/proc_fs.h>
  51 #include <linux/if_arp.h>
  52 #include <linux/in.h>
  53 #include <linux/ip.h>
  54 #include <linux/ipv6.h>
  55 #include <linux/irq.h>
  56 #include <linux/kthread.h>
  57 #include <linux/seq_file.h>
  58 #include <linux/interrupt.h>
  59 #include <net/net_namespace.h>
  60 #include <asm/hvcall.h>
  61 #include <linux/atomic.h>
  62 #include <asm/vio.h>
  63 #include <asm/iommu.h>
  64 #include <linux/uaccess.h>
  65 #include <asm/firmware.h>
  66 #include <linux/workqueue.h>
  67 #include <linux/if_vlan.h>
  68 #include <linux/utsname.h>
  69
  70 #include "ibmvnic.h"
  71
  72 static const char ibmvnic_driver_name[] = "ibmvnic";
  73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  74
  75 MODULE_AUTHOR("Santiago Leon");
  76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  77 MODULE_LICENSE("GPL");
  78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  79
  80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  81 static int ibmvnic_remove(struct vio_dev *);
  82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  88                        union sub_crq *sub_crq);
  89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
  90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
  91 static int enable_scrq_irq(struct ibmvnic_adapter *,
  92                            struct ibmvnic_sub_crq_queue *);
  93 static int disable_scrq_irq(struct ibmvnic_adapter *,
  94                             struct ibmvnic_sub_crq_queue *);
  95 static int pending_scrq(struct ibmvnic_adapter *,
  96                         struct ibmvnic_sub_crq_queue *);
  97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
  98                                         struct ibmvnic_sub_crq_queue *);
  99 static int ibmvnic_poll(struct napi_struct *napi, int data);
 100 static void send_map_query(struct ibmvnic_adapter *adapter);
 101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
 103 static int send_login(struct ibmvnic_adapter *adapter);
 104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 105 static int init_sub_crqs(struct ibmvnic_adapter *);
 106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 107 static int ibmvnic_init(struct ibmvnic_adapter *);
 108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
 109 static void release_crq_queue(struct ibmvnic_adapter *);
 110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
 111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
 113
 114 struct ibmvnic_stat {
 115         char name[ETH_GSTRING_LEN];
 116         int offset;
 117 };
 118
 119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 120                              offsetof(struct ibmvnic_statistics, stat))
 121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 122
 123 static const struct ibmvnic_stat ibmvnic_stats[] = {
 124         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 125         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 126         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 127         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 128         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 129         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 130         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 131         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 132         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 133         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 134         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 135         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 136         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 137         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 138         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 139         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 140         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 141         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 142         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 143         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 144         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 145         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 146 };
 147
 148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 149                           unsigned long length, unsigned long *number,
 150                           unsigned long *irq)
 151 {
 152         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 153         long rc;
 154
 155         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 156         *number = retbuf[0];
 157         *irq = retbuf[1];
 158
 159         return rc;
 160 }
 161
 162 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 163                                 struct ibmvnic_long_term_buff *ltb, int size)
 164 {
 165         struct device *dev = &adapter->vdev->dev;
 166         int rc;
 167
 168         ltb->size = size;
 169         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 170                                        GFP_KERNEL);
 171
 172         if (!ltb->buff) {
 173                 dev_err(dev, "Couldn't alloc long term buffer\n");
 174                 return -ENOMEM;
 175         }
 176         ltb->map_id = adapter->map_id;
 177         adapter->map_id++;
 178
 179         init_completion(&adapter->fw_done);
 180         rc = send_request_map(adapter, ltb->addr,
 181                               ltb->size, ltb->map_id);
 182         if (rc) {
 183                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 184                 return rc;
 185         }
 186         wait_for_completion(&adapter->fw_done);
 187
 188         if (adapter->fw_done_rc) {
 189                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 190                         adapter->fw_done_rc);
 191                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 192                 return -1;
 193         }
 194         return 0;
 195 }
 196
 197 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 198                                 struct ibmvnic_long_term_buff *ltb)
 199 {
 200         struct device *dev = &adapter->vdev->dev;
 201
 202         if (!ltb->buff)
 203                 return;
 204
 205         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 206             adapter->reset_reason != VNIC_RESET_MOBILITY)
 207                 send_request_unmap(adapter, ltb->map_id);
 208         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 209 }
 210
 211 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 212                                 struct ibmvnic_long_term_buff *ltb)
 213 {
 214         int rc;
 215
 216         memset(ltb->buff, 0, ltb->size);
 217
 218         init_completion(&adapter->fw_done);
 219         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 220         if (rc)
 221                 return rc;
 222         wait_for_completion(&adapter->fw_done);
 223
 224         if (adapter->fw_done_rc) {
 225                 dev_info(&adapter->vdev->dev,
 226                          "Reset failed, attempting to free and reallocate buffer\n");
 227                 free_long_term_buff(adapter, ltb);
 228                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 229         }
 230         return 0;
 231 }
 232
 233 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 234 {
 235         int i;
 236
 237         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 238              i++)
 239                 adapter->rx_pool[i].active = 0;
 240 }
 241
 242 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 243                               struct ibmvnic_rx_pool *pool)
 244 {
 245         int count = pool->size - atomic_read(&pool->available);
 246         struct device *dev = &adapter->vdev->dev;
 247         int buffers_added = 0;
 248         unsigned long lpar_rc;
 249         union sub_crq sub_crq;
 250         struct sk_buff *skb;
 251         unsigned int offset;
 252         dma_addr_t dma_addr;
 253         unsigned char *dst;
 254         u64 *handle_array;
 255         int shift = 0;
 256         int index;
 257         int i;
 258
 259         if (!pool->active)
 260                 return;
 261
 262         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 263                                       be32_to_cpu(adapter->login_rsp_buf->
 264                                       off_rxadd_subcrqs));
 265
 266         for (i = 0; i < count; ++i) {
 267                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 268                 if (!skb) {
 269                         dev_err(dev, "Couldn't replenish rx buff\n");
 270                         adapter->replenish_no_mem++;
 271                         break;
 272                 }
 273
 274                 index = pool->free_map[pool->next_free];
 275
 276                 if (pool->rx_buff[index].skb)
 277                         dev_err(dev, "Inconsistent free_map!\n");
 278
 279                 /* Copy the skb to the long term mapped DMA buffer */
 280                 offset = index * pool->buff_size;
 281                 dst = pool->long_term_buff.buff + offset;
 282                 memset(dst, 0, pool->buff_size);
 283                 dma_addr = pool->long_term_buff.addr + offset;
 284                 pool->rx_buff[index].data = dst;
 285
 286                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 287                 pool->rx_buff[index].dma = dma_addr;
 288                 pool->rx_buff[index].skb = skb;
 289                 pool->rx_buff[index].pool_index = pool->index;
 290                 pool->rx_buff[index].size = pool->buff_size;
 291
 292                 memset(&sub_crq, 0, sizeof(sub_crq));
 293                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 294                 sub_crq.rx_add.correlator =
 295                     cpu_to_be64((u64)&pool->rx_buff[index]);
 296                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 297                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 298
 299                 /* The length field of the sCRQ is defined to be 24 bits so the
 300                  * buffer size needs to be left shifted by a byte before it is
 301                  * converted to big endian to prevent the last byte from being
 302                  * truncated.
 303                  */
 304 #ifdef __LITTLE_ENDIAN__
 305                 shift = 8;
 306 #endif
 307                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 308
 309                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 310                                       &sub_crq);
 311                 if (lpar_rc != H_SUCCESS)
 312                         goto failure;
 313
 314                 buffers_added++;
 315                 adapter->replenish_add_buff_success++;
 316                 pool->next_free = (pool->next_free + 1) % pool->size;
 317         }
 318         atomic_add(buffers_added, &pool->available);
 319         return;
 320
 321 failure:
 322         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 323                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 324         pool->free_map[pool->next_free] = index;
 325         pool->rx_buff[index].skb = NULL;
 326
 327         dev_kfree_skb_any(skb);
 328         adapter->replenish_add_buff_failure++;
 329         atomic_add(buffers_added, &pool->available);
 330
 331         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 332                 /* Disable buffer pool replenishment and report carrier off if
 333                  * queue is closed or pending failover.
 334                  * Firmware guarantees that a signal will be sent to the
 335                  * driver, triggering a reset.
 336                  */
 337                 deactivate_rx_pools(adapter);
 338                 netif_carrier_off(adapter->netdev);
 339         }
 340 }
 341
 342 static void replenish_pools(struct ibmvnic_adapter *adapter)
 343 {
 344         int i;
 345
 346         adapter->replenish_task_cycles++;
 347         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 348              i++) {
 349                 if (adapter->rx_pool[i].active)
 350                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 351         }
 352 }
 353
 354 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 355 {
 356         kfree(adapter->tx_stats_buffers);
 357         kfree(adapter->rx_stats_buffers);
 358         adapter->tx_stats_buffers = NULL;
 359         adapter->rx_stats_buffers = NULL;
 360 }
 361
 362 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 363 {
 364         adapter->tx_stats_buffers =
 365                                 kcalloc(IBMVNIC_MAX_QUEUES,
 366                                         sizeof(struct ibmvnic_tx_queue_stats),
 367                                         GFP_KERNEL);
 368         if (!adapter->tx_stats_buffers)
 369                 return -ENOMEM;
 370
 371         adapter->rx_stats_buffers =
 372                                 kcalloc(IBMVNIC_MAX_QUEUES,
 373                                         sizeof(struct ibmvnic_rx_queue_stats),
 374                                         GFP_KERNEL);
 375         if (!adapter->rx_stats_buffers)
 376                 return -ENOMEM;
 377
 378         return 0;
 379 }
 380
 381 static void release_stats_token(struct ibmvnic_adapter *adapter)
 382 {
 383         struct device *dev = &adapter->vdev->dev;
 384
 385         if (!adapter->stats_token)
 386                 return;
 387
 388         dma_unmap_single(dev, adapter->stats_token,
 389                          sizeof(struct ibmvnic_statistics),
 390                          DMA_FROM_DEVICE);
 391         adapter->stats_token = 0;
 392 }
 393
 394 static int init_stats_token(struct ibmvnic_adapter *adapter)
 395 {
 396         struct device *dev = &adapter->vdev->dev;
 397         dma_addr_t stok;
 398
 399         stok = dma_map_single(dev, &adapter->stats,
 400                               sizeof(struct ibmvnic_statistics),
 401                               DMA_FROM_DEVICE);
 402         if (dma_mapping_error(dev, stok)) {
 403                 dev_err(dev, "Couldn't map stats buffer\n");
 404                 return -1;
 405         }
 406
 407         adapter->stats_token = stok;
 408         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 409         return 0;
 410 }
 411
 412 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 413 {
 414         struct ibmvnic_rx_pool *rx_pool;
 415         int rx_scrqs;
 416         int i, j, rc;
 417         u64 *size_array;
 418
 419         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 420                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 421
 422         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 423         for (i = 0; i < rx_scrqs; i++) {
 424                 rx_pool = &adapter->rx_pool[i];
 425
 426                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 427
 428                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 429                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 430                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 431                         rc = alloc_long_term_buff(adapter,
 432                                                   &rx_pool->long_term_buff,
 433                                                   rx_pool->size *
 434                                                   rx_pool->buff_size);
 435                 } else {
 436                         rc = reset_long_term_buff(adapter,
 437                                                   &rx_pool->long_term_buff);
 438                 }
 439
 440                 if (rc)
 441                         return rc;
 442
 443                 for (j = 0; j < rx_pool->size; j++)
 444                         rx_pool->free_map[j] = j;
 445
 446                 memset(rx_pool->rx_buff, 0,
 447                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 448
 449                 atomic_set(&rx_pool->available, 0);
 450                 rx_pool->next_alloc = 0;
 451                 rx_pool->next_free = 0;
 452                 rx_pool->active = 1;
 453         }
 454
 455         return 0;
 456 }
 457
 458 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 459 {
 460         struct ibmvnic_rx_pool *rx_pool;
 461         int i, j;
 462
 463         if (!adapter->rx_pool)
 464                 return;
 465
 466         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 467                 rx_pool = &adapter->rx_pool[i];
 468
 469                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 470
 471                 kfree(rx_pool->free_map);
 472                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 473
 474                 if (!rx_pool->rx_buff)
 475                         continue;
 476
 477                 for (j = 0; j < rx_pool->size; j++) {
 478                         if (rx_pool->rx_buff[j].skb) {
 479                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 480                                 rx_pool->rx_buff[j].skb = NULL;
 481                         }
 482                 }
 483
 484                 kfree(rx_pool->rx_buff);
 485         }
 486
 487         kfree(adapter->rx_pool);
 488         adapter->rx_pool = NULL;
 489         adapter->num_active_rx_pools = 0;
 490 }
 491
 492 static int init_rx_pools(struct net_device *netdev)
 493 {
 494         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 495         struct device *dev = &adapter->vdev->dev;
 496         struct ibmvnic_rx_pool *rx_pool;
 497         int rxadd_subcrqs;
 498         u64 *size_array;
 499         int i, j;
 500
 501         rxadd_subcrqs =
 502                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 503         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 504                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 505
 506         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 507                                    sizeof(struct ibmvnic_rx_pool),
 508                                    GFP_KERNEL);
 509         if (!adapter->rx_pool) {
 510                 dev_err(dev, "Failed to allocate rx pools\n");
 511                 return -1;
 512         }
 513
 514         adapter->num_active_rx_pools = rxadd_subcrqs;
 515
 516         for (i = 0; i < rxadd_subcrqs; i++) {
 517                 rx_pool = &adapter->rx_pool[i];
 518
 519                 netdev_dbg(adapter->netdev,
 520                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 521                            i, adapter->req_rx_add_entries_per_subcrq,
 522                            be64_to_cpu(size_array[i]));
 523
 524                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 525                 rx_pool->index = i;
 526                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 527                 rx_pool->active = 1;
 528
 529                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 530                                             GFP_KERNEL);
 531                 if (!rx_pool->free_map) {
 532                         release_rx_pools(adapter);
 533                         return -1;
 534                 }
 535
 536                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 537                                            sizeof(struct ibmvnic_rx_buff),
 538                                            GFP_KERNEL);
 539                 if (!rx_pool->rx_buff) {
 540                         dev_err(dev, "Couldn't alloc rx buffers\n");
 541                         release_rx_pools(adapter);
 542                         return -1;
 543                 }
 544
 545                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 546                                          rx_pool->size * rx_pool->buff_size)) {
 547                         release_rx_pools(adapter);
 548                         return -1;
 549                 }
 550
 551                 for (j = 0; j < rx_pool->size; ++j)
 552                         rx_pool->free_map[j] = j;
 553
 554                 atomic_set(&rx_pool->available, 0);
 555                 rx_pool->next_alloc = 0;
 556                 rx_pool->next_free = 0;
 557         }
 558
 559         return 0;
 560 }
 561
 562 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 563                              struct ibmvnic_tx_pool *tx_pool)
 564 {
 565         int rc, i;
 566
 567         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 568         if (rc)
 569                 return rc;
 570
 571         memset(tx_pool->tx_buff, 0,
 572                tx_pool->num_buffers *
 573                sizeof(struct ibmvnic_tx_buff));
 574
 575         for (i = 0; i < tx_pool->num_buffers; i++)
 576                 tx_pool->free_map[i] = i;
 577
 578         tx_pool->consumer_index = 0;
 579         tx_pool->producer_index = 0;
 580
 581         return 0;
 582 }
 583
 584 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 585 {
 586         int tx_scrqs;
 587         int i, rc;
 588
 589         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 590         for (i = 0; i < tx_scrqs; i++) {
 591                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 592                 if (rc)
 593                         return rc;
 594                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 595                 if (rc)
 596                         return rc;
 597         }
 598
 599         return 0;
 600 }
 601
 602 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 603 {
 604         if (!adapter->vpd)
 605                 return;
 606
 607         kfree(adapter->vpd->buff);
 608         kfree(adapter->vpd);
 609
 610         adapter->vpd = NULL;
 611 }
 612
 613 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 614                                 struct ibmvnic_tx_pool *tx_pool)
 615 {
 616         kfree(tx_pool->tx_buff);
 617         kfree(tx_pool->free_map);
 618         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 619 }
 620
 621 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 622 {
 623         int i;
 624
 625         if (!adapter->tx_pool)
 626                 return;
 627
 628         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 629                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 630                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 631         }
 632
 633         kfree(adapter->tx_pool);
 634         adapter->tx_pool = NULL;
 635         kfree(adapter->tso_pool);
 636         adapter->tso_pool = NULL;
 637         adapter->num_active_tx_pools = 0;
 638 }
 639
 640 static int init_one_tx_pool(struct net_device *netdev,
 641                             struct ibmvnic_tx_pool *tx_pool,
 642                             int num_entries, int buf_size)
 643 {
 644         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 645         int i;
 646
 647         tx_pool->tx_buff = kcalloc(num_entries,
 648                                    sizeof(struct ibmvnic_tx_buff),
 649                                    GFP_KERNEL);
 650         if (!tx_pool->tx_buff)
 651                 return -1;
 652
 653         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 654                                  num_entries * buf_size))
 655                 return -1;
 656
 657         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 658         if (!tx_pool->free_map)
 659                 return -1;
 660
 661         for (i = 0; i < num_entries; i++)
 662                 tx_pool->free_map[i] = i;
 663
 664         tx_pool->consumer_index = 0;
 665         tx_pool->producer_index = 0;
 666         tx_pool->num_buffers = num_entries;
 667         tx_pool->buf_size = buf_size;
 668
 669         return 0;
 670 }
 671
 672 static int init_tx_pools(struct net_device *netdev)
 673 {
 674         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 675         int tx_subcrqs;
 676         int i, rc;
 677
 678         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 679         adapter->tx_pool = kcalloc(tx_subcrqs,
 680                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 681         if (!adapter->tx_pool)
 682                 return -1;
 683
 684         adapter->tso_pool = kcalloc(tx_subcrqs,
 685                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 686         if (!adapter->tso_pool)
 687                 return -1;
 688
 689         adapter->num_active_tx_pools = tx_subcrqs;
 690
 691         for (i = 0; i < tx_subcrqs; i++) {
 692                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 693                                       adapter->req_tx_entries_per_subcrq,
 694                                       adapter->req_mtu + VLAN_HLEN);
 695                 if (rc) {
 696                         release_tx_pools(adapter);
 697                         return rc;
 698                 }
 699
 700                 rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
 701                                       IBMVNIC_TSO_BUFS,
 702                                       IBMVNIC_TSO_BUF_SZ);
 703                 if (rc) {
 704                         release_tx_pools(adapter);
 705                         return rc;
 706                 }
 707         }
 708
 709         return 0;
 710 }
 711
 712 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 713 {
 714         int i;
 715
 716         if (adapter->napi_enabled)
 717                 return;
 718
 719         for (i = 0; i < adapter->req_rx_queues; i++)
 720                 napi_enable(&adapter->napi[i]);
 721
 722         adapter->napi_enabled = true;
 723 }
 724
 725 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 726 {
 727         int i;
 728
 729         if (!adapter->napi_enabled)
 730                 return;
 731
 732         for (i = 0; i < adapter->req_rx_queues; i++) {
 733                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 734                 napi_disable(&adapter->napi[i]);
 735         }
 736
 737         adapter->napi_enabled = false;
 738 }
 739
 740 static int init_napi(struct ibmvnic_adapter *adapter)
 741 {
 742         int i;
 743
 744         adapter->napi = kcalloc(adapter->req_rx_queues,
 745                                 sizeof(struct napi_struct), GFP_KERNEL);
 746         if (!adapter->napi)
 747                 return -ENOMEM;
 748
 749         for (i = 0; i < adapter->req_rx_queues; i++) {
 750                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 751                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 752                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 753         }
 754
 755         adapter->num_active_rx_napi = adapter->req_rx_queues;
 756         return 0;
 757 }
 758
 759 static void release_napi(struct ibmvnic_adapter *adapter)
 760 {
 761         int i;
 762
 763         if (!adapter->napi)
 764                 return;
 765
 766         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 767                 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
 768                 netif_napi_del(&adapter->napi[i]);
 769         }
 770
 771         kfree(adapter->napi);
 772         adapter->napi = NULL;
 773         adapter->num_active_rx_napi = 0;
 774         adapter->napi_enabled = false;
 775 }
 776
 777 static int ibmvnic_login(struct net_device *netdev)
 778 {
 779         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 780         unsigned long timeout = msecs_to_jiffies(30000);
 781         int retry_count = 0;
 782         bool retry;
 783         int rc;
 784
 785         do {
 786                 retry = false;
 787                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 788                         netdev_warn(netdev, "Login attempts exceeded\n");
 789                         return -1;
 790                 }
 791
 792                 adapter->init_done_rc = 0;
 793                 reinit_completion(&adapter->init_done);
 794                 rc = send_login(adapter);
 795                 if (rc) {
 796                         netdev_warn(netdev, "Unable to login\n");
 797                         return rc;
 798                 }
 799
 800                 if (!wait_for_completion_timeout(&adapter->init_done,
 801                                                  timeout)) {
 802                         netdev_warn(netdev, "Login timed out\n");
 803                         return -1;
 804                 }
 805
 806                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 807                         retry_count++;
 808                         release_sub_crqs(adapter, 1);
 809
 810                         retry = true;
 811                         netdev_dbg(netdev,
 812                                    "Received partial success, retrying...\n");
 813                         adapter->init_done_rc = 0;
 814                         reinit_completion(&adapter->init_done);
 815                         send_cap_queries(adapter);
 816                         if (!wait_for_completion_timeout(&adapter->init_done,
 817                                                          timeout)) {
 818                                 netdev_warn(netdev,
 819                                             "Capabilities query timed out\n");
 820                                 return -1;
 821                         }
 822
 823                         rc = init_sub_crqs(adapter);
 824                         if (rc) {
 825                                 netdev_warn(netdev,
 826                                             "SCRQ initialization failed\n");
 827                                 return -1;
 828                         }
 829
 830                         rc = init_sub_crq_irqs(adapter);
 831                         if (rc) {
 832                                 netdev_warn(netdev,
 833                                             "SCRQ irq initialization failed\n");
 834                                 return -1;
 835                         }
 836                 } else if (adapter->init_done_rc) {
 837                         netdev_warn(netdev, "Adapter login failed\n");
 838                         return -1;
 839                 }
 840         } while (retry);
 841
 842         __ibmvnic_set_mac(netdev, adapter->mac_addr);
 843
 844         return 0;
 845 }
 846
 847 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 848 {
 849         kfree(adapter->login_buf);
 850         adapter->login_buf = NULL;
 851 }
 852
 853 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 854 {
 855         kfree(adapter->login_rsp_buf);
 856         adapter->login_rsp_buf = NULL;
 857 }
 858
 859 static void release_resources(struct ibmvnic_adapter *adapter)
 860 {
 861         release_vpd_data(adapter);
 862
 863         release_tx_pools(adapter);
 864         release_rx_pools(adapter);
 865
 866         release_napi(adapter);
 867         release_login_rsp_buffer(adapter);
 868 }
 869
 870 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 871 {
 872         struct net_device *netdev = adapter->netdev;
 873         unsigned long timeout = msecs_to_jiffies(30000);
 874         union ibmvnic_crq crq;
 875         bool resend;
 876         int rc;
 877
 878         netdev_dbg(netdev, "setting link state %d\n", link_state);
 879
 880         memset(&crq, 0, sizeof(crq));
 881         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 882         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 883         crq.logical_link_state.link_state = link_state;
 884
 885         do {
 886                 resend = false;
 887
 888                 reinit_completion(&adapter->init_done);
 889                 rc = ibmvnic_send_crq(adapter, &crq);
 890                 if (rc) {
 891                         netdev_err(netdev, "Failed to set link state\n");
 892                         return rc;
 893                 }
 894
 895                 if (!wait_for_completion_timeout(&adapter->init_done,
 896                                                  timeout)) {
 897                         netdev_err(netdev, "timeout setting link state\n");
 898                         return -1;
 899                 }
 900
 901                 if (adapter->init_done_rc == 1) {
 902                         /* Partuial success, delay and re-send */
 903                         mdelay(1000);
 904                         resend = true;
 905                 } else if (adapter->init_done_rc) {
 906                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 907                                     adapter->init_done_rc);
 908                         return adapter->init_done_rc;
 909                 }
 910         } while (resend);
 911
 912         return 0;
 913 }
 914
 915 static int set_real_num_queues(struct net_device *netdev)
 916 {
 917         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 918         int rc;
 919
 920         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 921                    adapter->req_tx_queues, adapter->req_rx_queues);
 922
 923         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 924         if (rc) {
 925                 netdev_err(netdev, "failed to set the number of tx queues\n");
 926                 return rc;
 927         }
 928
 929         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 930         if (rc)
 931                 netdev_err(netdev, "failed to set the number of rx queues\n");
 932
 933         return rc;
 934 }
 935
 936 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 937 {
 938         struct device *dev = &adapter->vdev->dev;
 939         union ibmvnic_crq crq;
 940         int len = 0;
 941         int rc;
 942
 943         if (adapter->vpd->buff)
 944                 len = adapter->vpd->len;
 945
 946         init_completion(&adapter->fw_done);
 947         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 948         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 949         rc = ibmvnic_send_crq(adapter, &crq);
 950         if (rc)
 951                 return rc;
 952         wait_for_completion(&adapter->fw_done);
 953
 954         if (!adapter->vpd->len)
 955                 return -ENODATA;
 956
 957         if (!adapter->vpd->buff)
 958                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 959         else if (adapter->vpd->len != len)
 960                 adapter->vpd->buff =
 961                         krealloc(adapter->vpd->buff,
 962                                  adapter->vpd->len, GFP_KERNEL);
 963
 964         if (!adapter->vpd->buff) {
 965                 dev_err(dev, "Could allocate VPD buffer\n");
 966                 return -ENOMEM;
 967         }
 968
 969         adapter->vpd->dma_addr =
 970                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 971                                DMA_FROM_DEVICE);
 972         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 973                 dev_err(dev, "Could not map VPD buffer\n");
 974                 kfree(adapter->vpd->buff);
 975                 adapter->vpd->buff = NULL;
 976                 return -ENOMEM;
 977         }
 978
 979         reinit_completion(&adapter->fw_done);
 980         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 981         crq.get_vpd.cmd = GET_VPD;
 982         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 983         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 984         rc = ibmvnic_send_crq(adapter, &crq);
 985         if (rc) {
 986                 kfree(adapter->vpd->buff);
 987                 adapter->vpd->buff = NULL;
 988                 return rc;
 989         }
 990         wait_for_completion(&adapter->fw_done);
 991
 992         return 0;
 993 }
 994
 995 static int init_resources(struct ibmvnic_adapter *adapter)
 996 {
 997         struct net_device *netdev = adapter->netdev;
 998         int rc;
 999
1000         rc = set_real_num_queues(netdev);
1001         if (rc)
1002                 return rc;
1003
1004         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1005         if (!adapter->vpd)
1006                 return -ENOMEM;
1007
1008         /* Vital Product Data (VPD) */
1009         rc = ibmvnic_get_vpd(adapter);
1010         if (rc) {
1011                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1012                 return rc;
1013         }
1014
1015         adapter->map_id = 1;
1016
1017         rc = init_napi(adapter);
1018         if (rc)
1019                 return rc;
1020
1021         send_map_query(adapter);
1022
1023         rc = init_rx_pools(netdev);
1024         if (rc)
1025                 return rc;
1026
1027         rc = init_tx_pools(netdev);
1028         return rc;
1029 }
1030
1031 static int __ibmvnic_open(struct net_device *netdev)
1032 {
1033         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1034         enum vnic_state prev_state = adapter->state;
1035         int i, rc;
1036
1037         adapter->state = VNIC_OPENING;
1038         replenish_pools(adapter);
1039         ibmvnic_napi_enable(adapter);
1040
1041         /* We're ready to receive frames, enable the sub-crq interrupts and
1042          * set the logical link state to up
1043          */
1044         for (i = 0; i < adapter->req_rx_queues; i++) {
1045                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1046                 if (prev_state == VNIC_CLOSED)
1047                         enable_irq(adapter->rx_scrq[i]->irq);
1048                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1049         }
1050
1051         for (i = 0; i < adapter->req_tx_queues; i++) {
1052                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1053                 if (prev_state == VNIC_CLOSED)
1054                         enable_irq(adapter->tx_scrq[i]->irq);
1055                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1056         }
1057
1058         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1059         if (rc) {
1060                 for (i = 0; i < adapter->req_rx_queues; i++)
1061                         napi_disable(&adapter->napi[i]);
1062                 release_resources(adapter);
1063                 return rc;
1064         }
1065
1066         netif_tx_start_all_queues(netdev);
1067
1068         if (prev_state == VNIC_CLOSED) {
1069                 for (i = 0; i < adapter->req_rx_queues; i++)
1070                         napi_schedule(&adapter->napi[i]);
1071         }
1072
1073         adapter->state = VNIC_OPEN;
1074         return rc;
1075 }
1076
1077 static int ibmvnic_open(struct net_device *netdev)
1078 {
1079         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1080         int rc;
1081
1082         /* If device failover is pending, just set device state and return.
1083          * Device operation will be handled by reset routine.
1084          */
1085         if (adapter->failover_pending) {
1086                 adapter->state = VNIC_OPEN;
1087                 return 0;
1088         }
1089
1090         if (adapter->state != VNIC_CLOSED) {
1091                 rc = ibmvnic_login(netdev);
1092                 if (rc)
1093                         return rc;
1094
1095                 rc = init_resources(adapter);
1096                 if (rc) {
1097                         netdev_err(netdev, "failed to initialize resources\n");
1098                         release_resources(adapter);
1099                         return rc;
1100                 }
1101         }
1102
1103         rc = __ibmvnic_open(netdev);
1104
1105         return rc;
1106 }
1107
1108 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1109 {
1110         struct ibmvnic_rx_pool *rx_pool;
1111         struct ibmvnic_rx_buff *rx_buff;
1112         u64 rx_entries;
1113         int rx_scrqs;
1114         int i, j;
1115
1116         if (!adapter->rx_pool)
1117                 return;
1118
1119         rx_scrqs = adapter->num_active_rx_pools;
1120         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1121
1122         /* Free any remaining skbs in the rx buffer pools */
1123         for (i = 0; i < rx_scrqs; i++) {
1124                 rx_pool = &adapter->rx_pool[i];
1125                 if (!rx_pool || !rx_pool->rx_buff)
1126                         continue;
1127
1128                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1129                 for (j = 0; j < rx_entries; j++) {
1130                         rx_buff = &rx_pool->rx_buff[j];
1131                         if (rx_buff && rx_buff->skb) {
1132                                 dev_kfree_skb_any(rx_buff->skb);
1133                                 rx_buff->skb = NULL;
1134                         }
1135                 }
1136         }
1137 }
1138
1139 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1140                               struct ibmvnic_tx_pool *tx_pool)
1141 {
1142         struct ibmvnic_tx_buff *tx_buff;
1143         u64 tx_entries;
1144         int i;
1145
1146         if (!tx_pool || !tx_pool->tx_buff)
1147                 return;
1148
1149         tx_entries = tx_pool->num_buffers;
1150
1151         for (i = 0; i < tx_entries; i++) {
1152                 tx_buff = &tx_pool->tx_buff[i];
1153                 if (tx_buff && tx_buff->skb) {
1154                         dev_kfree_skb_any(tx_buff->skb);
1155                         tx_buff->skb = NULL;
1156                 }
1157         }
1158 }
1159
1160 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1161 {
1162         int tx_scrqs;
1163         int i;
1164
1165         if (!adapter->tx_pool || !adapter->tso_pool)
1166                 return;
1167
1168         tx_scrqs = adapter->num_active_tx_pools;
1169
1170         /* Free any remaining skbs in the tx buffer pools */
1171         for (i = 0; i < tx_scrqs; i++) {
1172                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1173                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1174                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1175         }
1176 }
1177
1178 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1179 {
1180         struct net_device *netdev = adapter->netdev;
1181         int i;
1182
1183         if (adapter->tx_scrq) {
1184                 for (i = 0; i < adapter->req_tx_queues; i++)
1185                         if (adapter->tx_scrq[i]->irq) {
1186                                 netdev_dbg(netdev,
1187                                            "Disabling tx_scrq[%d] irq\n", i);
1188                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1189                                 disable_irq(adapter->tx_scrq[i]->irq);
1190                         }
1191         }
1192
1193         if (adapter->rx_scrq) {
1194                 for (i = 0; i < adapter->req_rx_queues; i++) {
1195                         if (adapter->rx_scrq[i]->irq) {
1196                                 netdev_dbg(netdev,
1197                                            "Disabling rx_scrq[%d] irq\n", i);
1198                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1199                                 disable_irq(adapter->rx_scrq[i]->irq);
1200                         }
1201                 }
1202         }
1203 }
1204
1205 static void ibmvnic_cleanup(struct net_device *netdev)
1206 {
1207         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1208
1209         /* ensure that transmissions are stopped if called by do_reset */
1210         if (adapter->resetting)
1211                 netif_tx_disable(netdev);
1212         else
1213                 netif_tx_stop_all_queues(netdev);
1214
1215         ibmvnic_napi_disable(adapter);
1216         ibmvnic_disable_irqs(adapter);
1217
1218         clean_rx_pools(adapter);
1219         clean_tx_pools(adapter);
1220 }
1221
1222 static int __ibmvnic_close(struct net_device *netdev)
1223 {
1224         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1225         int rc = 0;
1226
1227         adapter->state = VNIC_CLOSING;
1228         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1229         if (rc)
1230                 return rc;
1231         adapter->state = VNIC_CLOSED;
1232         return 0;
1233 }
1234
1235 static int ibmvnic_close(struct net_device *netdev)
1236 {
1237         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1238         int rc;
1239
1240         /* If device failover is pending, just set device state and return.
1241          * Device operation will be handled by reset routine.
1242          */
1243         if (adapter->failover_pending) {
1244                 adapter->state = VNIC_CLOSED;
1245                 return 0;
1246         }
1247
1248         rc = __ibmvnic_close(netdev);
1249         ibmvnic_cleanup(netdev);
1250
1251         return rc;
1252 }
1253
1254 /**
1255  * build_hdr_data - creates L2/L3/L4 header data buffer
1256  * @hdr_field - bitfield determining needed headers
1257  * @skb - socket buffer
1258  * @hdr_len - array of header lengths
1259  * @tot_len - total length of data
1260  *
1261  * Reads hdr_field to determine which headers are needed by firmware.
1262  * Builds a buffer containing these headers.  Saves individual header
1263  * lengths and total buffer length to be used to build descriptors.
1264  */
1265 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1266                           int *hdr_len, u8 *hdr_data)
1267 {
1268         int len = 0;
1269         u8 *hdr;
1270
1271         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1272                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1273         else
1274                 hdr_len[0] = sizeof(struct ethhdr);
1275
1276         if (skb->protocol == htons(ETH_P_IP)) {
1277                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1278                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1279                         hdr_len[2] = tcp_hdrlen(skb);
1280                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1281                         hdr_len[2] = sizeof(struct udphdr);
1282         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1283                 hdr_len[1] = sizeof(struct ipv6hdr);
1284                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1285                         hdr_len[2] = tcp_hdrlen(skb);
1286                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1287                         hdr_len[2] = sizeof(struct udphdr);
1288         } else if (skb->protocol == htons(ETH_P_ARP)) {
1289                 hdr_len[1] = arp_hdr_len(skb->dev);
1290                 hdr_len[2] = 0;
1291         }
1292
1293         memset(hdr_data, 0, 120);
1294         if ((hdr_field >> 6) & 1) {
1295                 hdr = skb_mac_header(skb);
1296                 memcpy(hdr_data, hdr, hdr_len[0]);
1297                 len += hdr_len[0];
1298         }
1299
1300         if ((hdr_field >> 5) & 1) {
1301                 hdr = skb_network_header(skb);
1302                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1303                 len += hdr_len[1];
1304         }
1305
1306         if ((hdr_field >> 4) & 1) {
1307                 hdr = skb_transport_header(skb);
1308                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1309                 len += hdr_len[2];
1310         }
1311         return len;
1312 }
1313
1314 /**
1315  * create_hdr_descs - create header and header extension descriptors
1316  * @hdr_field - bitfield determining needed headers
1317  * @data - buffer containing header data
1318  * @len - length of data buffer
1319  * @hdr_len - array of individual header lengths
1320  * @scrq_arr - descriptor array
1321  *
1322  * Creates header and, if needed, header extension descriptors and
1323  * places them in a descriptor array, scrq_arr
1324  */
1325
1326 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1327                             union sub_crq *scrq_arr)
1328 {
1329         union sub_crq hdr_desc;
1330         int tmp_len = len;
1331         int num_descs = 0;
1332         u8 *data, *cur;
1333         int tmp;
1334
1335         while (tmp_len > 0) {
1336                 cur = hdr_data + len - tmp_len;
1337
1338                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1339                 if (cur != hdr_data) {
1340                         data = hdr_desc.hdr_ext.data;
1341                         tmp = tmp_len > 29 ? 29 : tmp_len;
1342                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1343                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1344                         hdr_desc.hdr_ext.len = tmp;
1345                 } else {
1346                         data = hdr_desc.hdr.data;
1347                         tmp = tmp_len > 24 ? 24 : tmp_len;
1348                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1349                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1350                         hdr_desc.hdr.len = tmp;
1351                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1352                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1353                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1354                         hdr_desc.hdr.flag = hdr_field << 1;
1355                 }
1356                 memcpy(data, cur, tmp);
1357                 tmp_len -= tmp;
1358                 *scrq_arr = hdr_desc;
1359                 scrq_arr++;
1360                 num_descs++;
1361         }
1362
1363         return num_descs;
1364 }
1365
1366 /**
1367  * build_hdr_descs_arr - build a header descriptor array
1368  * @skb - socket buffer
1369  * @num_entries - number of descriptors to be sent
1370  * @subcrq - first TX descriptor
1371  * @hdr_field - bit field determining which headers will be sent
1372  *
1373  * This function will build a TX descriptor array with applicable
1374  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1375  */
1376
1377 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1378                                 int *num_entries, u8 hdr_field)
1379 {
1380         int hdr_len[3] = {0, 0, 0};
1381         int tot_len;
1382         u8 *hdr_data = txbuff->hdr_data;
1383
1384         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1385                                  txbuff->hdr_data);
1386         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1387                          txbuff->indir_arr + 1);
1388 }
1389
1390 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1391                                     struct net_device *netdev)
1392 {
1393         /* For some backing devices, mishandling of small packets
1394          * can result in a loss of connection or TX stall. Device
1395          * architects recommend that no packet should be smaller
1396          * than the minimum MTU value provided to the driver, so
1397          * pad any packets to that length
1398          */
1399         if (skb->len < netdev->min_mtu)
1400                 return skb_put_padto(skb, netdev->min_mtu);
1401
1402         return 0;
1403 }
1404
1405 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1406 {
1407         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1408         int queue_num = skb_get_queue_mapping(skb);
1409         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1410         struct device *dev = &adapter->vdev->dev;
1411         struct ibmvnic_tx_buff *tx_buff = NULL;
1412         struct ibmvnic_sub_crq_queue *tx_scrq;
1413         struct ibmvnic_tx_pool *tx_pool;
1414         unsigned int tx_send_failed = 0;
1415         unsigned int tx_map_failed = 0;
1416         unsigned int tx_dropped = 0;
1417         unsigned int tx_packets = 0;
1418         unsigned int tx_bytes = 0;
1419         dma_addr_t data_dma_addr;
1420         struct netdev_queue *txq;
1421         unsigned long lpar_rc;
1422         union sub_crq tx_crq;
1423         unsigned int offset;
1424         int num_entries = 1;
1425         unsigned char *dst;
1426         u64 *handle_array;
1427         int index = 0;
1428         u8 proto = 0;
1429         netdev_tx_t ret = NETDEV_TX_OK;
1430
1431         if (adapter->resetting) {
1432                 if (!netif_subqueue_stopped(netdev, skb))
1433                         netif_stop_subqueue(netdev, queue_num);
1434                 dev_kfree_skb_any(skb);
1435
1436                 tx_send_failed++;
1437                 tx_dropped++;
1438                 ret = NETDEV_TX_OK;
1439                 goto out;
1440         }
1441
1442         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1443                 tx_dropped++;
1444                 tx_send_failed++;
1445                 ret = NETDEV_TX_OK;
1446                 goto out;
1447         }
1448         if (skb_is_gso(skb))
1449                 tx_pool = &adapter->tso_pool[queue_num];
1450         else
1451                 tx_pool = &adapter->tx_pool[queue_num];
1452
1453         tx_scrq = adapter->tx_scrq[queue_num];
1454         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1455         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1456                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1457
1458         index = tx_pool->free_map[tx_pool->consumer_index];
1459
1460         if (index == IBMVNIC_INVALID_MAP) {
1461                 dev_kfree_skb_any(skb);
1462                 tx_send_failed++;
1463                 tx_dropped++;
1464                 ret = NETDEV_TX_OK;
1465                 goto out;
1466         }
1467
1468         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1469
1470         offset = index * tx_pool->buf_size;
1471         dst = tx_pool->long_term_buff.buff + offset;
1472         memset(dst, 0, tx_pool->buf_size);
1473         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1474
1475         if (skb_shinfo(skb)->nr_frags) {
1476                 int cur, i;
1477
1478                 /* Copy the head */
1479                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1480                 cur = skb_headlen(skb);
1481
1482                 /* Copy the frags */
1483                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1484                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1485
1486                         memcpy(dst + cur,
1487                                page_address(skb_frag_page(frag)) +
1488                                frag->page_offset, skb_frag_size(frag));
1489                         cur += skb_frag_size(frag);
1490                 }
1491         } else {
1492                 skb_copy_from_linear_data(skb, dst, skb->len);
1493         }
1494
1495         tx_pool->consumer_index =
1496             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1497
1498         tx_buff = &tx_pool->tx_buff[index];
1499         tx_buff->skb = skb;
1500         tx_buff->data_dma[0] = data_dma_addr;
1501         tx_buff->data_len[0] = skb->len;
1502         tx_buff->index = index;
1503         tx_buff->pool_index = queue_num;
1504         tx_buff->last_frag = true;
1505
1506         memset(&tx_crq, 0, sizeof(tx_crq));
1507         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1508         tx_crq.v1.type = IBMVNIC_TX_DESC;
1509         tx_crq.v1.n_crq_elem = 1;
1510         tx_crq.v1.n_sge = 1;
1511         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1512
1513         if (skb_is_gso(skb))
1514                 tx_crq.v1.correlator =
1515                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1516         else
1517                 tx_crq.v1.correlator = cpu_to_be32(index);
1518         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1519         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1520         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1521
1522         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1523                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1524                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1525         }
1526
1527         if (skb->protocol == htons(ETH_P_IP)) {
1528                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1529                 proto = ip_hdr(skb)->protocol;
1530         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1531                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1532                 proto = ipv6_hdr(skb)->nexthdr;
1533         }
1534
1535         if (proto == IPPROTO_TCP)
1536                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1537         else if (proto == IPPROTO_UDP)
1538                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1539
1540         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1541                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1542                 hdrs += 2;
1543         }
1544         if (skb_is_gso(skb)) {
1545                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1546                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1547                 hdrs += 2;
1548         }
1549         /* determine if l2/3/4 headers are sent to firmware */
1550         if ((*hdrs >> 7) & 1) {
1551                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1552                 tx_crq.v1.n_crq_elem = num_entries;
1553                 tx_buff->num_entries = num_entries;
1554                 tx_buff->indir_arr[0] = tx_crq;
1555                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1556                                                     sizeof(tx_buff->indir_arr),
1557                                                     DMA_TO_DEVICE);
1558                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1559                         dev_kfree_skb_any(skb);
1560                         tx_buff->skb = NULL;
1561                         if (!firmware_has_feature(FW_FEATURE_CMO))
1562                                 dev_err(dev, "tx: unable to map descriptor array\n");
1563                         tx_map_failed++;
1564                         tx_dropped++;
1565                         ret = NETDEV_TX_OK;
1566                         goto tx_err_out;
1567                 }
1568                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1569                                                (u64)tx_buff->indir_dma,
1570                                                (u64)num_entries);
1571                 dma_unmap_single(dev, tx_buff->indir_dma,
1572                                  sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1573         } else {
1574                 tx_buff->num_entries = num_entries;
1575                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1576                                       &tx_crq);
1577         }
1578         if (lpar_rc != H_SUCCESS) {
1579                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1580                         dev_err_ratelimited(dev, "tx: send failed\n");
1581                 dev_kfree_skb_any(skb);
1582                 tx_buff->skb = NULL;
1583
1584                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1585                         /* Disable TX and report carrier off if queue is closed
1586                          * or pending failover.
1587                          * Firmware guarantees that a signal will be sent to the
1588                          * driver, triggering a reset or some other action.
1589                          */
1590                         netif_tx_stop_all_queues(netdev);
1591                         netif_carrier_off(netdev);
1592                 }
1593
1594                 tx_send_failed++;
1595                 tx_dropped++;
1596                 ret = NETDEV_TX_OK;
1597                 goto tx_err_out;
1598         }
1599
1600         if (atomic_add_return(num_entries, &tx_scrq->used)
1601                                         >= adapter->req_tx_entries_per_subcrq) {
1602                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1603                 netif_stop_subqueue(netdev, queue_num);
1604         }
1605
1606         tx_packets++;
1607         tx_bytes += skb->len;
1608         txq->trans_start = jiffies;
1609         ret = NETDEV_TX_OK;
1610         goto out;
1611
1612 tx_err_out:
1613         /* roll back consumer index and map array*/
1614         if (tx_pool->consumer_index == 0)
1615                 tx_pool->consumer_index =
1616                         tx_pool->num_buffers - 1;
1617         else
1618                 tx_pool->consumer_index--;
1619         tx_pool->free_map[tx_pool->consumer_index] = index;
1620 out:
1621         netdev->stats.tx_dropped += tx_dropped;
1622         netdev->stats.tx_bytes += tx_bytes;
1623         netdev->stats.tx_packets += tx_packets;
1624         adapter->tx_send_failed += tx_send_failed;
1625         adapter->tx_map_failed += tx_map_failed;
1626         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1627         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1628         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1629
1630         return ret;
1631 }
1632
1633 static void ibmvnic_set_multi(struct net_device *netdev)
1634 {
1635         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1636         struct netdev_hw_addr *ha;
1637         union ibmvnic_crq crq;
1638
1639         memset(&crq, 0, sizeof(crq));
1640         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1641         crq.request_capability.cmd = REQUEST_CAPABILITY;
1642
1643         if (netdev->flags & IFF_PROMISC) {
1644                 if (!adapter->promisc_supported)
1645                         return;
1646         } else {
1647                 if (netdev->flags & IFF_ALLMULTI) {
1648                         /* Accept all multicast */
1649                         memset(&crq, 0, sizeof(crq));
1650                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1651                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1652                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1653                         ibmvnic_send_crq(adapter, &crq);
1654                 } else if (netdev_mc_empty(netdev)) {
1655                         /* Reject all multicast */
1656                         memset(&crq, 0, sizeof(crq));
1657                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1658                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1659                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1660                         ibmvnic_send_crq(adapter, &crq);
1661                 } else {
1662                         /* Accept one or more multicast(s) */
1663                         netdev_for_each_mc_addr(ha, netdev) {
1664                                 memset(&crq, 0, sizeof(crq));
1665                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1666                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1667                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1668                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1669                                                 ha->addr);
1670                                 ibmvnic_send_crq(adapter, &crq);
1671                         }
1672                 }
1673         }
1674 }
1675
1676 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1677 {
1678         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1679         union ibmvnic_crq crq;
1680         int rc;
1681
1682         if (!is_valid_ether_addr(dev_addr)) {
1683                 rc = -EADDRNOTAVAIL;
1684                 goto err;
1685         }
1686
1687         memset(&crq, 0, sizeof(crq));
1688         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1689         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1690         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1691
1692         init_completion(&adapter->fw_done);
1693         rc = ibmvnic_send_crq(adapter, &crq);
1694         if (rc) {
1695                 rc = -EIO;
1696                 goto err;
1697         }
1698
1699         wait_for_completion(&adapter->fw_done);
1700         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1701         if (adapter->fw_done_rc) {
1702                 rc = -EIO;
1703                 goto err;
1704         }
1705
1706         return 0;
1707 err:
1708         ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1709         return rc;
1710 }
1711
1712 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1713 {
1714         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1715         struct sockaddr *addr = p;
1716         int rc;
1717
1718         rc = 0;
1719         ether_addr_copy(adapter->mac_addr, addr->sa_data);
1720         if (adapter->state != VNIC_PROBED)
1721                 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1722
1723         return rc;
1724 }
1725
1726 /**
1727  * do_reset returns zero if we are able to keep processing reset events, or
1728  * non-zero if we hit a fatal error and must halt.
1729  */
1730 static int do_reset(struct ibmvnic_adapter *adapter,
1731                     struct ibmvnic_rwi *rwi, u32 reset_state)
1732 {
1733         u64 old_num_rx_queues, old_num_tx_queues;
1734         u64 old_num_rx_slots, old_num_tx_slots;
1735         struct net_device *netdev = adapter->netdev;
1736         int i, rc;
1737
1738         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1739                    rwi->reset_reason);
1740
1741         netif_carrier_off(netdev);
1742         adapter->reset_reason = rwi->reset_reason;
1743
1744         old_num_rx_queues = adapter->req_rx_queues;
1745         old_num_tx_queues = adapter->req_tx_queues;
1746         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1747         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1748
1749         ibmvnic_cleanup(netdev);
1750
1751         if (reset_state == VNIC_OPEN &&
1752             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1753             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1754                 rc = __ibmvnic_close(netdev);
1755                 if (rc)
1756                         return rc;
1757         }
1758
1759         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1760             adapter->wait_for_reset) {
1761                 release_resources(adapter);
1762                 release_sub_crqs(adapter, 1);
1763                 release_crq_queue(adapter);
1764         }
1765
1766         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1767                 /* remove the closed state so when we call open it appears
1768                  * we are coming from the probed state.
1769                  */
1770                 adapter->state = VNIC_PROBED;
1771
1772                 if (adapter->wait_for_reset) {
1773                         rc = init_crq_queue(adapter);
1774                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1775                         rc = ibmvnic_reenable_crq_queue(adapter);
1776                         release_sub_crqs(adapter, 1);
1777                 } else {
1778                         rc = ibmvnic_reset_crq(adapter);
1779                         if (!rc)
1780                                 rc = vio_enable_interrupts(adapter->vdev);
1781                 }
1782
1783                 if (rc) {
1784                         netdev_err(adapter->netdev,
1785                                    "Couldn't initialize crq. rc=%d\n", rc);
1786                         return rc;
1787                 }
1788
1789                 rc = ibmvnic_reset_init(adapter);
1790                 if (rc)
1791                         return IBMVNIC_INIT_FAILED;
1792
1793                 /* If the adapter was in PROBE state prior to the reset,
1794                  * exit here.
1795                  */
1796                 if (reset_state == VNIC_PROBED)
1797                         return 0;
1798
1799                 rc = ibmvnic_login(netdev);
1800                 if (rc) {
1801                         adapter->state = reset_state;
1802                         return rc;
1803                 }
1804
1805                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1806                     adapter->wait_for_reset) {
1807                         rc = init_resources(adapter);
1808                         if (rc)
1809                                 return rc;
1810                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1811                            adapter->req_tx_queues != old_num_tx_queues ||
1812                            adapter->req_rx_add_entries_per_subcrq !=
1813                                                         old_num_rx_slots ||
1814                            adapter->req_tx_entries_per_subcrq !=
1815                                                         old_num_tx_slots) {
1816                         release_rx_pools(adapter);
1817                         release_tx_pools(adapter);
1818                         release_napi(adapter);
1819                         release_vpd_data(adapter);
1820
1821                         rc = init_resources(adapter);
1822                         if (rc)
1823                                 return rc;
1824
1825                 } else {
1826                         rc = reset_tx_pools(adapter);
1827                         if (rc)
1828                                 return rc;
1829
1830                         rc = reset_rx_pools(adapter);
1831                         if (rc)
1832                                 return rc;
1833                 }
1834                 ibmvnic_disable_irqs(adapter);
1835         }
1836         adapter->state = VNIC_CLOSED;
1837
1838         if (reset_state == VNIC_CLOSED)
1839                 return 0;
1840
1841         rc = __ibmvnic_open(netdev);
1842         if (rc) {
1843                 if (list_empty(&adapter->rwi_list))
1844                         adapter->state = VNIC_CLOSED;
1845                 else
1846                         adapter->state = reset_state;
1847
1848                 return 0;
1849         }
1850
1851         /* refresh device's multicast list */
1852         ibmvnic_set_multi(netdev);
1853
1854         /* kick napi */
1855         for (i = 0; i < adapter->req_rx_queues; i++)
1856                 napi_schedule(&adapter->napi[i]);
1857
1858         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1859             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1860                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1861
1862         return 0;
1863 }
1864
1865 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1866                          struct ibmvnic_rwi *rwi, u32 reset_state)
1867 {
1868         struct net_device *netdev = adapter->netdev;
1869         int rc;
1870
1871         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1872                    rwi->reset_reason);
1873
1874         netif_carrier_off(netdev);
1875         adapter->reset_reason = rwi->reset_reason;
1876
1877         ibmvnic_cleanup(netdev);
1878         release_resources(adapter);
1879         release_sub_crqs(adapter, 0);
1880         release_crq_queue(adapter);
1881
1882         /* remove the closed state so when we call open it appears
1883          * we are coming from the probed state.
1884          */
1885         adapter->state = VNIC_PROBED;
1886
1887         reinit_completion(&adapter->init_done);
1888         rc = init_crq_queue(adapter);
1889         if (rc) {
1890                 netdev_err(adapter->netdev,
1891                            "Couldn't initialize crq. rc=%d\n", rc);
1892                 return rc;
1893         }
1894
1895         rc = ibmvnic_init(adapter);
1896         if (rc)
1897                 return rc;
1898
1899         /* If the adapter was in PROBE state prior to the reset,
1900          * exit here.
1901          */
1902         if (reset_state == VNIC_PROBED)
1903                 return 0;
1904
1905         rc = ibmvnic_login(netdev);
1906         if (rc) {
1907                 adapter->state = VNIC_PROBED;
1908                 return 0;
1909         }
1910
1911         rc = init_resources(adapter);
1912         if (rc)
1913                 return rc;
1914
1915         ibmvnic_disable_irqs(adapter);
1916         adapter->state = VNIC_CLOSED;
1917
1918         if (reset_state == VNIC_CLOSED)
1919                 return 0;
1920
1921         rc = __ibmvnic_open(netdev);
1922         if (rc) {
1923                 if (list_empty(&adapter->rwi_list))
1924                         adapter->state = VNIC_CLOSED;
1925                 else
1926                         adapter->state = reset_state;
1927
1928                 return 0;
1929         }
1930
1931         return 0;
1932 }
1933
1934 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1935 {
1936         struct ibmvnic_rwi *rwi;
1937         unsigned long flags;
1938
1939         spin_lock_irqsave(&adapter->rwi_lock, flags);
1940
1941         if (!list_empty(&adapter->rwi_list)) {
1942                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1943                                        list);
1944                 list_del(&rwi->list);
1945         } else {
1946                 rwi = NULL;
1947         }
1948
1949         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1950         return rwi;
1951 }
1952
1953 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1954 {
1955         struct ibmvnic_rwi *rwi;
1956
1957         rwi = get_next_rwi(adapter);
1958         while (rwi) {
1959                 kfree(rwi);
1960                 rwi = get_next_rwi(adapter);
1961         }
1962 }
1963
1964 static void __ibmvnic_reset(struct work_struct *work)
1965 {
1966         struct ibmvnic_rwi *rwi;
1967         struct ibmvnic_adapter *adapter;
1968         bool we_lock_rtnl = false;
1969         u32 reset_state;
1970         int rc = 0;
1971
1972         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1973
1974         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1975          * unless wait_for_reset is set, in which case the rtnl lock
1976          * has already been taken before initializing the reset
1977          */
1978         if (!adapter->wait_for_reset) {
1979                 rtnl_lock();
1980                 we_lock_rtnl = true;
1981         }
1982         reset_state = adapter->state;
1983
1984         rwi = get_next_rwi(adapter);
1985         while (rwi) {
1986                 if (adapter->state == VNIC_REMOVING ||
1987                     adapter->state == VNIC_REMOVED)
1988                         kfree(rwi);
1989                         rc = EBUSY;
1990                         break;
1991                 }
1992
1993                 if (adapter->force_reset_recovery) {
1994                         adapter->force_reset_recovery = false;
1995                         rc = do_hard_reset(adapter, rwi, reset_state);
1996                 } else {
1997                         rc = do_reset(adapter, rwi, reset_state);
1998                 }
1999                 kfree(rwi);
2000                 if (rc && rc != IBMVNIC_INIT_FAILED &&
2001                     !adapter->force_reset_recovery)
2002                         break;
2003
2004                 rwi = get_next_rwi(adapter);
2005         }
2006
2007         if (adapter->wait_for_reset) {
2008                 adapter->wait_for_reset = false;
2009                 adapter->reset_done_rc = rc;
2010                 complete(&adapter->reset_done);
2011         }
2012
2013         if (rc) {
2014                 netdev_dbg(adapter->netdev, "Reset failed\n");
2015                 free_all_rwi(adapter);
2016         }
2017
2018         adapter->resetting = false;
2019         if (we_lock_rtnl)
2020                 rtnl_unlock();
2021 }
2022
2023 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2024                          enum ibmvnic_reset_reason reason)
2025 {
2026         struct list_head *entry, *tmp_entry;
2027         struct ibmvnic_rwi *rwi, *tmp;
2028         struct net_device *netdev = adapter->netdev;
2029         unsigned long flags;
2030         int ret;
2031
2032         if (adapter->state == VNIC_REMOVING ||
2033             adapter->state == VNIC_REMOVED ||
2034             adapter->failover_pending) {
2035                 ret = EBUSY;
2036                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2037                 goto err;
2038         }
2039
2040         if (adapter->state == VNIC_PROBING) {
2041                 netdev_warn(netdev, "Adapter reset during probe\n");
2042                 ret = adapter->init_done_rc = EAGAIN;
2043                 goto err;
2044         }
2045
2046         spin_lock_irqsave(&adapter->rwi_lock, flags);
2047
2048         list_for_each(entry, &adapter->rwi_list) {
2049                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2050                 if (tmp->reset_reason == reason) {
2051                         netdev_dbg(netdev, "Skipping matching reset\n");
2052                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2053                         ret = EBUSY;
2054                         goto err;
2055                 }
2056         }
2057
2058         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2059         if (!rwi) {
2060                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2061                 ibmvnic_close(netdev);
2062                 ret = ENOMEM;
2063                 goto err;
2064         }
2065         /* if we just received a transport event,
2066          * flush reset queue and process this reset
2067          */
2068         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2069                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2070                         list_del(entry);
2071         }
2072         rwi->reset_reason = reason;
2073         list_add_tail(&rwi->list, &adapter->rwi_list);
2074         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2075         adapter->resetting = true;
2076         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2077         schedule_work(&adapter->ibmvnic_reset);
2078
2079         return 0;
2080 err:
2081         if (adapter->wait_for_reset)
2082                 adapter->wait_for_reset = false;
2083         return -ret;
2084 }
2085
2086 static void ibmvnic_tx_timeout(struct net_device *dev)
2087 {
2088         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2089
2090         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2091 }
2092
2093 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2094                                   struct ibmvnic_rx_buff *rx_buff)
2095 {
2096         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2097
2098         rx_buff->skb = NULL;
2099
2100         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2101         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2102
2103         atomic_dec(&pool->available);
2104 }
2105
2106 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2107 {
2108         struct net_device *netdev = napi->dev;
2109         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2110         int scrq_num = (int)(napi - adapter->napi);
2111         int frames_processed = 0;
2112
2113 restart_poll:
2114         while (frames_processed < budget) {
2115                 struct sk_buff *skb;
2116                 struct ibmvnic_rx_buff *rx_buff;
2117                 union sub_crq *next;
2118                 u32 length;
2119                 u16 offset;
2120                 u8 flags = 0;
2121
2122                 if (unlikely(adapter->resetting &&
2123                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2124                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2125                         napi_complete_done(napi, frames_processed);
2126                         return frames_processed;
2127                 }
2128
2129                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2130                         break;
2131                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2132                 rx_buff =
2133                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2134                                                           rx_comp.correlator);
2135                 /* do error checking */
2136                 if (next->rx_comp.rc) {
2137                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2138                                    be16_to_cpu(next->rx_comp.rc));
2139                         /* free the entry */
2140                         next->rx_comp.first = 0;
2141                         dev_kfree_skb_any(rx_buff->skb);
2142                         remove_buff_from_pool(adapter, rx_buff);
2143                         continue;
2144                 } else if (!rx_buff->skb) {
2145                         /* free the entry */
2146                         next->rx_comp.first = 0;
2147                         remove_buff_from_pool(adapter, rx_buff);
2148                         continue;
2149                 }
2150
2151                 length = be32_to_cpu(next->rx_comp.len);
2152                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2153                 flags = next->rx_comp.flags;
2154                 skb = rx_buff->skb;
2155                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2156                                         length);
2157
2158                 /* VLAN Header has been stripped by the system firmware and
2159                  * needs to be inserted by the driver
2160                  */
2161                 if (adapter->rx_vlan_header_insertion &&
2162                     (flags & IBMVNIC_VLAN_STRIPPED))
2163                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2164                                                ntohs(next->rx_comp.vlan_tci));
2165
2166                 /* free the entry */
2167                 next->rx_comp.first = 0;
2168                 remove_buff_from_pool(adapter, rx_buff);
2169
2170                 skb_put(skb, length);
2171                 skb->protocol = eth_type_trans(skb, netdev);
2172                 skb_record_rx_queue(skb, scrq_num);
2173
2174                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2175                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2176                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2177                 }
2178
2179                 length = skb->len;
2180                 napi_gro_receive(napi, skb); /* send it up */
2181                 netdev->stats.rx_packets++;
2182                 netdev->stats.rx_bytes += length;
2183                 adapter->rx_stats_buffers[scrq_num].packets++;
2184                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2185                 frames_processed++;
2186         }
2187
2188         if (adapter->state != VNIC_CLOSING)
2189                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2190
2191         if (frames_processed < budget) {
2192                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2193                 napi_complete_done(napi, frames_processed);
2194                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2195                     napi_reschedule(napi)) {
2196                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2197                         goto restart_poll;
2198                 }
2199         }
2200         return frames_processed;
2201 }
2202
2203 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2204 {
2205         int rc, ret;
2206
2207         adapter->fallback.mtu = adapter->req_mtu;
2208         adapter->fallback.rx_queues = adapter->req_rx_queues;
2209         adapter->fallback.tx_queues = adapter->req_tx_queues;
2210         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2211         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2212
2213         init_completion(&adapter->reset_done);
2214         adapter->wait_for_reset = true;
2215         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2216         if (rc)
2217                 return rc;
2218         wait_for_completion(&adapter->reset_done);
2219
2220         ret = 0;
2221         if (adapter->reset_done_rc) {
2222                 ret = -EIO;
2223                 adapter->desired.mtu = adapter->fallback.mtu;
2224                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2225                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2226                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2227                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2228
2229                 init_completion(&adapter->reset_done);
2230                 adapter->wait_for_reset = true;
2231                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2232                 if (rc)
2233                         return ret;
2234                 wait_for_completion(&adapter->reset_done);
2235         }
2236         adapter->wait_for_reset = false;
2237
2238         return ret;
2239 }
2240
2241 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2242 {
2243         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2244
2245         adapter->desired.mtu = new_mtu + ETH_HLEN;
2246
2247         return wait_for_reset(adapter);
2248 }
2249
2250 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2251                                                 struct net_device *dev,
2252                                                 netdev_features_t features)
2253 {
2254         /* Some backing hardware adapters can not
2255          * handle packets with a MSS less than 224
2256          * or with only one segment.
2257          */
2258         if (skb_is_gso(skb)) {
2259                 if (skb_shinfo(skb)->gso_size < 224 ||
2260                     skb_shinfo(skb)->gso_segs == 1)
2261                         features &= ~NETIF_F_GSO_MASK;
2262         }
2263
2264         return features;
2265 }
2266
2267 static const struct net_device_ops ibmvnic_netdev_ops = {
2268         .ndo_open               = ibmvnic_open,
2269         .ndo_stop               = ibmvnic_close,
2270         .ndo_start_xmit         = ibmvnic_xmit,
2271         .ndo_set_rx_mode        = ibmvnic_set_multi,
2272         .ndo_set_mac_address    = ibmvnic_set_mac,
2273         .ndo_validate_addr      = eth_validate_addr,
2274         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2275         .ndo_change_mtu         = ibmvnic_change_mtu,
2276         .ndo_features_check     = ibmvnic_features_check,
2277 };
2278
2279 /* ethtool functions */
2280
2281 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2282                                       struct ethtool_link_ksettings *cmd)
2283 {
2284         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2285         int rc;
2286
2287         rc = send_query_phys_parms(adapter);
2288         if (rc) {
2289                 adapter->speed = SPEED_UNKNOWN;
2290                 adapter->duplex = DUPLEX_UNKNOWN;
2291         }
2292         cmd->base.speed = adapter->speed;
2293         cmd->base.duplex = adapter->duplex;
2294         cmd->base.port = PORT_FIBRE;
2295         cmd->base.phy_address = 0;
2296         cmd->base.autoneg = AUTONEG_ENABLE;
2297
2298         return 0;
2299 }
2300
2301 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2302                                 struct ethtool_drvinfo *info)
2303 {
2304         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2305
2306         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2307         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2308         strlcpy(info->fw_version, adapter->fw_version,
2309                 sizeof(info->fw_version));
2310 }
2311
2312 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2313 {
2314         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2315
2316         return adapter->msg_enable;
2317 }
2318
2319 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2320 {
2321         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2322
2323         adapter->msg_enable = data;
2324 }
2325
2326 static u32 ibmvnic_get_link(struct net_device *netdev)
2327 {
2328         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2329
2330         /* Don't need to send a query because we request a logical link up at
2331          * init and then we wait for link state indications
2332          */
2333         return adapter->logical_link_state;
2334 }
2335
2336 static void ibmvnic_get_ringparam(struct net_device *netdev,
2337                                   struct ethtool_ringparam *ring)
2338 {
2339         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2340
2341         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2342                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2343                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2344         } else {
2345                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2346                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2347         }
2348         ring->rx_mini_max_pending = 0;
2349         ring->rx_jumbo_max_pending = 0;
2350         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2351         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2352         ring->rx_mini_pending = 0;
2353         ring->rx_jumbo_pending = 0;
2354 }
2355
2356 static int ibmvnic_set_ringparam(struct net_device *netdev,
2357                                  struct ethtool_ringparam *ring)
2358 {
2359         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2360         int ret;
2361
2362         ret = 0;
2363         adapter->desired.rx_entries = ring->rx_pending;
2364         adapter->desired.tx_entries = ring->tx_pending;
2365
2366         ret = wait_for_reset(adapter);
2367
2368         if (!ret &&
2369             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2370              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2371                 netdev_info(netdev,
2372                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2373                             ring->rx_pending, ring->tx_pending,
2374                             adapter->req_rx_add_entries_per_subcrq,
2375                             adapter->req_tx_entries_per_subcrq);
2376         return ret;
2377 }
2378
2379 static void ibmvnic_get_channels(struct net_device *netdev,
2380                                  struct ethtool_channels *channels)
2381 {
2382         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2383
2384         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2385                 channels->max_rx = adapter->max_rx_queues;
2386                 channels->max_tx = adapter->max_tx_queues;
2387         } else {
2388                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2389                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2390         }
2391
2392         channels->max_other = 0;
2393         channels->max_combined = 0;
2394         channels->rx_count = adapter->req_rx_queues;
2395         channels->tx_count = adapter->req_tx_queues;
2396         channels->other_count = 0;
2397         channels->combined_count = 0;
2398 }
2399
2400 static int ibmvnic_set_channels(struct net_device *netdev,
2401                                 struct ethtool_channels *channels)
2402 {
2403         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2404         int ret;
2405
2406         ret = 0;
2407         adapter->desired.rx_queues = channels->rx_count;
2408         adapter->desired.tx_queues = channels->tx_count;
2409
2410         ret = wait_for_reset(adapter);
2411
2412         if (!ret &&
2413             (adapter->req_rx_queues != channels->rx_count ||
2414              adapter->req_tx_queues != channels->tx_count))
2415                 netdev_info(netdev,
2416                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2417                             channels->rx_count, channels->tx_count,
2418                             adapter->req_rx_queues, adapter->req_tx_queues);
2419         return ret;
2420
2421 }
2422
2423 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2424 {
2425         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2426         int i;
2427
2428         switch (stringset) {
2429         case ETH_SS_STATS:
2430                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2431                                 i++, data += ETH_GSTRING_LEN)
2432                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2433
2434                 for (i = 0; i < adapter->req_tx_queues; i++) {
2435                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2436                         data += ETH_GSTRING_LEN;
2437
2438                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2439                         data += ETH_GSTRING_LEN;
2440
2441                         snprintf(data, ETH_GSTRING_LEN,
2442                                  "tx%d_dropped_packets", i);
2443                         data += ETH_GSTRING_LEN;
2444                 }
2445
2446                 for (i = 0; i < adapter->req_rx_queues; i++) {
2447                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2448                         data += ETH_GSTRING_LEN;
2449
2450                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2451                         data += ETH_GSTRING_LEN;
2452
2453                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2454                         data += ETH_GSTRING_LEN;
2455                 }
2456                 break;
2457
2458         case ETH_SS_PRIV_FLAGS:
2459                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2460                         strcpy(data + i * ETH_GSTRING_LEN,
2461                                ibmvnic_priv_flags[i]);
2462                 break;
2463         default:
2464                 return;
2465         }
2466 }
2467
2468 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2469 {
2470         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2471
2472         switch (sset) {
2473         case ETH_SS_STATS:
2474                 return ARRAY_SIZE(ibmvnic_stats) +
2475                        adapter->req_tx_queues * NUM_TX_STATS +
2476                        adapter->req_rx_queues * NUM_RX_STATS;
2477         case ETH_SS_PRIV_FLAGS:
2478                 return ARRAY_SIZE(ibmvnic_priv_flags);
2479         default:
2480                 return -EOPNOTSUPP;
2481         }
2482 }
2483
2484 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2485                                       struct ethtool_stats *stats, u64 *data)
2486 {
2487         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2488         union ibmvnic_crq crq;
2489         int i, j;
2490         int rc;
2491
2492         memset(&crq, 0, sizeof(crq));
2493         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2494         crq.request_statistics.cmd = REQUEST_STATISTICS;
2495         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2496         crq.request_statistics.len =
2497             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2498
2499         /* Wait for data to be written */
2500         init_completion(&adapter->stats_done);
2501         rc = ibmvnic_send_crq(adapter, &crq);
2502         if (rc)
2503                 return;
2504         wait_for_completion(&adapter->stats_done);
2505
2506         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2507                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2508                                                 ibmvnic_stats[i].offset));
2509
2510         for (j = 0; j < adapter->req_tx_queues; j++) {
2511                 data[i] = adapter->tx_stats_buffers[j].packets;
2512                 i++;
2513                 data[i] = adapter->tx_stats_buffers[j].bytes;
2514                 i++;
2515                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2516                 i++;
2517         }
2518
2519         for (j = 0; j < adapter->req_rx_queues; j++) {
2520                 data[i] = adapter->rx_stats_buffers[j].packets;
2521                 i++;
2522                 data[i] = adapter->rx_stats_buffers[j].bytes;
2523                 i++;
2524                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2525                 i++;
2526         }
2527 }
2528
2529 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2530 {
2531         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2532
2533         return adapter->priv_flags;
2534 }
2535
2536 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2537 {
2538         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2539         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2540
2541         if (which_maxes)
2542                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2543         else
2544                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2545
2546         return 0;
2547 }
2548 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2549         .get_drvinfo            = ibmvnic_get_drvinfo,
2550         .get_msglevel           = ibmvnic_get_msglevel,
2551         .set_msglevel           = ibmvnic_set_msglevel,
2552         .get_link               = ibmvnic_get_link,
2553         .get_ringparam          = ibmvnic_get_ringparam,
2554         .set_ringparam          = ibmvnic_set_ringparam,
2555         .get_channels           = ibmvnic_get_channels,
2556         .set_channels           = ibmvnic_set_channels,
2557         .get_strings            = ibmvnic_get_strings,
2558         .get_sset_count         = ibmvnic_get_sset_count,
2559         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2560         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2561         .get_priv_flags         = ibmvnic_get_priv_flags,
2562         .set_priv_flags         = ibmvnic_set_priv_flags,
2563 };
2564
2565 /* Routines for managing CRQs/sCRQs  */
2566
2567 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2568                                    struct ibmvnic_sub_crq_queue *scrq)
2569 {
2570         int rc;
2571
2572         if (scrq->irq) {
2573                 free_irq(scrq->irq, scrq);
2574                 irq_dispose_mapping(scrq->irq);
2575                 scrq->irq = 0;
2576         }
2577
2578         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2579         atomic_set(&scrq->used, 0);
2580         scrq->cur = 0;
2581
2582         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2583                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2584         return rc;
2585 }
2586
2587 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2588 {
2589         int i, rc;
2590
2591         for (i = 0; i < adapter->req_tx_queues; i++) {
2592                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2593                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2594                 if (rc)
2595                         return rc;
2596         }
2597
2598         for (i = 0; i < adapter->req_rx_queues; i++) {
2599                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2600                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2601                 if (rc)
2602                         return rc;
2603         }
2604
2605         return rc;
2606 }
2607
2608 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2609                                   struct ibmvnic_sub_crq_queue *scrq,
2610                                   bool do_h_free)
2611 {
2612         struct device *dev = &adapter->vdev->dev;
2613         long rc;
2614
2615         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2616
2617         if (do_h_free) {
2618                 /* Close the sub-crqs */
2619                 do {
2620                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2621                                                 adapter->vdev->unit_address,
2622                                                 scrq->crq_num);
2623                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2624
2625                 if (rc) {
2626                         netdev_err(adapter->netdev,
2627                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2628                                    scrq->crq_num, rc);
2629                 }
2630         }
2631
2632         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2633                          DMA_BIDIRECTIONAL);
2634         free_pages((unsigned long)scrq->msgs, 2);
2635         kfree(scrq);
2636 }
2637
2638 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2639                                                         *adapter)
2640 {
2641         struct device *dev = &adapter->vdev->dev;
2642         struct ibmvnic_sub_crq_queue *scrq;
2643         int rc;
2644
2645         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2646         if (!scrq)
2647                 return NULL;
2648
2649         scrq->msgs =
2650                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2651         if (!scrq->msgs) {
2652                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2653                 goto zero_page_failed;
2654         }
2655
2656         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2657                                          DMA_BIDIRECTIONAL);
2658         if (dma_mapping_error(dev, scrq->msg_token)) {
2659                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2660                 goto map_failed;
2661         }
2662
2663         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2664                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2665
2666         if (rc == H_RESOURCE)
2667                 rc = ibmvnic_reset_crq(adapter);
2668
2669         if (rc == H_CLOSED) {
2670                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2671         } else if (rc) {
2672                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2673                 goto reg_failed;
2674         }
2675
2676         scrq->adapter = adapter;
2677         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2678         spin_lock_init(&scrq->lock);
2679
2680         netdev_dbg(adapter->netdev,
2681                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2682                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2683
2684         return scrq;
2685
2686 reg_failed:
2687         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2688                          DMA_BIDIRECTIONAL);
2689 map_failed:
2690         free_pages((unsigned long)scrq->msgs, 2);
2691 zero_page_failed:
2692         kfree(scrq);
2693
2694         return NULL;
2695 }
2696
2697 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2698 {
2699         int i;
2700
2701         if (adapter->tx_scrq) {
2702                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2703                         if (!adapter->tx_scrq[i])
2704                                 continue;
2705
2706                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2707                                    i);
2708                         if (adapter->tx_scrq[i]->irq) {
2709                                 free_irq(adapter->tx_scrq[i]->irq,
2710                                          adapter->tx_scrq[i]);
2711                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2712                                 adapter->tx_scrq[i]->irq = 0;
2713                         }
2714
2715                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2716                                               do_h_free);
2717                 }
2718
2719                 kfree(adapter->tx_scrq);
2720                 adapter->tx_scrq = NULL;
2721                 adapter->num_active_tx_scrqs = 0;
2722         }
2723
2724         if (adapter->rx_scrq) {
2725                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2726                         if (!adapter->rx_scrq[i])
2727                                 continue;
2728
2729                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2730                                    i);
2731                         if (adapter->rx_scrq[i]->irq) {
2732                                 free_irq(adapter->rx_scrq[i]->irq,
2733                                          adapter->rx_scrq[i]);
2734                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2735                                 adapter->rx_scrq[i]->irq = 0;
2736                         }
2737
2738                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2739                                               do_h_free);
2740                 }
2741
2742                 kfree(adapter->rx_scrq);
2743                 adapter->rx_scrq = NULL;
2744                 adapter->num_active_rx_scrqs = 0;
2745         }
2746 }
2747
2748 static int disable_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         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2755                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2756         if (rc)
2757                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2758                         scrq->hw_irq, rc);
2759         return rc;
2760 }
2761
2762 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2763                            struct ibmvnic_sub_crq_queue *scrq)
2764 {
2765         struct device *dev = &adapter->vdev->dev;
2766         unsigned long rc;
2767
2768         if (scrq->hw_irq > 0x100000000ULL) {
2769                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2770                 return 1;
2771         }
2772
2773         if (adapter->resetting &&
2774             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2775                 u64 val = (0xff000000) | scrq->hw_irq;
2776
2777                 rc = plpar_hcall_norets(H_EOI, val);
2778                 if (rc)
2779                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2780                                 val, rc);
2781         }
2782
2783         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2784                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2785         if (rc)
2786                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2787                         scrq->hw_irq, rc);
2788         return rc;
2789 }
2790
2791 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2792                                struct ibmvnic_sub_crq_queue *scrq)
2793 {
2794         struct device *dev = &adapter->vdev->dev;
2795         struct ibmvnic_tx_pool *tx_pool;
2796         struct ibmvnic_tx_buff *txbuff;
2797         union sub_crq *next;
2798         int index;
2799         int i, j;
2800
2801 restart_loop:
2802         while (pending_scrq(adapter, scrq)) {
2803                 unsigned int pool = scrq->pool_index;
2804                 int num_entries = 0;
2805
2806                 next = ibmvnic_next_scrq(adapter, scrq);
2807                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2808                         if (next->tx_comp.rcs[i]) {
2809                                 dev_err(dev, "tx error %x\n",
2810                                         next->tx_comp.rcs[i]);
2811                                 continue;
2812                         }
2813                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2814                         if (index & IBMVNIC_TSO_POOL_MASK) {
2815                                 tx_pool = &adapter->tso_pool[pool];
2816                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2817                         } else {
2818                                 tx_pool = &adapter->tx_pool[pool];
2819                         }
2820
2821                         txbuff = &tx_pool->tx_buff[index];
2822
2823                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2824                                 if (!txbuff->data_dma[j])
2825                                         continue;
2826
2827                                 txbuff->data_dma[j] = 0;
2828                         }
2829
2830                         if (txbuff->last_frag) {
2831                                 dev_kfree_skb_any(txbuff->skb);
2832                                 txbuff->skb = NULL;
2833                         }
2834
2835                         num_entries += txbuff->num_entries;
2836
2837                         tx_pool->free_map[tx_pool->producer_index] = index;
2838                         tx_pool->producer_index =
2839                                 (tx_pool->producer_index + 1) %
2840                                         tx_pool->num_buffers;
2841                 }
2842                 /* remove tx_comp scrq*/
2843                 next->tx_comp.first = 0;
2844
2845                 if (atomic_sub_return(num_entries, &scrq->used) <=
2846                     (adapter->req_tx_entries_per_subcrq / 2) &&
2847                     __netif_subqueue_stopped(adapter->netdev,
2848                                              scrq->pool_index)) {
2849                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2850                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2851                                    scrq->pool_index);
2852                 }
2853         }
2854
2855         enable_scrq_irq(adapter, scrq);
2856
2857         if (pending_scrq(adapter, scrq)) {
2858                 disable_scrq_irq(adapter, scrq);
2859                 goto restart_loop;
2860         }
2861
2862         return 0;
2863 }
2864
2865 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2866 {
2867         struct ibmvnic_sub_crq_queue *scrq = instance;
2868         struct ibmvnic_adapter *adapter = scrq->adapter;
2869
2870         disable_scrq_irq(adapter, scrq);
2871         ibmvnic_complete_tx(adapter, scrq);
2872
2873         return IRQ_HANDLED;
2874 }
2875
2876 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2877 {
2878         struct ibmvnic_sub_crq_queue *scrq = instance;
2879         struct ibmvnic_adapter *adapter = scrq->adapter;
2880
2881         /* When booting a kdump kernel we can hit pending interrupts
2882          * prior to completing driver initialization.
2883          */
2884         if (unlikely(adapter->state != VNIC_OPEN))
2885                 return IRQ_NONE;
2886
2887         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2888
2889         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2890                 disable_scrq_irq(adapter, scrq);
2891                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2892         }
2893
2894         return IRQ_HANDLED;
2895 }
2896
2897 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2898 {
2899         struct device *dev = &adapter->vdev->dev;
2900         struct ibmvnic_sub_crq_queue *scrq;
2901         int i = 0, j = 0;
2902         int rc = 0;
2903
2904         for (i = 0; i < adapter->req_tx_queues; i++) {
2905                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2906                            i);
2907                 scrq = adapter->tx_scrq[i];
2908                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2909
2910                 if (!scrq->irq) {
2911                         rc = -EINVAL;
2912                         dev_err(dev, "Error mapping irq\n");
2913                         goto req_tx_irq_failed;
2914                 }
2915
2916                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
2917                          adapter->vdev->unit_address, i);
2918                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2919                                  0, scrq->name, scrq);
2920
2921                 if (rc) {
2922                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2923                                 scrq->irq, rc);
2924                         irq_dispose_mapping(scrq->irq);
2925                         goto req_tx_irq_failed;
2926                 }
2927         }
2928
2929         for (i = 0; i < adapter->req_rx_queues; i++) {
2930                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2931                            i);
2932                 scrq = adapter->rx_scrq[i];
2933                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2934                 if (!scrq->irq) {
2935                         rc = -EINVAL;
2936                         dev_err(dev, "Error mapping irq\n");
2937                         goto req_rx_irq_failed;
2938                 }
2939                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
2940                          adapter->vdev->unit_address, i);
2941                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2942                                  0, scrq->name, scrq);
2943                 if (rc) {
2944                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2945                                 scrq->irq, rc);
2946                         irq_dispose_mapping(scrq->irq);
2947                         goto req_rx_irq_failed;
2948                 }
2949         }
2950         return rc;
2951
2952 req_rx_irq_failed:
2953         for (j = 0; j < i; j++) {
2954                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2955                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2956         }
2957         i = adapter->req_tx_queues;
2958 req_tx_irq_failed:
2959         for (j = 0; j < i; j++) {
2960                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2961                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2962         }
2963         release_sub_crqs(adapter, 1);
2964         return rc;
2965 }
2966
2967 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2968 {
2969         struct device *dev = &adapter->vdev->dev;
2970         struct ibmvnic_sub_crq_queue **allqueues;
2971         int registered_queues = 0;
2972         int total_queues;
2973         int more = 0;
2974         int i;
2975
2976         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2977
2978         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2979         if (!allqueues)
2980                 return -1;
2981
2982         for (i = 0; i < total_queues; i++) {
2983                 allqueues[i] = init_sub_crq_queue(adapter);
2984                 if (!allqueues[i]) {
2985                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2986                         break;
2987                 }
2988                 registered_queues++;
2989         }
2990
2991         /* Make sure we were able to register the minimum number of queues */
2992         if (registered_queues <
2993             adapter->min_tx_queues + adapter->min_rx_queues) {
2994                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2995                 goto tx_failed;
2996         }
2997
2998         /* Distribute the failed allocated queues*/
2999         for (i = 0; i < total_queues - registered_queues + more ; i++) {
3000                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3001                 switch (i % 3) {
3002                 case 0:
3003                         if (adapter->req_rx_queues > adapter->min_rx_queues)
3004                                 adapter->req_rx_queues--;
3005                         else
3006                                 more++;
3007                         break;
3008                 case 1:
3009                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3010                                 adapter->req_tx_queues--;
3011                         else
3012                                 more++;
3013                         break;
3014                 }
3015         }
3016
3017         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3018                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3019         if (!adapter->tx_scrq)
3020                 goto tx_failed;
3021
3022         for (i = 0; i < adapter->req_tx_queues; i++) {
3023                 adapter->tx_scrq[i] = allqueues[i];
3024                 adapter->tx_scrq[i]->pool_index = i;
3025                 adapter->num_active_tx_scrqs++;
3026         }
3027
3028         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3029                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3030         if (!adapter->rx_scrq)
3031                 goto rx_failed;
3032
3033         for (i = 0; i < adapter->req_rx_queues; i++) {
3034                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3035                 adapter->rx_scrq[i]->scrq_num = i;
3036                 adapter->num_active_rx_scrqs++;
3037         }
3038
3039         kfree(allqueues);
3040         return 0;
3041
3042 rx_failed:
3043         kfree(adapter->tx_scrq);
3044         adapter->tx_scrq = NULL;
3045 tx_failed:
3046         for (i = 0; i < registered_queues; i++)
3047                 release_sub_crq_queue(adapter, allqueues[i], 1);
3048         kfree(allqueues);
3049         return -1;
3050 }
3051
3052 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3053 {
3054         struct device *dev = &adapter->vdev->dev;
3055         union ibmvnic_crq crq;
3056         int max_entries;
3057
3058         if (!retry) {
3059                 /* Sub-CRQ entries are 32 byte long */
3060                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3061
3062                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3063                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3064                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3065                         return;
3066                 }
3067
3068                 if (adapter->desired.mtu)
3069                         adapter->req_mtu = adapter->desired.mtu;
3070                 else
3071                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3072
3073                 if (!adapter->desired.tx_entries)
3074                         adapter->desired.tx_entries =
3075                                         adapter->max_tx_entries_per_subcrq;
3076                 if (!adapter->desired.rx_entries)
3077                         adapter->desired.rx_entries =
3078                                         adapter->max_rx_add_entries_per_subcrq;
3079
3080                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3081                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3082
3083                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3084                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3085                         adapter->desired.tx_entries = max_entries;
3086                 }
3087
3088                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3089                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3090                         adapter->desired.rx_entries = max_entries;
3091                 }
3092
3093                 if (adapter->desired.tx_entries)
3094                         adapter->req_tx_entries_per_subcrq =
3095                                         adapter->desired.tx_entries;
3096                 else
3097                         adapter->req_tx_entries_per_subcrq =
3098                                         adapter->max_tx_entries_per_subcrq;
3099
3100                 if (adapter->desired.rx_entries)
3101                         adapter->req_rx_add_entries_per_subcrq =
3102                                         adapter->desired.rx_entries;
3103                 else
3104                         adapter->req_rx_add_entries_per_subcrq =
3105                                         adapter->max_rx_add_entries_per_subcrq;
3106
3107                 if (adapter->desired.tx_queues)
3108                         adapter->req_tx_queues =
3109                                         adapter->desired.tx_queues;
3110                 else
3111                         adapter->req_tx_queues =
3112                                         adapter->opt_tx_comp_sub_queues;
3113
3114                 if (adapter->desired.rx_queues)
3115                         adapter->req_rx_queues =
3116                                         adapter->desired.rx_queues;
3117                 else
3118                         adapter->req_rx_queues =
3119                                         adapter->opt_rx_comp_queues;
3120
3121                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3122         }
3123
3124         memset(&crq, 0, sizeof(crq));
3125         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3126         crq.request_capability.cmd = REQUEST_CAPABILITY;
3127
3128         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3129         crq.request_capability.number = cpu_to_be64(adapter->req_tx_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_QUEUES);
3134         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3135         atomic_inc(&adapter->running_cap_crqs);
3136         ibmvnic_send_crq(adapter, &crq);
3137
3138         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3139         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3140         atomic_inc(&adapter->running_cap_crqs);
3141         ibmvnic_send_crq(adapter, &crq);
3142
3143         crq.request_capability.capability =
3144             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3145         crq.request_capability.number =
3146             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3147         atomic_inc(&adapter->running_cap_crqs);
3148         ibmvnic_send_crq(adapter, &crq);
3149
3150         crq.request_capability.capability =
3151             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3152         crq.request_capability.number =
3153             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3154         atomic_inc(&adapter->running_cap_crqs);
3155         ibmvnic_send_crq(adapter, &crq);
3156
3157         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3158         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3159         atomic_inc(&adapter->running_cap_crqs);
3160         ibmvnic_send_crq(adapter, &crq);
3161
3162         if (adapter->netdev->flags & IFF_PROMISC) {
3163                 if (adapter->promisc_supported) {
3164                         crq.request_capability.capability =
3165                             cpu_to_be16(PROMISC_REQUESTED);
3166                         crq.request_capability.number = cpu_to_be64(1);
3167                         atomic_inc(&adapter->running_cap_crqs);
3168                         ibmvnic_send_crq(adapter, &crq);
3169                 }
3170         } else {
3171                 crq.request_capability.capability =
3172                     cpu_to_be16(PROMISC_REQUESTED);
3173                 crq.request_capability.number = cpu_to_be64(0);
3174                 atomic_inc(&adapter->running_cap_crqs);
3175                 ibmvnic_send_crq(adapter, &crq);
3176         }
3177 }
3178
3179 static int pending_scrq(struct ibmvnic_adapter *adapter,
3180                         struct ibmvnic_sub_crq_queue *scrq)
3181 {
3182         union sub_crq *entry = &scrq->msgs[scrq->cur];
3183
3184         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3185                 return 1;
3186         else
3187                 return 0;
3188 }
3189
3190 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3191                                         struct ibmvnic_sub_crq_queue *scrq)
3192 {
3193         union sub_crq *entry;
3194         unsigned long flags;
3195
3196         spin_lock_irqsave(&scrq->lock, flags);
3197         entry = &scrq->msgs[scrq->cur];
3198         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3199                 if (++scrq->cur == scrq->size)
3200                         scrq->cur = 0;
3201         } else {
3202                 entry = NULL;
3203         }
3204         spin_unlock_irqrestore(&scrq->lock, flags);
3205
3206         return entry;
3207 }
3208
3209 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3210 {
3211         struct ibmvnic_crq_queue *queue = &adapter->crq;
3212         union ibmvnic_crq *crq;
3213
3214         crq = &queue->msgs[queue->cur];
3215         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3216                 if (++queue->cur == queue->size)
3217                         queue->cur = 0;
3218         } else {
3219                 crq = NULL;
3220         }
3221
3222         return crq;
3223 }
3224
3225 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3226 {
3227         switch (rc) {
3228         case H_PARAMETER:
3229                 dev_warn_ratelimited(dev,
3230                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3231                                      func, rc);
3232                 break;
3233         case H_CLOSED:
3234                 dev_warn_ratelimited(dev,
3235                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3236                                      func, rc);
3237                 break;
3238         default:
3239                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3240                 break;
3241         }
3242 }
3243
3244 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3245                        union sub_crq *sub_crq)
3246 {
3247         unsigned int ua = adapter->vdev->unit_address;
3248         struct device *dev = &adapter->vdev->dev;
3249         u64 *u64_crq = (u64 *)sub_crq;
3250         int rc;
3251
3252         netdev_dbg(adapter->netdev,
3253                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3254                    (unsigned long int)cpu_to_be64(remote_handle),
3255                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3256                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3257                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3258                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3259
3260         /* Make sure the hypervisor sees the complete request */
3261         mb();
3262
3263         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3264                                 cpu_to_be64(remote_handle),
3265                                 cpu_to_be64(u64_crq[0]),
3266                                 cpu_to_be64(u64_crq[1]),
3267                                 cpu_to_be64(u64_crq[2]),
3268                                 cpu_to_be64(u64_crq[3]));
3269
3270         if (rc)
3271                 print_subcrq_error(dev, rc, __func__);
3272
3273         return rc;
3274 }
3275
3276 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3277                                 u64 remote_handle, u64 ioba, u64 num_entries)
3278 {
3279         unsigned int ua = adapter->vdev->unit_address;
3280         struct device *dev = &adapter->vdev->dev;
3281         int rc;
3282
3283         /* Make sure the hypervisor sees the complete request */
3284         mb();
3285         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3286                                 cpu_to_be64(remote_handle),
3287                                 ioba, num_entries);
3288
3289         if (rc)
3290                 print_subcrq_error(dev, rc, __func__);
3291
3292         return rc;
3293 }
3294
3295 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3296                             union ibmvnic_crq *crq)
3297 {
3298         unsigned int ua = adapter->vdev->unit_address;
3299         struct device *dev = &adapter->vdev->dev;
3300         u64 *u64_crq = (u64 *)crq;
3301         int rc;
3302
3303         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3304                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3305                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3306
3307         if (!adapter->crq.active &&
3308             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3309                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3310                 return -EINVAL;
3311         }
3312
3313         /* Make sure the hypervisor sees the complete request */
3314         mb();
3315
3316         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3317                                 cpu_to_be64(u64_crq[0]),
3318                                 cpu_to_be64(u64_crq[1]));
3319
3320         if (rc) {
3321                 if (rc == H_CLOSED) {
3322                         dev_warn(dev, "CRQ Queue closed\n");
3323                         if (adapter->resetting)
3324                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3325                 }
3326
3327                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3328         }
3329
3330         return rc;
3331 }
3332
3333 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3334 {
3335         union ibmvnic_crq crq;
3336
3337         memset(&crq, 0, sizeof(crq));
3338         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3339         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3340         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3341
3342         return ibmvnic_send_crq(adapter, &crq);
3343 }
3344
3345 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3346 {
3347         union ibmvnic_crq crq;
3348
3349         memset(&crq, 0, sizeof(crq));
3350         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3351         crq.version_exchange.cmd = VERSION_EXCHANGE;
3352         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3353
3354         return ibmvnic_send_crq(adapter, &crq);
3355 }
3356
3357 struct vnic_login_client_data {
3358         u8      type;
3359         __be16  len;
3360         char    name[];
3361 } __packed;
3362
3363 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3364 {
3365         int len;
3366
3367         /* Calculate the amount of buffer space needed for the
3368          * vnic client data in the login buffer. There are four entries,
3369          * OS name, LPAR name, device name, and a null last entry.
3370          */
3371         len = 4 * sizeof(struct vnic_login_client_data);
3372         len += 6; /* "Linux" plus NULL */
3373         len += strlen(utsname()->nodename) + 1;
3374         len += strlen(adapter->netdev->name) + 1;
3375
3376         return len;
3377 }
3378
3379 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3380                                  struct vnic_login_client_data *vlcd)
3381 {
3382         const char *os_name = "Linux";
3383         int len;
3384
3385         /* Type 1 - LPAR OS */
3386         vlcd->type = 1;
3387         len = strlen(os_name) + 1;
3388         vlcd->len = cpu_to_be16(len);
3389         strncpy(vlcd->name, os_name, len);
3390         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3391
3392         /* Type 2 - LPAR name */
3393         vlcd->type = 2;
3394         len = strlen(utsname()->nodename) + 1;
3395         vlcd->len = cpu_to_be16(len);
3396         strncpy(vlcd->name, utsname()->nodename, len);
3397         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3398
3399         /* Type 3 - device name */
3400         vlcd->type = 3;
3401         len = strlen(adapter->netdev->name) + 1;
3402         vlcd->len = cpu_to_be16(len);
3403         strncpy(vlcd->name, adapter->netdev->name, len);
3404 }
3405
3406 static int send_login(struct ibmvnic_adapter *adapter)
3407 {
3408         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3409         struct ibmvnic_login_buffer *login_buffer;
3410         struct device *dev = &adapter->vdev->dev;
3411         dma_addr_t rsp_buffer_token;
3412         dma_addr_t buffer_token;
3413         size_t rsp_buffer_size;
3414         union ibmvnic_crq crq;
3415         size_t buffer_size;
3416         __be64 *tx_list_p;
3417         __be64 *rx_list_p;
3418         int client_data_len;
3419         struct vnic_login_client_data *vlcd;
3420         int i;
3421
3422         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3423                 netdev_err(adapter->netdev,
3424                            "RX or TX queues are not allocated, device login failed\n");
3425                 return -1;
3426         }
3427
3428         release_login_rsp_buffer(adapter);
3429         client_data_len = vnic_client_data_len(adapter);
3430
3431         buffer_size =
3432             sizeof(struct ibmvnic_login_buffer) +
3433             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3434             client_data_len;
3435
3436         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3437         if (!login_buffer)
3438                 goto buf_alloc_failed;
3439
3440         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3441                                       DMA_TO_DEVICE);
3442         if (dma_mapping_error(dev, buffer_token)) {
3443                 dev_err(dev, "Couldn't map login buffer\n");
3444                 goto buf_map_failed;
3445         }
3446
3447         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3448                           sizeof(u64) * adapter->req_tx_queues +
3449                           sizeof(u64) * adapter->req_rx_queues +
3450                           sizeof(u64) * adapter->req_rx_queues +
3451                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3452
3453         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3454         if (!login_rsp_buffer)
3455                 goto buf_rsp_alloc_failed;
3456
3457         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3458                                           rsp_buffer_size, DMA_FROM_DEVICE);
3459         if (dma_mapping_error(dev, rsp_buffer_token)) {
3460                 dev_err(dev, "Couldn't map login rsp buffer\n");
3461                 goto buf_rsp_map_failed;
3462         }
3463
3464         adapter->login_buf = login_buffer;
3465         adapter->login_buf_token = buffer_token;
3466         adapter->login_buf_sz = buffer_size;
3467         adapter->login_rsp_buf = login_rsp_buffer;
3468         adapter->login_rsp_buf_token = rsp_buffer_token;
3469         adapter->login_rsp_buf_sz = rsp_buffer_size;
3470
3471         login_buffer->len = cpu_to_be32(buffer_size);
3472         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3473         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3474         login_buffer->off_txcomp_subcrqs =
3475             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3476         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3477         login_buffer->off_rxcomp_subcrqs =
3478             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3479                         sizeof(u64) * adapter->req_tx_queues);
3480         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3481         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3482
3483         tx_list_p = (__be64 *)((char *)login_buffer +
3484                                       sizeof(struct ibmvnic_login_buffer));
3485         rx_list_p = (__be64 *)((char *)login_buffer +
3486                                       sizeof(struct ibmvnic_login_buffer) +
3487                                       sizeof(u64) * adapter->req_tx_queues);
3488
3489         for (i = 0; i < adapter->req_tx_queues; i++) {
3490                 if (adapter->tx_scrq[i]) {
3491                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3492                                                    crq_num);
3493                 }
3494         }
3495
3496         for (i = 0; i < adapter->req_rx_queues; i++) {
3497                 if (adapter->rx_scrq[i]) {
3498                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3499                                                    crq_num);
3500                 }
3501         }
3502
3503         /* Insert vNIC login client data */
3504         vlcd = (struct vnic_login_client_data *)
3505                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3506         login_buffer->client_data_offset =
3507                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3508         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3509
3510         vnic_add_client_data(adapter, vlcd);
3511
3512         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3513         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3514                 netdev_dbg(adapter->netdev, "%016lx\n",
3515                            ((unsigned long int *)(adapter->login_buf))[i]);
3516         }
3517
3518         memset(&crq, 0, sizeof(crq));
3519         crq.login.first = IBMVNIC_CRQ_CMD;
3520         crq.login.cmd = LOGIN;
3521         crq.login.ioba = cpu_to_be32(buffer_token);
3522         crq.login.len = cpu_to_be32(buffer_size);
3523         ibmvnic_send_crq(adapter, &crq);
3524
3525         return 0;
3526
3527 buf_rsp_map_failed:
3528         kfree(login_rsp_buffer);
3529 buf_rsp_alloc_failed:
3530         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3531 buf_map_failed:
3532         kfree(login_buffer);
3533 buf_alloc_failed:
3534         return -1;
3535 }
3536
3537 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3538                             u32 len, u8 map_id)
3539 {
3540         union ibmvnic_crq crq;
3541
3542         memset(&crq, 0, sizeof(crq));
3543         crq.request_map.first = IBMVNIC_CRQ_CMD;
3544         crq.request_map.cmd = REQUEST_MAP;
3545         crq.request_map.map_id = map_id;
3546         crq.request_map.ioba = cpu_to_be32(addr);
3547         crq.request_map.len = cpu_to_be32(len);
3548         return ibmvnic_send_crq(adapter, &crq);
3549 }
3550
3551 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3552 {
3553         union ibmvnic_crq crq;
3554
3555         memset(&crq, 0, sizeof(crq));
3556         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3557         crq.request_unmap.cmd = REQUEST_UNMAP;
3558         crq.request_unmap.map_id = map_id;
3559         return ibmvnic_send_crq(adapter, &crq);
3560 }
3561
3562 static void send_map_query(struct ibmvnic_adapter *adapter)
3563 {
3564         union ibmvnic_crq crq;
3565
3566         memset(&crq, 0, sizeof(crq));
3567         crq.query_map.first = IBMVNIC_CRQ_CMD;
3568         crq.query_map.cmd = QUERY_MAP;
3569         ibmvnic_send_crq(adapter, &crq);
3570 }
3571
3572 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3573 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3574 {
3575         union ibmvnic_crq crq;
3576
3577         atomic_set(&adapter->running_cap_crqs, 0);
3578         memset(&crq, 0, sizeof(crq));
3579         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3580         crq.query_capability.cmd = QUERY_CAPABILITY;
3581
3582         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3583         atomic_inc(&adapter->running_cap_crqs);
3584         ibmvnic_send_crq(adapter, &crq);
3585
3586         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3587         atomic_inc(&adapter->running_cap_crqs);
3588         ibmvnic_send_crq(adapter, &crq);
3589
3590         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3591         atomic_inc(&adapter->running_cap_crqs);
3592         ibmvnic_send_crq(adapter, &crq);
3593
3594         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3595         atomic_inc(&adapter->running_cap_crqs);
3596         ibmvnic_send_crq(adapter, &crq);
3597
3598         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3599         atomic_inc(&adapter->running_cap_crqs);
3600         ibmvnic_send_crq(adapter, &crq);
3601
3602         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3603         atomic_inc(&adapter->running_cap_crqs);
3604         ibmvnic_send_crq(adapter, &crq);
3605
3606         crq.query_capability.capability =
3607             cpu_to_be16(MIN_TX_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(MIN_RX_ADD_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_TX_ENTRIES_PER_SUBCRQ);
3618         atomic_inc(&adapter->running_cap_crqs);
3619         ibmvnic_send_crq(adapter, &crq);
3620
3621         crq.query_capability.capability =
3622             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3623         atomic_inc(&adapter->running_cap_crqs);
3624         ibmvnic_send_crq(adapter, &crq);
3625
3626         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3627         atomic_inc(&adapter->running_cap_crqs);
3628         ibmvnic_send_crq(adapter, &crq);
3629
3630         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3631         atomic_inc(&adapter->running_cap_crqs);
3632         ibmvnic_send_crq(adapter, &crq);
3633
3634         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3635         atomic_inc(&adapter->running_cap_crqs);
3636         ibmvnic_send_crq(adapter, &crq);
3637
3638         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3639         atomic_inc(&adapter->running_cap_crqs);
3640         ibmvnic_send_crq(adapter, &crq);
3641
3642         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3643         atomic_inc(&adapter->running_cap_crqs);
3644         ibmvnic_send_crq(adapter, &crq);
3645
3646         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3647         atomic_inc(&adapter->running_cap_crqs);
3648         ibmvnic_send_crq(adapter, &crq);
3649
3650         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3651         atomic_inc(&adapter->running_cap_crqs);
3652         ibmvnic_send_crq(adapter, &crq);
3653
3654         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3655         atomic_inc(&adapter->running_cap_crqs);
3656         ibmvnic_send_crq(adapter, &crq);
3657
3658         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3659         atomic_inc(&adapter->running_cap_crqs);
3660         ibmvnic_send_crq(adapter, &crq);
3661
3662         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3663         atomic_inc(&adapter->running_cap_crqs);
3664         ibmvnic_send_crq(adapter, &crq);
3665
3666         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3667         atomic_inc(&adapter->running_cap_crqs);
3668         ibmvnic_send_crq(adapter, &crq);
3669
3670         crq.query_capability.capability =
3671                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3672         atomic_inc(&adapter->running_cap_crqs);
3673         ibmvnic_send_crq(adapter, &crq);
3674
3675         crq.query_capability.capability =
3676                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3677         atomic_inc(&adapter->running_cap_crqs);
3678         ibmvnic_send_crq(adapter, &crq);
3679
3680         crq.query_capability.capability =
3681                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3682         atomic_inc(&adapter->running_cap_crqs);
3683         ibmvnic_send_crq(adapter, &crq);
3684
3685         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3686         atomic_inc(&adapter->running_cap_crqs);
3687         ibmvnic_send_crq(adapter, &crq);
3688 }
3689
3690 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3691                                 struct ibmvnic_adapter *adapter)
3692 {
3693         struct device *dev = &adapter->vdev->dev;
3694
3695         if (crq->get_vpd_size_rsp.rc.code) {
3696                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3697                         crq->get_vpd_size_rsp.rc.code);
3698                 complete(&adapter->fw_done);
3699                 return;
3700         }
3701
3702         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3703         complete(&adapter->fw_done);
3704 }
3705
3706 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3707                            struct ibmvnic_adapter *adapter)
3708 {
3709         struct device *dev = &adapter->vdev->dev;
3710         unsigned char *substr = NULL;
3711         u8 fw_level_len = 0;
3712
3713         memset(adapter->fw_version, 0, 32);
3714
3715         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3716                          DMA_FROM_DEVICE);
3717
3718         if (crq->get_vpd_rsp.rc.code) {
3719                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3720                         crq->get_vpd_rsp.rc.code);
3721                 goto complete;
3722         }
3723
3724         /* get the position of the firmware version info
3725          * located after the ASCII 'RM' substring in the buffer
3726          */
3727         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3728         if (!substr) {
3729                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3730                 goto complete;
3731         }
3732
3733         /* get length of firmware level ASCII substring */
3734         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3735                 fw_level_len = *(substr + 2);
3736         } else {
3737                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3738                 goto complete;
3739         }
3740
3741         /* copy firmware version string from vpd into adapter */
3742         if ((substr + 3 + fw_level_len) <
3743             (adapter->vpd->buff + adapter->vpd->len)) {
3744                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3745         } else {
3746                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3747         }
3748
3749 complete:
3750         if (adapter->fw_version[0] == '\0')
3751                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3752         complete(&adapter->fw_done);
3753 }
3754
3755 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3756 {
3757         struct device *dev = &adapter->vdev->dev;
3758         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3759         netdev_features_t old_hw_features = 0;
3760         union ibmvnic_crq crq;
3761         int i;
3762
3763         dma_unmap_single(dev, adapter->ip_offload_tok,
3764                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3765
3766         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3767         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3768                 netdev_dbg(adapter->netdev, "%016lx\n",
3769                            ((unsigned long int *)(buf))[i]);
3770
3771         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3772         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3773         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3774                    buf->tcp_ipv4_chksum);
3775         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3776                    buf->tcp_ipv6_chksum);
3777         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3778                    buf->udp_ipv4_chksum);
3779         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3780                    buf->udp_ipv6_chksum);
3781         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3782                    buf->large_tx_ipv4);
3783         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3784                    buf->large_tx_ipv6);
3785         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3786                    buf->large_rx_ipv4);
3787         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3788                    buf->large_rx_ipv6);
3789         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3790                    buf->max_ipv4_header_size);
3791         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3792                    buf->max_ipv6_header_size);
3793         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3794                    buf->max_tcp_header_size);
3795         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3796                    buf->max_udp_header_size);
3797         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3798                    buf->max_large_tx_size);
3799         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3800                    buf->max_large_rx_size);
3801         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3802                    buf->ipv6_extension_header);
3803         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3804                    buf->tcp_pseudosum_req);
3805         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3806                    buf->num_ipv6_ext_headers);
3807         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3808                    buf->off_ipv6_ext_headers);
3809
3810         adapter->ip_offload_ctrl_tok =
3811             dma_map_single(dev, &adapter->ip_offload_ctrl,
3812                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3813
3814         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3815                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3816                 return;
3817         }
3818
3819         adapter->ip_offload_ctrl.len =
3820             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3821         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3822         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3823         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3824         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3825         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3826         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3827         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3828         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3829         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3830
3831         /* large_rx disabled for now, additional features needed */
3832         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3833         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3834
3835         if (adapter->state != VNIC_PROBING) {
3836                 old_hw_features = adapter->netdev->hw_features;
3837                 adapter->netdev->hw_features = 0;
3838         }
3839
3840         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
3841
3842         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3843                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
3844
3845         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3846                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
3847
3848         if ((adapter->netdev->features &
3849             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3850                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
3851
3852         if (buf->large_tx_ipv4)
3853                 adapter->netdev->hw_features |= NETIF_F_TSO;
3854         if (buf->large_tx_ipv6)
3855                 adapter->netdev->hw_features |= NETIF_F_TSO6;
3856
3857         if (adapter->state == VNIC_PROBING) {
3858                 adapter->netdev->features |= adapter->netdev->hw_features;
3859         } else if (old_hw_features != adapter->netdev->hw_features) {
3860                 netdev_features_t tmp = 0;
3861
3862                 /* disable features no longer supported */
3863                 adapter->netdev->features &= adapter->netdev->hw_features;
3864                 /* turn on features now supported if previously enabled */
3865                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
3866                         adapter->netdev->hw_features;
3867                 adapter->netdev->features |=
3868                                 tmp & adapter->netdev->wanted_features;
3869         }
3870
3871         memset(&crq, 0, sizeof(crq));
3872         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3873         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3874         crq.control_ip_offload.len =
3875             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3876         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3877         ibmvnic_send_crq(adapter, &crq);
3878 }
3879
3880 static const char *ibmvnic_fw_err_cause(u16 cause)
3881 {
3882         switch (cause) {
3883         case ADAPTER_PROBLEM:
3884                 return "adapter problem";
3885         case BUS_PROBLEM:
3886                 return "bus problem";
3887         case FW_PROBLEM:
3888                 return "firmware problem";
3889         case DD_PROBLEM:
3890                 return "device driver problem";
3891         case EEH_RECOVERY:
3892                 return "EEH recovery";
3893         case FW_UPDATED:
3894                 return "firmware updated";
3895         case LOW_MEMORY:
3896                 return "low Memory";
3897         default:
3898                 return "unknown";
3899         }
3900 }
3901
3902 static void handle_error_indication(union ibmvnic_crq *crq,
3903                                     struct ibmvnic_adapter *adapter)
3904 {
3905         struct device *dev = &adapter->vdev->dev;
3906         u16 cause;
3907
3908         cause = be16_to_cpu(crq->error_indication.error_cause);
3909
3910         dev_warn_ratelimited(dev,
3911                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3912                              crq->error_indication.flags
3913                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3914                              ibmvnic_fw_err_cause(cause));
3915
3916         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3917                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3918         else
3919                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3920 }
3921
3922 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3923                                  struct ibmvnic_adapter *adapter)
3924 {
3925         struct net_device *netdev = adapter->netdev;
3926         struct device *dev = &adapter->vdev->dev;
3927         long rc;
3928
3929         rc = crq->change_mac_addr_rsp.rc.code;
3930         if (rc) {
3931                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3932                 goto out;
3933         }
3934         ether_addr_copy(netdev->dev_addr,
3935                         &crq->change_mac_addr_rsp.mac_addr[0]);
3936 out:
3937         complete(&adapter->fw_done);
3938         return rc;
3939 }
3940
3941 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3942                                    struct ibmvnic_adapter *adapter)
3943 {
3944         struct device *dev = &adapter->vdev->dev;
3945         u64 *req_value;
3946         char *name;
3947
3948         atomic_dec(&adapter->running_cap_crqs);
3949         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3950         case REQ_TX_QUEUES:
3951                 req_value = &adapter->req_tx_queues;
3952                 name = "tx";
3953                 break;
3954         case REQ_RX_QUEUES:
3955                 req_value = &adapter->req_rx_queues;
3956                 name = "rx";
3957                 break;
3958         case REQ_RX_ADD_QUEUES:
3959                 req_value = &adapter->req_rx_add_queues;
3960                 name = "rx_add";
3961                 break;
3962         case REQ_TX_ENTRIES_PER_SUBCRQ:
3963                 req_value = &adapter->req_tx_entries_per_subcrq;
3964                 name = "tx_entries_per_subcrq";
3965                 break;
3966         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3967                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3968                 name = "rx_add_entries_per_subcrq";
3969                 break;
3970         case REQ_MTU:
3971                 req_value = &adapter->req_mtu;
3972                 name = "mtu";
3973                 break;
3974         case PROMISC_REQUESTED:
3975                 req_value = &adapter->promisc;
3976                 name = "promisc";
3977                 break;
3978         default:
3979                 dev_err(dev, "Got invalid cap request rsp %d\n",
3980                         crq->request_capability.capability);
3981                 return;
3982         }
3983
3984         switch (crq->request_capability_rsp.rc.code) {
3985         case SUCCESS:
3986                 break;
3987         case PARTIALSUCCESS:
3988                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3989                          *req_value,
3990                          (long int)be64_to_cpu(crq->request_capability_rsp.
3991                                                number), name);
3992
3993                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3994                     REQ_MTU) {
3995                         pr_err("mtu of %llu is not supported. Reverting.\n",
3996                                *req_value);
3997                         *req_value = adapter->fallback.mtu;
3998                 } else {
3999                         *req_value =
4000                                 be64_to_cpu(crq->request_capability_rsp.number);
4001                 }
4002
4003                 ibmvnic_send_req_caps(adapter, 1);
4004                 return;
4005         default:
4006                 dev_err(dev, "Error %d in request cap rsp\n",
4007                         crq->request_capability_rsp.rc.code);
4008                 return;
4009         }
4010
4011         /* Done receiving requested capabilities, query IP offload support */
4012         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4013                 union ibmvnic_crq newcrq;
4014                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4015                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4016                     &adapter->ip_offload_buf;
4017
4018                 adapter->wait_capability = false;
4019                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4020                                                          buf_sz,
4021                                                          DMA_FROM_DEVICE);
4022
4023                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4024                         if (!firmware_has_feature(FW_FEATURE_CMO))
4025                                 dev_err(dev, "Couldn't map offload buffer\n");
4026                         return;
4027                 }
4028
4029                 memset(&newcrq, 0, sizeof(newcrq));
4030                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4031                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4032                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4033                 newcrq.query_ip_offload.ioba =
4034                     cpu_to_be32(adapter->ip_offload_tok);
4035
4036                 ibmvnic_send_crq(adapter, &newcrq);
4037         }
4038 }
4039
4040 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4041                             struct ibmvnic_adapter *adapter)
4042 {
4043         struct device *dev = &adapter->vdev->dev;
4044         struct net_device *netdev = adapter->netdev;
4045         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4046         struct ibmvnic_login_buffer *login = adapter->login_buf;
4047         int i;
4048
4049         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4050                          DMA_TO_DEVICE);
4051         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4052                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4053
4054         /* If the number of queues requested can't be allocated by the
4055          * server, the login response will return with code 1. We will need
4056          * to resend the login buffer with fewer queues requested.
4057          */
4058         if (login_rsp_crq->generic.rc.code) {
4059                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4060                 complete(&adapter->init_done);
4061                 return 0;
4062         }
4063
4064         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4065
4066         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4067         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4068                 netdev_dbg(adapter->netdev, "%016lx\n",
4069                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4070         }
4071
4072         /* Sanity checks */
4073         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4074             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4075              adapter->req_rx_add_queues !=
4076              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4077                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4078                 ibmvnic_remove(adapter->vdev);
4079                 return -EIO;
4080         }
4081         release_login_buffer(adapter);
4082         complete(&adapter->init_done);
4083
4084         return 0;
4085 }
4086
4087 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4088                                      struct ibmvnic_adapter *adapter)
4089 {
4090         struct device *dev = &adapter->vdev->dev;
4091         long rc;
4092
4093         rc = crq->request_unmap_rsp.rc.code;
4094         if (rc)
4095                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4096 }
4097
4098 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4099                                  struct ibmvnic_adapter *adapter)
4100 {
4101         struct net_device *netdev = adapter->netdev;
4102         struct device *dev = &adapter->vdev->dev;
4103         long rc;
4104
4105         rc = crq->query_map_rsp.rc.code;
4106         if (rc) {
4107                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4108                 return;
4109         }
4110         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4111                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4112                    crq->query_map_rsp.free_pages);
4113 }
4114
4115 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4116                                  struct ibmvnic_adapter *adapter)
4117 {
4118         struct net_device *netdev = adapter->netdev;
4119         struct device *dev = &adapter->vdev->dev;
4120         long rc;
4121
4122         atomic_dec(&adapter->running_cap_crqs);
4123         netdev_dbg(netdev, "Outstanding queries: %d\n",
4124                    atomic_read(&adapter->running_cap_crqs));
4125         rc = crq->query_capability.rc.code;
4126         if (rc) {
4127                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4128                 goto out;
4129         }
4130
4131         switch (be16_to_cpu(crq->query_capability.capability)) {
4132         case MIN_TX_QUEUES:
4133                 adapter->min_tx_queues =
4134                     be64_to_cpu(crq->query_capability.number);
4135                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4136                            adapter->min_tx_queues);
4137                 break;
4138         case MIN_RX_QUEUES:
4139                 adapter->min_rx_queues =
4140                     be64_to_cpu(crq->query_capability.number);
4141                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4142                            adapter->min_rx_queues);
4143                 break;
4144         case MIN_RX_ADD_QUEUES:
4145                 adapter->min_rx_add_queues =
4146                     be64_to_cpu(crq->query_capability.number);
4147                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4148                            adapter->min_rx_add_queues);
4149                 break;
4150         case MAX_TX_QUEUES:
4151                 adapter->max_tx_queues =
4152                     be64_to_cpu(crq->query_capability.number);
4153                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4154                            adapter->max_tx_queues);
4155                 break;
4156         case MAX_RX_QUEUES:
4157                 adapter->max_rx_queues =
4158                     be64_to_cpu(crq->query_capability.number);
4159                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4160                            adapter->max_rx_queues);
4161                 break;
4162         case MAX_RX_ADD_QUEUES:
4163                 adapter->max_rx_add_queues =
4164                     be64_to_cpu(crq->query_capability.number);
4165                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4166                            adapter->max_rx_add_queues);
4167                 break;
4168         case MIN_TX_ENTRIES_PER_SUBCRQ:
4169                 adapter->min_tx_entries_per_subcrq =
4170                     be64_to_cpu(crq->query_capability.number);
4171                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4172                            adapter->min_tx_entries_per_subcrq);
4173                 break;
4174         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4175                 adapter->min_rx_add_entries_per_subcrq =
4176                     be64_to_cpu(crq->query_capability.number);
4177                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4178                            adapter->min_rx_add_entries_per_subcrq);
4179                 break;
4180         case MAX_TX_ENTRIES_PER_SUBCRQ:
4181                 adapter->max_tx_entries_per_subcrq =
4182                     be64_to_cpu(crq->query_capability.number);
4183                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4184                            adapter->max_tx_entries_per_subcrq);
4185                 break;
4186         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4187                 adapter->max_rx_add_entries_per_subcrq =
4188                     be64_to_cpu(crq->query_capability.number);
4189                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4190                            adapter->max_rx_add_entries_per_subcrq);
4191                 break;
4192         case TCP_IP_OFFLOAD:
4193                 adapter->tcp_ip_offload =
4194                     be64_to_cpu(crq->query_capability.number);
4195                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4196                            adapter->tcp_ip_offload);
4197                 break;
4198         case PROMISC_SUPPORTED:
4199                 adapter->promisc_supported =
4200                     be64_to_cpu(crq->query_capability.number);
4201                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4202                            adapter->promisc_supported);
4203                 break;
4204         case MIN_MTU:
4205                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4206                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4207                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4208                 break;
4209         case MAX_MTU:
4210                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4211                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4212                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4213                 break;
4214         case MAX_MULTICAST_FILTERS:
4215                 adapter->max_multicast_filters =
4216                     be64_to_cpu(crq->query_capability.number);
4217                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4218                            adapter->max_multicast_filters);
4219                 break;
4220         case VLAN_HEADER_INSERTION:
4221                 adapter->vlan_header_insertion =
4222                     be64_to_cpu(crq->query_capability.number);
4223                 if (adapter->vlan_header_insertion)
4224                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4225                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4226                            adapter->vlan_header_insertion);
4227                 break;
4228         case RX_VLAN_HEADER_INSERTION:
4229                 adapter->rx_vlan_header_insertion =
4230                     be64_to_cpu(crq->query_capability.number);
4231                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4232                            adapter->rx_vlan_header_insertion);
4233                 break;
4234         case MAX_TX_SG_ENTRIES:
4235                 adapter->max_tx_sg_entries =
4236                     be64_to_cpu(crq->query_capability.number);
4237                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4238                            adapter->max_tx_sg_entries);
4239                 break;
4240         case RX_SG_SUPPORTED:
4241                 adapter->rx_sg_supported =
4242                     be64_to_cpu(crq->query_capability.number);
4243                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4244                            adapter->rx_sg_supported);
4245                 break;
4246         case OPT_TX_COMP_SUB_QUEUES:
4247                 adapter->opt_tx_comp_sub_queues =
4248                     be64_to_cpu(crq->query_capability.number);
4249                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4250                            adapter->opt_tx_comp_sub_queues);
4251                 break;
4252         case OPT_RX_COMP_QUEUES:
4253                 adapter->opt_rx_comp_queues =
4254                     be64_to_cpu(crq->query_capability.number);
4255                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4256                            adapter->opt_rx_comp_queues);
4257                 break;
4258         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4259                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4260                     be64_to_cpu(crq->query_capability.number);
4261                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4262                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4263                 break;
4264         case OPT_TX_ENTRIES_PER_SUBCRQ:
4265                 adapter->opt_tx_entries_per_subcrq =
4266                     be64_to_cpu(crq->query_capability.number);
4267                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4268                            adapter->opt_tx_entries_per_subcrq);
4269                 break;
4270         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4271                 adapter->opt_rxba_entries_per_subcrq =
4272                     be64_to_cpu(crq->query_capability.number);
4273                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4274                            adapter->opt_rxba_entries_per_subcrq);
4275                 break;
4276         case TX_RX_DESC_REQ:
4277                 adapter->tx_rx_desc_req = crq->query_capability.number;
4278                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4279                            adapter->tx_rx_desc_req);
4280                 break;
4281
4282         default:
4283                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4284                            crq->query_capability.capability);
4285         }
4286
4287 out:
4288         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4289                 adapter->wait_capability = false;
4290                 ibmvnic_send_req_caps(adapter, 0);
4291         }
4292 }
4293
4294 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4295 {
4296         union ibmvnic_crq crq;
4297         int rc;
4298
4299         memset(&crq, 0, sizeof(crq));
4300         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4301         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4302         init_completion(&adapter->fw_done);
4303         rc = ibmvnic_send_crq(adapter, &crq);
4304         if (rc)
4305                 return rc;
4306         wait_for_completion(&adapter->fw_done);
4307         return adapter->fw_done_rc ? -EIO : 0;
4308 }
4309
4310 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4311                                        struct ibmvnic_adapter *adapter)
4312 {
4313         struct net_device *netdev = adapter->netdev;
4314         int rc;
4315
4316         rc = crq->query_phys_parms_rsp.rc.code;
4317         if (rc) {
4318                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4319                 return rc;
4320         }
4321         switch (cpu_to_be32(crq->query_phys_parms_rsp.speed)) {
4322         case IBMVNIC_10MBPS:
4323                 adapter->speed = SPEED_10;
4324                 break;
4325         case IBMVNIC_100MBPS:
4326                 adapter->speed = SPEED_100;
4327                 break;
4328         case IBMVNIC_1GBPS:
4329                 adapter->speed = SPEED_1000;
4330                 break;
4331         case IBMVNIC_10GBP:
4332                 adapter->speed = SPEED_10000;
4333                 break;
4334         case IBMVNIC_25GBPS:
4335                 adapter->speed = SPEED_25000;
4336                 break;
4337         case IBMVNIC_40GBPS:
4338                 adapter->speed = SPEED_40000;
4339                 break;
4340         case IBMVNIC_50GBPS:
4341                 adapter->speed = SPEED_50000;
4342                 break;
4343         case IBMVNIC_100GBPS:
4344                 adapter->speed = SPEED_100000;
4345                 break;
4346         default:
4347                 netdev_warn(netdev, "Unknown speed 0x%08x\n",
4348                             cpu_to_be32(crq->query_phys_parms_rsp.speed));
4349                 adapter->speed = SPEED_UNKNOWN;
4350         }
4351         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4352                 adapter->duplex = DUPLEX_FULL;
4353         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4354                 adapter->duplex = DUPLEX_HALF;
4355         else
4356                 adapter->duplex = DUPLEX_UNKNOWN;
4357
4358         return rc;
4359 }
4360
4361 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4362                                struct ibmvnic_adapter *adapter)
4363 {
4364         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4365         struct net_device *netdev = adapter->netdev;
4366         struct device *dev = &adapter->vdev->dev;
4367         u64 *u64_crq = (u64 *)crq;
4368         long rc;
4369
4370         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4371                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4372                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4373         switch (gen_crq->first) {
4374         case IBMVNIC_CRQ_INIT_RSP:
4375                 switch (gen_crq->cmd) {
4376                 case IBMVNIC_CRQ_INIT:
4377                         dev_info(dev, "Partner initialized\n");
4378                         adapter->from_passive_init = true;
4379                         adapter->failover_pending = false;
4380                         if (!completion_done(&adapter->init_done)) {
4381                                 complete(&adapter->init_done);
4382                                 adapter->init_done_rc = -EIO;
4383                         }
4384                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4385                         break;
4386                 case IBMVNIC_CRQ_INIT_COMPLETE:
4387                         dev_info(dev, "Partner initialization complete\n");
4388                         adapter->crq.active = true;
4389                         send_version_xchg(adapter);
4390                         break;
4391                 default:
4392                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4393                 }
4394                 return;
4395         case IBMVNIC_CRQ_XPORT_EVENT:
4396                 netif_carrier_off(netdev);
4397                 adapter->crq.active = false;
4398                 if (adapter->resetting)
4399                         adapter->force_reset_recovery = true;
4400                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4401                         dev_info(dev, "Migrated, re-enabling adapter\n");
4402                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4403                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4404                         dev_info(dev, "Backing device failover detected\n");
4405                         adapter->failover_pending = true;
4406                 } else {
4407                         /* The adapter lost the connection */
4408                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4409                                 gen_crq->cmd);
4410                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4411                 }
4412                 return;
4413         case IBMVNIC_CRQ_CMD_RSP:
4414                 break;
4415         default:
4416                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4417                         gen_crq->first);
4418                 return;
4419         }
4420
4421         switch (gen_crq->cmd) {
4422         case VERSION_EXCHANGE_RSP:
4423                 rc = crq->version_exchange_rsp.rc.code;
4424                 if (rc) {
4425                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4426                         break;
4427                 }
4428                 dev_info(dev, "Partner protocol version is %d\n",
4429                          crq->version_exchange_rsp.version);
4430                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4431                     ibmvnic_version)
4432                         ibmvnic_version =
4433                             be16_to_cpu(crq->version_exchange_rsp.version);
4434                 send_cap_queries(adapter);
4435                 break;
4436         case QUERY_CAPABILITY_RSP:
4437                 handle_query_cap_rsp(crq, adapter);
4438                 break;
4439         case QUERY_MAP_RSP:
4440                 handle_query_map_rsp(crq, adapter);
4441                 break;
4442         case REQUEST_MAP_RSP:
4443                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4444                 complete(&adapter->fw_done);
4445                 break;
4446         case REQUEST_UNMAP_RSP:
4447                 handle_request_unmap_rsp(crq, adapter);
4448                 break;
4449         case REQUEST_CAPABILITY_RSP:
4450                 handle_request_cap_rsp(crq, adapter);
4451                 break;
4452         case LOGIN_RSP:
4453                 netdev_dbg(netdev, "Got Login Response\n");
4454                 handle_login_rsp(crq, adapter);
4455                 break;
4456         case LOGICAL_LINK_STATE_RSP:
4457                 netdev_dbg(netdev,
4458                            "Got Logical Link State Response, state: %d rc: %d\n",
4459                            crq->logical_link_state_rsp.link_state,
4460                            crq->logical_link_state_rsp.rc.code);
4461                 adapter->logical_link_state =
4462                     crq->logical_link_state_rsp.link_state;
4463                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4464                 complete(&adapter->init_done);
4465                 break;
4466         case LINK_STATE_INDICATION:
4467                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4468                 adapter->phys_link_state =
4469                     crq->link_state_indication.phys_link_state;
4470                 adapter->logical_link_state =
4471                     crq->link_state_indication.logical_link_state;
4472                 if (adapter->phys_link_state && adapter->logical_link_state)
4473                         netif_carrier_on(netdev);
4474                 else
4475                         netif_carrier_off(netdev);
4476                 break;
4477         case CHANGE_MAC_ADDR_RSP:
4478                 netdev_dbg(netdev, "Got MAC address change Response\n");
4479                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4480                 break;
4481         case ERROR_INDICATION:
4482                 netdev_dbg(netdev, "Got Error Indication\n");
4483                 handle_error_indication(crq, adapter);
4484                 break;
4485         case REQUEST_STATISTICS_RSP:
4486                 netdev_dbg(netdev, "Got Statistics Response\n");
4487                 complete(&adapter->stats_done);
4488                 break;
4489         case QUERY_IP_OFFLOAD_RSP:
4490                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4491                 handle_query_ip_offload_rsp(adapter);
4492                 break;
4493         case MULTICAST_CTRL_RSP:
4494                 netdev_dbg(netdev, "Got multicast control Response\n");
4495                 break;
4496         case CONTROL_IP_OFFLOAD_RSP:
4497                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4498                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4499                                  sizeof(adapter->ip_offload_ctrl),
4500                                  DMA_TO_DEVICE);
4501                 complete(&adapter->init_done);
4502                 break;
4503         case COLLECT_FW_TRACE_RSP:
4504                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4505                 complete(&adapter->fw_done);
4506                 break;
4507         case GET_VPD_SIZE_RSP:
4508                 handle_vpd_size_rsp(crq, adapter);
4509                 break;
4510         case GET_VPD_RSP:
4511                 handle_vpd_rsp(crq, adapter);
4512                 break;
4513         case QUERY_PHYS_PARMS_RSP:
4514                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4515                 complete(&adapter->fw_done);
4516                 break;
4517         default:
4518                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4519                            gen_crq->cmd);
4520         }
4521 }
4522
4523 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4524 {
4525         struct ibmvnic_adapter *adapter = instance;
4526
4527         tasklet_schedule(&adapter->tasklet);
4528         return IRQ_HANDLED;
4529 }
4530
4531 static void ibmvnic_tasklet(void *data)
4532 {
4533         struct ibmvnic_adapter *adapter = data;
4534         struct ibmvnic_crq_queue *queue = &adapter->crq;
4535         union ibmvnic_crq *crq;
4536         unsigned long flags;
4537         bool done = false;
4538
4539         spin_lock_irqsave(&queue->lock, flags);
4540         while (!done) {
4541                 /* Pull all the valid messages off the CRQ */
4542                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4543                         ibmvnic_handle_crq(crq, adapter);
4544                         crq->generic.first = 0;
4545                 }
4546
4547                 /* remain in tasklet until all
4548                  * capabilities responses are received
4549                  */
4550                 if (!adapter->wait_capability)
4551                         done = true;
4552         }
4553         /* if capabilities CRQ's were sent in this tasklet, the following
4554          * tasklet must wait until all responses are received
4555          */
4556         if (atomic_read(&adapter->running_cap_crqs) != 0)
4557                 adapter->wait_capability = true;
4558         spin_unlock_irqrestore(&queue->lock, flags);
4559 }
4560
4561 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4562 {
4563         struct vio_dev *vdev = adapter->vdev;
4564         int rc;
4565
4566         do {
4567                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4568         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4569
4570         if (rc)
4571                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4572
4573         return rc;
4574 }
4575
4576 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4577 {
4578         struct ibmvnic_crq_queue *crq = &adapter->crq;
4579         struct device *dev = &adapter->vdev->dev;
4580         struct vio_dev *vdev = adapter->vdev;
4581         int rc;
4582
4583         /* Close the CRQ */
4584         do {
4585                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4586         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4587
4588         /* Clean out the queue */
4589         memset(crq->msgs, 0, PAGE_SIZE);
4590         crq->cur = 0;
4591         crq->active = false;
4592
4593         /* And re-open it again */
4594         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4595                                 crq->msg_token, PAGE_SIZE);
4596
4597         if (rc == H_CLOSED)
4598                 /* Adapter is good, but other end is not ready */
4599                 dev_warn(dev, "Partner adapter not ready\n");
4600         else if (rc != 0)
4601                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4602
4603         return rc;
4604 }
4605
4606 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4607 {
4608         struct ibmvnic_crq_queue *crq = &adapter->crq;
4609         struct vio_dev *vdev = adapter->vdev;
4610         long rc;
4611
4612         if (!crq->msgs)
4613                 return;
4614
4615         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4616         free_irq(vdev->irq, adapter);
4617         tasklet_kill(&adapter->tasklet);
4618         do {
4619                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4620         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4621
4622         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4623                          DMA_BIDIRECTIONAL);
4624         free_page((unsigned long)crq->msgs);
4625         crq->msgs = NULL;
4626         crq->active = false;
4627 }
4628
4629 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4630 {
4631         struct ibmvnic_crq_queue *crq = &adapter->crq;
4632         struct device *dev = &adapter->vdev->dev;
4633         struct vio_dev *vdev = adapter->vdev;
4634         int rc, retrc = -ENOMEM;
4635
4636         if (crq->msgs)
4637                 return 0;
4638
4639         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4640         /* Should we allocate more than one page? */
4641
4642         if (!crq->msgs)
4643                 return -ENOMEM;
4644
4645         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4646         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4647                                         DMA_BIDIRECTIONAL);
4648         if (dma_mapping_error(dev, crq->msg_token))
4649                 goto map_failed;
4650
4651         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4652                                 crq->msg_token, PAGE_SIZE);
4653
4654         if (rc == H_RESOURCE)
4655                 /* maybe kexecing and resource is busy. try a reset */
4656                 rc = ibmvnic_reset_crq(adapter);
4657         retrc = rc;
4658
4659         if (rc == H_CLOSED) {
4660                 dev_warn(dev, "Partner adapter not ready\n");
4661         } else if (rc) {
4662                 dev_warn(dev, "Error %d opening adapter\n", rc);
4663                 goto reg_crq_failed;
4664         }
4665
4666         retrc = 0;
4667
4668         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4669                      (unsigned long)adapter);
4670
4671         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4672         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4673                  adapter->vdev->unit_address);
4674         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4675         if (rc) {
4676                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4677                         vdev->irq, rc);
4678                 goto req_irq_failed;
4679         }
4680
4681         rc = vio_enable_interrupts(vdev);
4682         if (rc) {
4683                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4684                 goto req_irq_failed;
4685         }
4686
4687         crq->cur = 0;
4688         spin_lock_init(&crq->lock);
4689
4690         return retrc;
4691
4692 req_irq_failed:
4693         tasklet_kill(&adapter->tasklet);
4694         do {
4695                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4696         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4697 reg_crq_failed:
4698         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4699 map_failed:
4700         free_page((unsigned long)crq->msgs);
4701         crq->msgs = NULL;
4702         return retrc;
4703 }
4704
4705 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4706 {
4707         struct device *dev = &adapter->vdev->dev;
4708         unsigned long timeout = msecs_to_jiffies(30000);
4709         u64 old_num_rx_queues, old_num_tx_queues;
4710         int rc;
4711
4712         adapter->from_passive_init = false;
4713
4714         old_num_rx_queues = adapter->req_rx_queues;
4715         old_num_tx_queues = adapter->req_tx_queues;
4716
4717         reinit_completion(&adapter->init_done);
4718         adapter->init_done_rc = 0;
4719         ibmvnic_send_crq_init(adapter);
4720         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4721                 dev_err(dev, "Initialization sequence timed out\n");
4722                 return -1;
4723         }
4724
4725         if (adapter->init_done_rc) {
4726                 release_crq_queue(adapter);
4727                 return adapter->init_done_rc;
4728         }
4729
4730         if (adapter->from_passive_init) {
4731                 adapter->state = VNIC_OPEN;
4732                 adapter->from_passive_init = false;
4733                 return -1;
4734         }
4735
4736         if (adapter->resetting && !adapter->wait_for_reset &&
4737             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4738                 if (adapter->req_rx_queues != old_num_rx_queues ||
4739                     adapter->req_tx_queues != old_num_tx_queues) {
4740                         release_sub_crqs(adapter, 0);
4741                         rc = init_sub_crqs(adapter);
4742                 } else {
4743                         rc = reset_sub_crq_queues(adapter);
4744                 }
4745         } else {
4746                 rc = init_sub_crqs(adapter);
4747         }
4748
4749         if (rc) {
4750                 dev_err(dev, "Initialization of sub crqs failed\n");
4751                 release_crq_queue(adapter);
4752                 return rc;
4753         }
4754
4755         rc = init_sub_crq_irqs(adapter);
4756         if (rc) {
4757                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4758                 release_crq_queue(adapter);
4759         }
4760
4761         return rc;
4762 }
4763
4764 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4765 {
4766         struct device *dev = &adapter->vdev->dev;
4767         unsigned long timeout = msecs_to_jiffies(30000);
4768         int rc;
4769
4770         adapter->from_passive_init = false;
4771
4772         adapter->init_done_rc = 0;
4773         ibmvnic_send_crq_init(adapter);
4774         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4775                 dev_err(dev, "Initialization sequence timed out\n");
4776                 return -1;
4777         }
4778
4779         if (adapter->init_done_rc) {
4780                 release_crq_queue(adapter);
4781                 return adapter->init_done_rc;
4782         }
4783
4784         if (adapter->from_passive_init) {
4785                 adapter->state = VNIC_OPEN;
4786                 adapter->from_passive_init = false;
4787                 return -1;
4788         }
4789
4790         rc = init_sub_crqs(adapter);
4791         if (rc) {
4792                 dev_err(dev, "Initialization of sub crqs failed\n");
4793                 release_crq_queue(adapter);
4794                 return rc;
4795         }
4796
4797         rc = init_sub_crq_irqs(adapter);
4798         if (rc) {
4799                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4800                 release_crq_queue(adapter);
4801         }
4802
4803         return rc;
4804 }
4805
4806 static struct device_attribute dev_attr_failover;
4807
4808 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4809 {
4810         struct ibmvnic_adapter *adapter;
4811         struct net_device *netdev;
4812         unsigned char *mac_addr_p;
4813         int rc;
4814
4815         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4816                 dev->unit_address);
4817
4818         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4819                                                         VETH_MAC_ADDR, NULL);
4820         if (!mac_addr_p) {
4821                 dev_err(&dev->dev,
4822                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4823                         __FILE__, __LINE__);
4824                 return 0;
4825         }
4826
4827         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4828                                    IBMVNIC_MAX_QUEUES);
4829         if (!netdev)
4830                 return -ENOMEM;
4831
4832         adapter = netdev_priv(netdev);
4833         adapter->state = VNIC_PROBING;
4834         dev_set_drvdata(&dev->dev, netdev);
4835         adapter->vdev = dev;
4836         adapter->netdev = netdev;
4837
4838         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4839         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4840         netdev->irq = dev->irq;
4841         netdev->netdev_ops = &ibmvnic_netdev_ops;
4842         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4843         SET_NETDEV_DEV(netdev, &dev->dev);
4844
4845         spin_lock_init(&adapter->stats_lock);
4846
4847         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4848         INIT_LIST_HEAD(&adapter->rwi_list);
4849         spin_lock_init(&adapter->rwi_lock);
4850         init_completion(&adapter->init_done);
4851         adapter->resetting = false;
4852
4853         do {
4854                 rc = init_crq_queue(adapter);
4855                 if (rc) {
4856                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4857                                 rc);
4858                         goto ibmvnic_init_fail;
4859                 }
4860
4861                 rc = ibmvnic_init(adapter);
4862                 if (rc && rc != EAGAIN)
4863                         goto ibmvnic_init_fail;
4864         } while (rc == EAGAIN);
4865
4866         rc = init_stats_buffers(adapter);
4867         if (rc)
4868                 goto ibmvnic_init_fail;
4869
4870         rc = init_stats_token(adapter);
4871         if (rc)
4872                 goto ibmvnic_stats_fail;
4873
4874         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4875         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4876         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4877
4878         rc = device_create_file(&dev->dev, &dev_attr_failover);
4879         if (rc)
4880                 goto ibmvnic_dev_file_err;
4881
4882         netif_carrier_off(netdev);
4883         rc = register_netdev(netdev);
4884         if (rc) {
4885                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4886                 goto ibmvnic_register_fail;
4887         }
4888         dev_info(&dev->dev, "ibmvnic registered\n");
4889
4890         adapter->state = VNIC_PROBED;
4891
4892         adapter->wait_for_reset = false;
4893
4894         return 0;
4895
4896 ibmvnic_register_fail:
4897         device_remove_file(&dev->dev, &dev_attr_failover);
4898
4899 ibmvnic_dev_file_err:
4900         release_stats_token(adapter);
4901
4902 ibmvnic_stats_fail:
4903         release_stats_buffers(adapter);
4904
4905 ibmvnic_init_fail:
4906         release_sub_crqs(adapter, 1);
4907         release_crq_queue(adapter);
4908         free_netdev(netdev);
4909
4910         return rc;
4911 }
4912
4913 static int ibmvnic_remove(struct vio_dev *dev)
4914 {
4915         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4916         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4917
4918         adapter->state = VNIC_REMOVING;
4919         rtnl_lock();
4920         unregister_netdevice(netdev);
4921
4922         release_resources(adapter);
4923         release_sub_crqs(adapter, 1);
4924         release_crq_queue(adapter);
4925
4926         release_stats_token(adapter);
4927         release_stats_buffers(adapter);
4928
4929         adapter->state = VNIC_REMOVED;
4930
4931         rtnl_unlock();
4932         device_remove_file(&dev->dev, &dev_attr_failover);
4933         free_netdev(netdev);
4934         dev_set_drvdata(&dev->dev, NULL);
4935
4936         return 0;
4937 }
4938
4939 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4940                               const char *buf, size_t count)
4941 {
4942         struct net_device *netdev = dev_get_drvdata(dev);
4943         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4944         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4945         __be64 session_token;
4946         long rc;
4947
4948         if (!sysfs_streq(buf, "1"))
4949                 return -EINVAL;
4950
4951         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4952                          H_GET_SESSION_TOKEN, 0, 0, 0);
4953         if (rc) {
4954                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4955                            rc);
4956                 return -EINVAL;
4957         }
4958
4959         session_token = (__be64)retbuf[0];
4960         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4961                    be64_to_cpu(session_token));
4962         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4963                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4964         if (rc) {
4965                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4966                            rc);
4967                 return -EINVAL;
4968         }
4969
4970         return count;
4971 }
4972
4973 static DEVICE_ATTR_WO(failover);
4974
4975 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4976 {
4977         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4978         struct ibmvnic_adapter *adapter;
4979         struct iommu_table *tbl;
4980         unsigned long ret = 0;
4981         int i;
4982
4983         tbl = get_iommu_table_base(&vdev->dev);
4984
4985         /* netdev inits at probe time along with the structures we need below*/
4986         if (!netdev)
4987                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4988
4989         adapter = netdev_priv(netdev);
4990
4991         ret += PAGE_SIZE; /* the crq message queue */
4992         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4993
4994         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4995                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4996
4997         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4998              i++)
4999                 ret += adapter->rx_pool[i].size *
5000                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5001
5002         return ret;
5003 }
5004
5005 static int ibmvnic_resume(struct device *dev)
5006 {
5007         struct net_device *netdev = dev_get_drvdata(dev);
5008         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5009
5010         if (adapter->state != VNIC_OPEN)
5011                 return 0;
5012
5013         tasklet_schedule(&adapter->tasklet);
5014
5015         return 0;
5016 }
5017
5018 static const struct vio_device_id ibmvnic_device_table[] = {
5019         {"network", "IBM,vnic"},
5020         {"", "" }
5021 };
5022 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5023
5024 static const struct dev_pm_ops ibmvnic_pm_ops = {
5025         .resume = ibmvnic_resume
5026 };
5027
5028 static struct vio_driver ibmvnic_driver = {
5029         .id_table       = ibmvnic_device_table,
5030         .probe          = ibmvnic_probe,
5031         .remove         = ibmvnic_remove,
5032         .get_desired_dma = ibmvnic_get_desired_dma,
5033         .name           = ibmvnic_driver_name,
5034         .pm             = &ibmvnic_pm_ops,
5035 };
5036
5037 /* module functions */
5038 static int __init ibmvnic_module_init(void)
5039 {
5040         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5041                 IBMVNIC_DRIVER_VERSION);
5042
5043         return vio_register_driver(&ibmvnic_driver);
5044 }
5045
5046 static void __exit ibmvnic_module_exit(void)
5047 {
5048         vio_unregister_driver(&ibmvnic_driver);
5049 }
5050
5051 module_init(ibmvnic_module_init);
5052 module_exit(ibmvnic_module_exit);