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_query_map(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_query_cap(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_reset_init(struct ibmvnic_adapter *, bool reset);
 108 static void release_crq_queue(struct ibmvnic_adapter *);
 109 static int __ibmvnic_set_mac(struct net_device *, u8 *);
 110 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 111 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
 112
 113 struct ibmvnic_stat {
 114         char name[ETH_GSTRING_LEN];
 115         int offset;
 116 };
 117
 118 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 119                              offsetof(struct ibmvnic_statistics, stat))
 120 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 121
 122 static const struct ibmvnic_stat ibmvnic_stats[] = {
 123         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 124         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 125         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 126         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 127         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 128         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 129         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 130         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 131         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 132         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 133         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 134         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 135         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 136         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 137         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 138         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 139         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 140         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 141         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 142         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 143         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 144         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 145 };
 146
 147 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 148                           unsigned long length, unsigned long *number,
 149                           unsigned long *irq)
 150 {
 151         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 152         long rc;
 153
 154         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 155         *number = retbuf[0];
 156         *irq = retbuf[1];
 157
 158         return rc;
 159 }
 160
 161 /**
 162  * ibmvnic_wait_for_completion - Check device state and wait for completion
 163  * @adapter: private device data
 164  * @comp_done: completion structure to wait for
 165  * @timeout: time to wait in milliseconds
 166  *
 167  * Wait for a completion signal or until the timeout limit is reached
 168  * while checking that the device is still active.
 169  */
 170 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
 171                                        struct completion *comp_done,
 172                                        unsigned long timeout)
 173 {
 174         struct net_device *netdev;
 175         unsigned long div_timeout;
 176         u8 retry;
 177
 178         netdev = adapter->netdev;
 179         retry = 5;
 180         div_timeout = msecs_to_jiffies(timeout / retry);
 181         while (true) {
 182                 if (!adapter->crq.active) {
 183                         netdev_err(netdev, "Device down!\n");
 184                         return -ENODEV;
 185                 }
 186                 if (!retry--)
 187                         break;
 188                 if (wait_for_completion_timeout(comp_done, div_timeout))
 189                         return 0;
 190         }
 191         netdev_err(netdev, "Operation timed out.\n");
 192         return -ETIMEDOUT;
 193 }
 194
 195 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 196                                 struct ibmvnic_long_term_buff *ltb, int size)
 197 {
 198         struct device *dev = &adapter->vdev->dev;
 199         int rc;
 200
 201         ltb->size = size;
 202         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 203                                        GFP_KERNEL);
 204
 205         if (!ltb->buff) {
 206                 dev_err(dev, "Couldn't alloc long term buffer\n");
 207                 return -ENOMEM;
 208         }
 209         ltb->map_id = adapter->map_id;
 210         adapter->map_id++;
 211
 212         mutex_lock(&adapter->fw_lock);
 213         adapter->fw_done_rc = 0;
 214         reinit_completion(&adapter->fw_done);
 215         rc = send_request_map(adapter, ltb->addr,
 216                               ltb->size, ltb->map_id);
 217         if (rc) {
 218                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 219                 mutex_unlock(&adapter->fw_lock);
 220                 return rc;
 221         }
 222
 223         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
 224         if (rc) {
 225                 dev_err(dev,
 226                         "Long term map request aborted or timed out,rc = %d\n",
 227                         rc);
 228                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 229                 mutex_unlock(&adapter->fw_lock);
 230                 return rc;
 231         }
 232
 233         if (adapter->fw_done_rc) {
 234                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 235                         adapter->fw_done_rc);
 236                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 237                 mutex_unlock(&adapter->fw_lock);
 238                 return -1;
 239         }
 240         mutex_unlock(&adapter->fw_lock);
 241         return 0;
 242 }
 243
 244 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 245                                 struct ibmvnic_long_term_buff *ltb)
 246 {
 247         struct device *dev = &adapter->vdev->dev;
 248
 249         if (!ltb->buff)
 250                 return;
 251
 252         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 253             adapter->reset_reason != VNIC_RESET_MOBILITY)
 254                 send_request_unmap(adapter, ltb->map_id);
 255         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 256 }
 257
 258 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 259                                 struct ibmvnic_long_term_buff *ltb)
 260 {
 261         struct device *dev = &adapter->vdev->dev;
 262         int rc;
 263
 264         memset(ltb->buff, 0, ltb->size);
 265
 266         mutex_lock(&adapter->fw_lock);
 267         adapter->fw_done_rc = 0;
 268
 269         reinit_completion(&adapter->fw_done);
 270         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 271         if (rc) {
 272                 mutex_unlock(&adapter->fw_lock);
 273                 return rc;
 274         }
 275
 276         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
 277         if (rc) {
 278                 dev_info(dev,
 279                          "Reset failed, long term map request timed out or aborted\n");
 280                 mutex_unlock(&adapter->fw_lock);
 281                 return rc;
 282         }
 283
 284         if (adapter->fw_done_rc) {
 285                 dev_info(dev,
 286                          "Reset failed, attempting to free and reallocate buffer\n");
 287                 free_long_term_buff(adapter, ltb);
 288                 mutex_unlock(&adapter->fw_lock);
 289                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 290         }
 291         mutex_unlock(&adapter->fw_lock);
 292         return 0;
 293 }
 294
 295 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 296 {
 297         int i;
 298
 299         for (i = 0; i < adapter->num_active_rx_pools; i++)
 300                 adapter->rx_pool[i].active = 0;
 301 }
 302
 303 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 304                               struct ibmvnic_rx_pool *pool)
 305 {
 306         int count = pool->size - atomic_read(&pool->available);
 307         u64 handle = adapter->rx_scrq[pool->index]->handle;
 308         struct device *dev = &adapter->vdev->dev;
 309         int buffers_added = 0;
 310         unsigned long lpar_rc;
 311         union sub_crq sub_crq;
 312         struct sk_buff *skb;
 313         unsigned int offset;
 314         dma_addr_t dma_addr;
 315         unsigned char *dst;
 316         int shift = 0;
 317         int index;
 318         int i;
 319
 320         if (!pool->active)
 321                 return;
 322
 323         for (i = 0; i < count; ++i) {
 324                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 325                 if (!skb) {
 326                         dev_err(dev, "Couldn't replenish rx buff\n");
 327                         adapter->replenish_no_mem++;
 328                         break;
 329                 }
 330
 331                 index = pool->free_map[pool->next_free];
 332
 333                 if (pool->rx_buff[index].skb)
 334                         dev_err(dev, "Inconsistent free_map!\n");
 335
 336                 /* Copy the skb to the long term mapped DMA buffer */
 337                 offset = index * pool->buff_size;
 338                 dst = pool->long_term_buff.buff + offset;
 339                 memset(dst, 0, pool->buff_size);
 340                 dma_addr = pool->long_term_buff.addr + offset;
 341                 pool->rx_buff[index].data = dst;
 342
 343                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 344                 pool->rx_buff[index].dma = dma_addr;
 345                 pool->rx_buff[index].skb = skb;
 346                 pool->rx_buff[index].pool_index = pool->index;
 347                 pool->rx_buff[index].size = pool->buff_size;
 348
 349                 memset(&sub_crq, 0, sizeof(sub_crq));
 350                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 351                 sub_crq.rx_add.correlator =
 352                     cpu_to_be64((u64)&pool->rx_buff[index]);
 353                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 354                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 355
 356                 /* The length field of the sCRQ is defined to be 24 bits so the
 357                  * buffer size needs to be left shifted by a byte before it is
 358                  * converted to big endian to prevent the last byte from being
 359                  * truncated.
 360                  */
 361 #ifdef __LITTLE_ENDIAN__
 362                 shift = 8;
 363 #endif
 364                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 365
 366                 lpar_rc = send_subcrq(adapter, handle, &sub_crq);
 367                 if (lpar_rc != H_SUCCESS)
 368                         goto failure;
 369
 370                 buffers_added++;
 371                 adapter->replenish_add_buff_success++;
 372                 pool->next_free = (pool->next_free + 1) % pool->size;
 373         }
 374         atomic_add(buffers_added, &pool->available);
 375         return;
 376
 377 failure:
 378         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 379                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 380         pool->free_map[pool->next_free] = index;
 381         pool->rx_buff[index].skb = NULL;
 382
 383         dev_kfree_skb_any(skb);
 384         adapter->replenish_add_buff_failure++;
 385         atomic_add(buffers_added, &pool->available);
 386
 387         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 388                 /* Disable buffer pool replenishment and report carrier off if
 389                  * queue is closed or pending failover.
 390                  * Firmware guarantees that a signal will be sent to the
 391                  * driver, triggering a reset.
 392                  */
 393                 deactivate_rx_pools(adapter);
 394                 netif_carrier_off(adapter->netdev);
 395         }
 396 }
 397
 398 static void replenish_pools(struct ibmvnic_adapter *adapter)
 399 {
 400         int i;
 401
 402         adapter->replenish_task_cycles++;
 403         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 404                 if (adapter->rx_pool[i].active)
 405                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 406         }
 407 }
 408
 409 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 410 {
 411         kfree(adapter->tx_stats_buffers);
 412         kfree(adapter->rx_stats_buffers);
 413         adapter->tx_stats_buffers = NULL;
 414         adapter->rx_stats_buffers = NULL;
 415 }
 416
 417 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 418 {
 419         adapter->tx_stats_buffers =
 420                                 kcalloc(IBMVNIC_MAX_QUEUES,
 421                                         sizeof(struct ibmvnic_tx_queue_stats),
 422                                         GFP_KERNEL);
 423         if (!adapter->tx_stats_buffers)
 424                 return -ENOMEM;
 425
 426         adapter->rx_stats_buffers =
 427                                 kcalloc(IBMVNIC_MAX_QUEUES,
 428                                         sizeof(struct ibmvnic_rx_queue_stats),
 429                                         GFP_KERNEL);
 430         if (!adapter->rx_stats_buffers)
 431                 return -ENOMEM;
 432
 433         return 0;
 434 }
 435
 436 static void release_stats_token(struct ibmvnic_adapter *adapter)
 437 {
 438         struct device *dev = &adapter->vdev->dev;
 439
 440         if (!adapter->stats_token)
 441                 return;
 442
 443         dma_unmap_single(dev, adapter->stats_token,
 444                          sizeof(struct ibmvnic_statistics),
 445                          DMA_FROM_DEVICE);
 446         adapter->stats_token = 0;
 447 }
 448
 449 static int init_stats_token(struct ibmvnic_adapter *adapter)
 450 {
 451         struct device *dev = &adapter->vdev->dev;
 452         dma_addr_t stok;
 453
 454         stok = dma_map_single(dev, &adapter->stats,
 455                               sizeof(struct ibmvnic_statistics),
 456                               DMA_FROM_DEVICE);
 457         if (dma_mapping_error(dev, stok)) {
 458                 dev_err(dev, "Couldn't map stats buffer\n");
 459                 return -1;
 460         }
 461
 462         adapter->stats_token = stok;
 463         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 464         return 0;
 465 }
 466
 467 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 468 {
 469         struct ibmvnic_rx_pool *rx_pool;
 470         u64 buff_size;
 471         int rx_scrqs;
 472         int i, j, rc;
 473
 474         if (!adapter->rx_pool)
 475                 return -1;
 476
 477         buff_size = adapter->cur_rx_buf_sz;
 478         rx_scrqs = adapter->num_active_rx_pools;
 479         for (i = 0; i < rx_scrqs; i++) {
 480                 rx_pool = &adapter->rx_pool[i];
 481
 482                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 483
 484                 if (rx_pool->buff_size != buff_size) {
 485                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 486                         rx_pool->buff_size = buff_size;
 487                         rc = alloc_long_term_buff(adapter,
 488                                                   &rx_pool->long_term_buff,
 489                                                   rx_pool->size *
 490                                                   rx_pool->buff_size);
 491                 } else {
 492                         rc = reset_long_term_buff(adapter,
 493                                                   &rx_pool->long_term_buff);
 494                 }
 495
 496                 if (rc)
 497                         return rc;
 498
 499                 for (j = 0; j < rx_pool->size; j++)
 500                         rx_pool->free_map[j] = j;
 501
 502                 memset(rx_pool->rx_buff, 0,
 503                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 504
 505                 atomic_set(&rx_pool->available, 0);
 506                 rx_pool->next_alloc = 0;
 507                 rx_pool->next_free = 0;
 508                 rx_pool->active = 1;
 509         }
 510
 511         return 0;
 512 }
 513
 514 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 515 {
 516         struct ibmvnic_rx_pool *rx_pool;
 517         int i, j;
 518
 519         if (!adapter->rx_pool)
 520                 return;
 521
 522         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 523                 rx_pool = &adapter->rx_pool[i];
 524
 525                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 526
 527                 kfree(rx_pool->free_map);
 528                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 529
 530                 if (!rx_pool->rx_buff)
 531                         continue;
 532
 533                 for (j = 0; j < rx_pool->size; j++) {
 534                         if (rx_pool->rx_buff[j].skb) {
 535                                 dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
 536                                 rx_pool->rx_buff[j].skb = NULL;
 537                         }
 538                 }
 539
 540                 kfree(rx_pool->rx_buff);
 541         }
 542
 543         kfree(adapter->rx_pool);
 544         adapter->rx_pool = NULL;
 545         adapter->num_active_rx_pools = 0;
 546 }
 547
 548 static int init_rx_pools(struct net_device *netdev)
 549 {
 550         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 551         struct device *dev = &adapter->vdev->dev;
 552         struct ibmvnic_rx_pool *rx_pool;
 553         int rxadd_subcrqs;
 554         u64 buff_size;
 555         int i, j;
 556
 557         rxadd_subcrqs = adapter->num_active_rx_scrqs;
 558         buff_size = adapter->cur_rx_buf_sz;
 559
 560         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 561                                    sizeof(struct ibmvnic_rx_pool),
 562                                    GFP_KERNEL);
 563         if (!adapter->rx_pool) {
 564                 dev_err(dev, "Failed to allocate rx pools\n");
 565                 return -1;
 566         }
 567
 568         adapter->num_active_rx_pools = rxadd_subcrqs;
 569
 570         for (i = 0; i < rxadd_subcrqs; i++) {
 571                 rx_pool = &adapter->rx_pool[i];
 572
 573                 netdev_dbg(adapter->netdev,
 574                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 575                            i, adapter->req_rx_add_entries_per_subcrq,
 576                            buff_size);
 577
 578                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 579                 rx_pool->index = i;
 580                 rx_pool->buff_size = buff_size;
 581                 rx_pool->active = 1;
 582
 583                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 584                                             GFP_KERNEL);
 585                 if (!rx_pool->free_map) {
 586                         release_rx_pools(adapter);
 587                         return -1;
 588                 }
 589
 590                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 591                                            sizeof(struct ibmvnic_rx_buff),
 592                                            GFP_KERNEL);
 593                 if (!rx_pool->rx_buff) {
 594                         dev_err(dev, "Couldn't alloc rx buffers\n");
 595                         release_rx_pools(adapter);
 596                         return -1;
 597                 }
 598
 599                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 600                                          rx_pool->size * rx_pool->buff_size)) {
 601                         release_rx_pools(adapter);
 602                         return -1;
 603                 }
 604
 605                 for (j = 0; j < rx_pool->size; ++j)
 606                         rx_pool->free_map[j] = j;
 607
 608                 atomic_set(&rx_pool->available, 0);
 609                 rx_pool->next_alloc = 0;
 610                 rx_pool->next_free = 0;
 611         }
 612
 613         return 0;
 614 }
 615
 616 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 617                              struct ibmvnic_tx_pool *tx_pool)
 618 {
 619         int rc, i;
 620
 621         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 622         if (rc)
 623                 return rc;
 624
 625         memset(tx_pool->tx_buff, 0,
 626                tx_pool->num_buffers *
 627                sizeof(struct ibmvnic_tx_buff));
 628
 629         for (i = 0; i < tx_pool->num_buffers; i++)
 630                 tx_pool->free_map[i] = i;
 631
 632         tx_pool->consumer_index = 0;
 633         tx_pool->producer_index = 0;
 634
 635         return 0;
 636 }
 637
 638 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 639 {
 640         int tx_scrqs;
 641         int i, rc;
 642
 643         if (!adapter->tx_pool)
 644                 return -1;
 645
 646         tx_scrqs = adapter->num_active_tx_pools;
 647         for (i = 0; i < tx_scrqs; i++) {
 648                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 649                 if (rc)
 650                         return rc;
 651                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 652                 if (rc)
 653                         return rc;
 654         }
 655
 656         return 0;
 657 }
 658
 659 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 660 {
 661         if (!adapter->vpd)
 662                 return;
 663
 664         kfree(adapter->vpd->buff);
 665         kfree(adapter->vpd);
 666
 667         adapter->vpd = NULL;
 668 }
 669
 670 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 671                                 struct ibmvnic_tx_pool *tx_pool)
 672 {
 673         kfree(tx_pool->tx_buff);
 674         kfree(tx_pool->free_map);
 675         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 676 }
 677
 678 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 679 {
 680         int i;
 681
 682         if (!adapter->tx_pool)
 683                 return;
 684
 685         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 686                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 687                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 688         }
 689
 690         kfree(adapter->tx_pool);
 691         adapter->tx_pool = NULL;
 692         kfree(adapter->tso_pool);
 693         adapter->tso_pool = NULL;
 694         adapter->num_active_tx_pools = 0;
 695 }
 696
 697 static int init_one_tx_pool(struct net_device *netdev,
 698                             struct ibmvnic_tx_pool *tx_pool,
 699                             int num_entries, int buf_size)
 700 {
 701         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 702         int i;
 703
 704         tx_pool->tx_buff = kcalloc(num_entries,
 705                                    sizeof(struct ibmvnic_tx_buff),
 706                                    GFP_KERNEL);
 707         if (!tx_pool->tx_buff)
 708                 return -1;
 709
 710         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 711                                  num_entries * buf_size))
 712                 return -1;
 713
 714         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 715         if (!tx_pool->free_map)
 716                 return -1;
 717
 718         for (i = 0; i < num_entries; i++)
 719                 tx_pool->free_map[i] = i;
 720
 721         tx_pool->consumer_index = 0;
 722         tx_pool->producer_index = 0;
 723         tx_pool->num_buffers = num_entries;
 724         tx_pool->buf_size = buf_size;
 725
 726         return 0;
 727 }
 728
 729 static int init_tx_pools(struct net_device *netdev)
 730 {
 731         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 732         int tx_subcrqs;
 733         int i, rc;
 734
 735         tx_subcrqs = adapter->num_active_tx_scrqs;
 736         adapter->tx_pool = kcalloc(tx_subcrqs,
 737                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 738         if (!adapter->tx_pool)
 739                 return -1;
 740
 741         adapter->tso_pool = kcalloc(tx_subcrqs,
 742                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 743         if (!adapter->tso_pool)
 744                 return -1;
 745
 746         adapter->num_active_tx_pools = tx_subcrqs;
 747
 748         for (i = 0; i < tx_subcrqs; i++) {
 749                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 750                                       adapter->req_tx_entries_per_subcrq,
 751                                       adapter->req_mtu + VLAN_HLEN);
 752                 if (rc) {
 753                         release_tx_pools(adapter);
 754                         return rc;
 755                 }
 756
 757                 rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
 758                                       IBMVNIC_TSO_BUFS,
 759                                       IBMVNIC_TSO_BUF_SZ);
 760                 if (rc) {
 761                         release_tx_pools(adapter);
 762                         return rc;
 763                 }
 764         }
 765
 766         return 0;
 767 }
 768
 769 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 770 {
 771         int i;
 772
 773         if (adapter->napi_enabled)
 774                 return;
 775
 776         for (i = 0; i < adapter->req_rx_queues; i++)
 777                 napi_enable(&adapter->napi[i]);
 778
 779         adapter->napi_enabled = true;
 780 }
 781
 782 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 783 {
 784         int i;
 785
 786         if (!adapter->napi_enabled)
 787                 return;
 788
 789         for (i = 0; i < adapter->req_rx_queues; i++) {
 790                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 791                 napi_disable(&adapter->napi[i]);
 792         }
 793
 794         adapter->napi_enabled = false;
 795 }
 796
 797 static int init_napi(struct ibmvnic_adapter *adapter)
 798 {
 799         int i;
 800
 801         adapter->napi = kcalloc(adapter->req_rx_queues,
 802                                 sizeof(struct napi_struct), GFP_KERNEL);
 803         if (!adapter->napi)
 804                 return -ENOMEM;
 805
 806         for (i = 0; i < adapter->req_rx_queues; i++) {
 807                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 808                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 809                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 810         }
 811
 812         adapter->num_active_rx_napi = adapter->req_rx_queues;
 813         return 0;
 814 }
 815
 816 static void release_napi(struct ibmvnic_adapter *adapter)
 817 {
 818         int i;
 819
 820         if (!adapter->napi)
 821                 return;
 822
 823         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 824                 netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
 825                 netif_napi_del(&adapter->napi[i]);
 826         }
 827
 828         kfree(adapter->napi);
 829         adapter->napi = NULL;
 830         adapter->num_active_rx_napi = 0;
 831         adapter->napi_enabled = false;
 832 }
 833
 834 static int ibmvnic_login(struct net_device *netdev)
 835 {
 836         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 837         unsigned long timeout = msecs_to_jiffies(30000);
 838         int retry_count = 0;
 839         int retries = 10;
 840         bool retry;
 841         int rc;
 842
 843         do {
 844                 retry = false;
 845                 if (retry_count > retries) {
 846                         netdev_warn(netdev, "Login attempts exceeded\n");
 847                         return -1;
 848                 }
 849
 850                 adapter->init_done_rc = 0;
 851                 reinit_completion(&adapter->init_done);
 852                 rc = send_login(adapter);
 853                 if (rc) {
 854                         netdev_warn(netdev, "Unable to login\n");
 855                         return rc;
 856                 }
 857
 858                 if (!wait_for_completion_timeout(&adapter->init_done,
 859                                                  timeout)) {
 860                         netdev_warn(netdev, "Login timed out, retrying...\n");
 861                         retry = true;
 862                         adapter->init_done_rc = 0;
 863                         retry_count++;
 864                         continue;
 865                 }
 866
 867                 if (adapter->init_done_rc == ABORTED) {
 868                         netdev_warn(netdev, "Login aborted, retrying...\n");
 869                         retry = true;
 870                         adapter->init_done_rc = 0;
 871                         retry_count++;
 872                         /* FW or device may be busy, so
 873                          * wait a bit before retrying login
 874                          */
 875                         msleep(500);
 876                 } else if (adapter->init_done_rc == PARTIALSUCCESS) {
 877                         retry_count++;
 878                         release_sub_crqs(adapter, 1);
 879
 880                         retry = true;
 881                         netdev_dbg(netdev,
 882                                    "Received partial success, retrying...\n");
 883                         adapter->init_done_rc = 0;
 884                         reinit_completion(&adapter->init_done);
 885                         send_query_cap(adapter);
 886                         if (!wait_for_completion_timeout(&adapter->init_done,
 887                                                          timeout)) {
 888                                 netdev_warn(netdev,
 889                                             "Capabilities query timed out\n");
 890                                 return -1;
 891                         }
 892
 893                         rc = init_sub_crqs(adapter);
 894                         if (rc) {
 895                                 netdev_warn(netdev,
 896                                             "SCRQ initialization failed\n");
 897                                 return -1;
 898                         }
 899
 900                         rc = init_sub_crq_irqs(adapter);
 901                         if (rc) {
 902                                 netdev_warn(netdev,
 903                                             "SCRQ irq initialization failed\n");
 904                                 return -1;
 905                         }
 906                 } else if (adapter->init_done_rc) {
 907                         netdev_warn(netdev, "Adapter login failed\n");
 908                         return -1;
 909                 }
 910         } while (retry);
 911
 912         __ibmvnic_set_mac(netdev, adapter->mac_addr);
 913
 914         return 0;
 915 }
 916
 917 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 918 {
 919         kfree(adapter->login_buf);
 920         adapter->login_buf = NULL;
 921 }
 922
 923 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 924 {
 925         kfree(adapter->login_rsp_buf);
 926         adapter->login_rsp_buf = NULL;
 927 }
 928
 929 static void release_resources(struct ibmvnic_adapter *adapter)
 930 {
 931         release_vpd_data(adapter);
 932
 933         release_tx_pools(adapter);
 934         release_rx_pools(adapter);
 935
 936         release_napi(adapter);
 937         release_login_rsp_buffer(adapter);
 938 }
 939
 940 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 941 {
 942         struct net_device *netdev = adapter->netdev;
 943         unsigned long timeout = msecs_to_jiffies(30000);
 944         union ibmvnic_crq crq;
 945         bool resend;
 946         int rc;
 947
 948         netdev_dbg(netdev, "setting link state %d\n", link_state);
 949
 950         memset(&crq, 0, sizeof(crq));
 951         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 952         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 953         crq.logical_link_state.link_state = link_state;
 954
 955         do {
 956                 resend = false;
 957
 958                 reinit_completion(&adapter->init_done);
 959                 rc = ibmvnic_send_crq(adapter, &crq);
 960                 if (rc) {
 961                         netdev_err(netdev, "Failed to set link state\n");
 962                         return rc;
 963                 }
 964
 965                 if (!wait_for_completion_timeout(&adapter->init_done,
 966                                                  timeout)) {
 967                         netdev_err(netdev, "timeout setting link state\n");
 968                         return -1;
 969                 }
 970
 971                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 972                         /* Partuial success, delay and re-send */
 973                         mdelay(1000);
 974                         resend = true;
 975                 } else if (adapter->init_done_rc) {
 976                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 977                                     adapter->init_done_rc);
 978                         return adapter->init_done_rc;
 979                 }
 980         } while (resend);
 981
 982         return 0;
 983 }
 984
 985 static int set_real_num_queues(struct net_device *netdev)
 986 {
 987         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 988         int rc;
 989
 990         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 991                    adapter->req_tx_queues, adapter->req_rx_queues);
 992
 993         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 994         if (rc) {
 995                 netdev_err(netdev, "failed to set the number of tx queues\n");
 996                 return rc;
 997         }
 998
 999         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
1000         if (rc)
1001                 netdev_err(netdev, "failed to set the number of rx queues\n");
1002
1003         return rc;
1004 }
1005
1006 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1007 {
1008         struct device *dev = &adapter->vdev->dev;
1009         union ibmvnic_crq crq;
1010         int len = 0;
1011         int rc;
1012
1013         if (adapter->vpd->buff)
1014                 len = adapter->vpd->len;
1015
1016         mutex_lock(&adapter->fw_lock);
1017         adapter->fw_done_rc = 0;
1018         reinit_completion(&adapter->fw_done);
1019
1020         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1021         crq.get_vpd_size.cmd = GET_VPD_SIZE;
1022         rc = ibmvnic_send_crq(adapter, &crq);
1023         if (rc) {
1024                 mutex_unlock(&adapter->fw_lock);
1025                 return rc;
1026         }
1027
1028         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1029         if (rc) {
1030                 dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1031                 mutex_unlock(&adapter->fw_lock);
1032                 return rc;
1033         }
1034         mutex_unlock(&adapter->fw_lock);
1035
1036         if (!adapter->vpd->len)
1037                 return -ENODATA;
1038
1039         if (!adapter->vpd->buff)
1040                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1041         else if (adapter->vpd->len != len)
1042                 adapter->vpd->buff =
1043                         krealloc(adapter->vpd->buff,
1044                                  adapter->vpd->len, GFP_KERNEL);
1045
1046         if (!adapter->vpd->buff) {
1047                 dev_err(dev, "Could allocate VPD buffer\n");
1048                 return -ENOMEM;
1049         }
1050
1051         adapter->vpd->dma_addr =
1052                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1053                                DMA_FROM_DEVICE);
1054         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1055                 dev_err(dev, "Could not map VPD buffer\n");
1056                 kfree(adapter->vpd->buff);
1057                 adapter->vpd->buff = NULL;
1058                 return -ENOMEM;
1059         }
1060
1061         mutex_lock(&adapter->fw_lock);
1062         adapter->fw_done_rc = 0;
1063         reinit_completion(&adapter->fw_done);
1064
1065         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1066         crq.get_vpd.cmd = GET_VPD;
1067         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1068         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1069         rc = ibmvnic_send_crq(adapter, &crq);
1070         if (rc) {
1071                 kfree(adapter->vpd->buff);
1072                 adapter->vpd->buff = NULL;
1073                 mutex_unlock(&adapter->fw_lock);
1074                 return rc;
1075         }
1076
1077         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1078         if (rc) {
1079                 dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1080                 kfree(adapter->vpd->buff);
1081                 adapter->vpd->buff = NULL;
1082                 mutex_unlock(&adapter->fw_lock);
1083                 return rc;
1084         }
1085
1086         mutex_unlock(&adapter->fw_lock);
1087         return 0;
1088 }
1089
1090 static int init_resources(struct ibmvnic_adapter *adapter)
1091 {
1092         struct net_device *netdev = adapter->netdev;
1093         int rc;
1094
1095         rc = set_real_num_queues(netdev);
1096         if (rc)
1097                 return rc;
1098
1099         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1100         if (!adapter->vpd)
1101                 return -ENOMEM;
1102
1103         /* Vital Product Data (VPD) */
1104         rc = ibmvnic_get_vpd(adapter);
1105         if (rc) {
1106                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1107                 return rc;
1108         }
1109
1110         adapter->map_id = 1;
1111
1112         rc = init_napi(adapter);
1113         if (rc)
1114                 return rc;
1115
1116         send_query_map(adapter);
1117
1118         rc = init_rx_pools(netdev);
1119         if (rc)
1120                 return rc;
1121
1122         rc = init_tx_pools(netdev);
1123         return rc;
1124 }
1125
1126 static int __ibmvnic_open(struct net_device *netdev)
1127 {
1128         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1129         enum vnic_state prev_state = adapter->state;
1130         int i, rc;
1131
1132         adapter->state = VNIC_OPENING;
1133         replenish_pools(adapter);
1134         ibmvnic_napi_enable(adapter);
1135
1136         /* We're ready to receive frames, enable the sub-crq interrupts and
1137          * set the logical link state to up
1138          */
1139         for (i = 0; i < adapter->req_rx_queues; i++) {
1140                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1141                 if (prev_state == VNIC_CLOSED)
1142                         enable_irq(adapter->rx_scrq[i]->irq);
1143                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1144         }
1145
1146         for (i = 0; i < adapter->req_tx_queues; i++) {
1147                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1148                 if (prev_state == VNIC_CLOSED)
1149                         enable_irq(adapter->tx_scrq[i]->irq);
1150                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1151         }
1152
1153         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1154         if (rc) {
1155                 for (i = 0; i < adapter->req_rx_queues; i++)
1156                         napi_disable(&adapter->napi[i]);
1157                 release_resources(adapter);
1158                 return rc;
1159         }
1160
1161         netif_tx_start_all_queues(netdev);
1162
1163         if (prev_state == VNIC_CLOSED) {
1164                 for (i = 0; i < adapter->req_rx_queues; i++)
1165                         napi_schedule(&adapter->napi[i]);
1166         }
1167
1168         adapter->state = VNIC_OPEN;
1169         return rc;
1170 }
1171
1172 static int ibmvnic_open(struct net_device *netdev)
1173 {
1174         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1175         int rc;
1176
1177         /* If device failover is pending, just set device state and return.
1178          * Device operation will be handled by reset routine.
1179          */
1180         if (adapter->failover_pending) {
1181                 adapter->state = VNIC_OPEN;
1182                 return 0;
1183         }
1184
1185         if (adapter->state != VNIC_CLOSED) {
1186                 rc = ibmvnic_login(netdev);
1187                 if (rc)
1188                         return rc;
1189
1190                 rc = init_resources(adapter);
1191                 if (rc) {
1192                         netdev_err(netdev, "failed to initialize resources\n");
1193                         release_resources(adapter);
1194                         return rc;
1195                 }
1196         }
1197
1198         rc = __ibmvnic_open(netdev);
1199
1200         return rc;
1201 }
1202
1203 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1204 {
1205         struct ibmvnic_rx_pool *rx_pool;
1206         struct ibmvnic_rx_buff *rx_buff;
1207         u64 rx_entries;
1208         int rx_scrqs;
1209         int i, j;
1210
1211         if (!adapter->rx_pool)
1212                 return;
1213
1214         rx_scrqs = adapter->num_active_rx_pools;
1215         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1216
1217         /* Free any remaining skbs in the rx buffer pools */
1218         for (i = 0; i < rx_scrqs; i++) {
1219                 rx_pool = &adapter->rx_pool[i];
1220                 if (!rx_pool || !rx_pool->rx_buff)
1221                         continue;
1222
1223                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1224                 for (j = 0; j < rx_entries; j++) {
1225                         rx_buff = &rx_pool->rx_buff[j];
1226                         if (rx_buff && rx_buff->skb) {
1227                                 dev_kfree_skb_any(rx_buff->skb);
1228                                 rx_buff->skb = NULL;
1229                         }
1230                 }
1231         }
1232 }
1233
1234 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1235                               struct ibmvnic_tx_pool *tx_pool)
1236 {
1237         struct ibmvnic_tx_buff *tx_buff;
1238         u64 tx_entries;
1239         int i;
1240
1241         if (!tx_pool || !tx_pool->tx_buff)
1242                 return;
1243
1244         tx_entries = tx_pool->num_buffers;
1245
1246         for (i = 0; i < tx_entries; i++) {
1247                 tx_buff = &tx_pool->tx_buff[i];
1248                 if (tx_buff && tx_buff->skb) {
1249                         dev_kfree_skb_any(tx_buff->skb);
1250                         tx_buff->skb = NULL;
1251                 }
1252         }
1253 }
1254
1255 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1256 {
1257         int tx_scrqs;
1258         int i;
1259
1260         if (!adapter->tx_pool || !adapter->tso_pool)
1261                 return;
1262
1263         tx_scrqs = adapter->num_active_tx_pools;
1264
1265         /* Free any remaining skbs in the tx buffer pools */
1266         for (i = 0; i < tx_scrqs; i++) {
1267                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1268                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1269                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1270         }
1271 }
1272
1273 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1274 {
1275         struct net_device *netdev = adapter->netdev;
1276         int i;
1277
1278         if (adapter->tx_scrq) {
1279                 for (i = 0; i < adapter->req_tx_queues; i++)
1280                         if (adapter->tx_scrq[i]->irq) {
1281                                 netdev_dbg(netdev,
1282                                            "Disabling tx_scrq[%d] irq\n", i);
1283                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1284                                 disable_irq(adapter->tx_scrq[i]->irq);
1285                         }
1286         }
1287
1288         if (adapter->rx_scrq) {
1289                 for (i = 0; i < adapter->req_rx_queues; i++) {
1290                         if (adapter->rx_scrq[i]->irq) {
1291                                 netdev_dbg(netdev,
1292                                            "Disabling rx_scrq[%d] irq\n", i);
1293                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1294                                 disable_irq(adapter->rx_scrq[i]->irq);
1295                         }
1296                 }
1297         }
1298 }
1299
1300 static void ibmvnic_cleanup(struct net_device *netdev)
1301 {
1302         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1303
1304         /* ensure that transmissions are stopped if called by do_reset */
1305         if (test_bit(0, &adapter->resetting))
1306                 netif_tx_disable(netdev);
1307         else
1308                 netif_tx_stop_all_queues(netdev);
1309
1310         ibmvnic_napi_disable(adapter);
1311         ibmvnic_disable_irqs(adapter);
1312
1313         clean_rx_pools(adapter);
1314         clean_tx_pools(adapter);
1315 }
1316
1317 static int __ibmvnic_close(struct net_device *netdev)
1318 {
1319         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1320         int rc = 0;
1321
1322         adapter->state = VNIC_CLOSING;
1323         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1324         if (rc)
1325                 return rc;
1326         adapter->state = VNIC_CLOSED;
1327         return 0;
1328 }
1329
1330 static int ibmvnic_close(struct net_device *netdev)
1331 {
1332         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1333         int rc;
1334
1335         /* If device failover is pending, just set device state and return.
1336          * Device operation will be handled by reset routine.
1337          */
1338         if (adapter->failover_pending) {
1339                 adapter->state = VNIC_CLOSED;
1340                 return 0;
1341         }
1342
1343         rc = __ibmvnic_close(netdev);
1344         ibmvnic_cleanup(netdev);
1345
1346         return rc;
1347 }
1348
1349 /**
1350  * build_hdr_data - creates L2/L3/L4 header data buffer
1351  * @hdr_field - bitfield determining needed headers
1352  * @skb - socket buffer
1353  * @hdr_len - array of header lengths
1354  * @tot_len - total length of data
1355  *
1356  * Reads hdr_field to determine which headers are needed by firmware.
1357  * Builds a buffer containing these headers.  Saves individual header
1358  * lengths and total buffer length to be used to build descriptors.
1359  */
1360 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1361                           int *hdr_len, u8 *hdr_data)
1362 {
1363         int len = 0;
1364         u8 *hdr;
1365
1366         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1367                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1368         else
1369                 hdr_len[0] = sizeof(struct ethhdr);
1370
1371         if (skb->protocol == htons(ETH_P_IP)) {
1372                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1373                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1374                         hdr_len[2] = tcp_hdrlen(skb);
1375                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1376                         hdr_len[2] = sizeof(struct udphdr);
1377         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1378                 hdr_len[1] = sizeof(struct ipv6hdr);
1379                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1380                         hdr_len[2] = tcp_hdrlen(skb);
1381                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1382                         hdr_len[2] = sizeof(struct udphdr);
1383         } else if (skb->protocol == htons(ETH_P_ARP)) {
1384                 hdr_len[1] = arp_hdr_len(skb->dev);
1385                 hdr_len[2] = 0;
1386         }
1387
1388         memset(hdr_data, 0, 120);
1389         if ((hdr_field >> 6) & 1) {
1390                 hdr = skb_mac_header(skb);
1391                 memcpy(hdr_data, hdr, hdr_len[0]);
1392                 len += hdr_len[0];
1393         }
1394
1395         if ((hdr_field >> 5) & 1) {
1396                 hdr = skb_network_header(skb);
1397                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1398                 len += hdr_len[1];
1399         }
1400
1401         if ((hdr_field >> 4) & 1) {
1402                 hdr = skb_transport_header(skb);
1403                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1404                 len += hdr_len[2];
1405         }
1406         return len;
1407 }
1408
1409 /**
1410  * create_hdr_descs - create header and header extension descriptors
1411  * @hdr_field - bitfield determining needed headers
1412  * @data - buffer containing header data
1413  * @len - length of data buffer
1414  * @hdr_len - array of individual header lengths
1415  * @scrq_arr - descriptor array
1416  *
1417  * Creates header and, if needed, header extension descriptors and
1418  * places them in a descriptor array, scrq_arr
1419  */
1420
1421 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1422                             union sub_crq *scrq_arr)
1423 {
1424         union sub_crq hdr_desc;
1425         int tmp_len = len;
1426         int num_descs = 0;
1427         u8 *data, *cur;
1428         int tmp;
1429
1430         while (tmp_len > 0) {
1431                 cur = hdr_data + len - tmp_len;
1432
1433                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1434                 if (cur != hdr_data) {
1435                         data = hdr_desc.hdr_ext.data;
1436                         tmp = tmp_len > 29 ? 29 : tmp_len;
1437                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1438                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1439                         hdr_desc.hdr_ext.len = tmp;
1440                 } else {
1441                         data = hdr_desc.hdr.data;
1442                         tmp = tmp_len > 24 ? 24 : tmp_len;
1443                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1444                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1445                         hdr_desc.hdr.len = tmp;
1446                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1447                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1448                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1449                         hdr_desc.hdr.flag = hdr_field << 1;
1450                 }
1451                 memcpy(data, cur, tmp);
1452                 tmp_len -= tmp;
1453                 *scrq_arr = hdr_desc;
1454                 scrq_arr++;
1455                 num_descs++;
1456         }
1457
1458         return num_descs;
1459 }
1460
1461 /**
1462  * build_hdr_descs_arr - build a header descriptor array
1463  * @skb - socket buffer
1464  * @num_entries - number of descriptors to be sent
1465  * @subcrq - first TX descriptor
1466  * @hdr_field - bit field determining which headers will be sent
1467  *
1468  * This function will build a TX descriptor array with applicable
1469  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1470  */
1471
1472 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1473                                 int *num_entries, u8 hdr_field)
1474 {
1475         int hdr_len[3] = {0, 0, 0};
1476         int tot_len;
1477         u8 *hdr_data = txbuff->hdr_data;
1478
1479         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1480                                  txbuff->hdr_data);
1481         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1482                          txbuff->indir_arr + 1);
1483 }
1484
1485 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1486                                     struct net_device *netdev)
1487 {
1488         /* For some backing devices, mishandling of small packets
1489          * can result in a loss of connection or TX stall. Device
1490          * architects recommend that no packet should be smaller
1491          * than the minimum MTU value provided to the driver, so
1492          * pad any packets to that length
1493          */
1494         if (skb->len < netdev->min_mtu)
1495                 return skb_put_padto(skb, netdev->min_mtu);
1496
1497         return 0;
1498 }
1499
1500 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1501 {
1502         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1503         int queue_num = skb_get_queue_mapping(skb);
1504         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1505         struct device *dev = &adapter->vdev->dev;
1506         struct ibmvnic_tx_buff *tx_buff = NULL;
1507         struct ibmvnic_sub_crq_queue *tx_scrq;
1508         struct ibmvnic_tx_pool *tx_pool;
1509         unsigned int tx_send_failed = 0;
1510         unsigned int tx_map_failed = 0;
1511         unsigned int tx_dropped = 0;
1512         unsigned int tx_packets = 0;
1513         unsigned int tx_bytes = 0;
1514         dma_addr_t data_dma_addr;
1515         struct netdev_queue *txq;
1516         unsigned long lpar_rc;
1517         union sub_crq tx_crq;
1518         unsigned int offset;
1519         int num_entries = 1;
1520         unsigned char *dst;
1521         int index = 0;
1522         u8 proto = 0;
1523         u64 handle;
1524         netdev_tx_t ret = NETDEV_TX_OK;
1525
1526         if (test_bit(0, &adapter->resetting)) {
1527                 if (!netif_subqueue_stopped(netdev, skb))
1528                         netif_stop_subqueue(netdev, queue_num);
1529                 dev_kfree_skb_any(skb);
1530
1531                 tx_send_failed++;
1532                 tx_dropped++;
1533                 ret = NETDEV_TX_OK;
1534                 goto out;
1535         }
1536
1537         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1538                 tx_dropped++;
1539                 tx_send_failed++;
1540                 ret = NETDEV_TX_OK;
1541                 goto out;
1542         }
1543         if (skb_is_gso(skb))
1544                 tx_pool = &adapter->tso_pool[queue_num];
1545         else
1546                 tx_pool = &adapter->tx_pool[queue_num];
1547
1548         tx_scrq = adapter->tx_scrq[queue_num];
1549         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1550         handle = tx_scrq->handle;
1551
1552         index = tx_pool->free_map[tx_pool->consumer_index];
1553
1554         if (index == IBMVNIC_INVALID_MAP) {
1555                 dev_kfree_skb_any(skb);
1556                 tx_send_failed++;
1557                 tx_dropped++;
1558                 ret = NETDEV_TX_OK;
1559                 goto out;
1560         }
1561
1562         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1563
1564         offset = index * tx_pool->buf_size;
1565         dst = tx_pool->long_term_buff.buff + offset;
1566         memset(dst, 0, tx_pool->buf_size);
1567         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1568
1569         if (skb_shinfo(skb)->nr_frags) {
1570                 int cur, i;
1571
1572                 /* Copy the head */
1573                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1574                 cur = skb_headlen(skb);
1575
1576                 /* Copy the frags */
1577                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1578                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1579
1580                         memcpy(dst + cur,
1581                                page_address(skb_frag_page(frag)) +
1582                                skb_frag_off(frag), skb_frag_size(frag));
1583                         cur += skb_frag_size(frag);
1584                 }
1585         } else {
1586                 skb_copy_from_linear_data(skb, dst, skb->len);
1587         }
1588
1589         tx_pool->consumer_index =
1590             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1591
1592         tx_buff = &tx_pool->tx_buff[index];
1593         tx_buff->skb = skb;
1594         tx_buff->data_dma[0] = data_dma_addr;
1595         tx_buff->data_len[0] = skb->len;
1596         tx_buff->index = index;
1597         tx_buff->pool_index = queue_num;
1598         tx_buff->last_frag = true;
1599
1600         memset(&tx_crq, 0, sizeof(tx_crq));
1601         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1602         tx_crq.v1.type = IBMVNIC_TX_DESC;
1603         tx_crq.v1.n_crq_elem = 1;
1604         tx_crq.v1.n_sge = 1;
1605         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1606
1607         if (skb_is_gso(skb))
1608                 tx_crq.v1.correlator =
1609                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1610         else
1611                 tx_crq.v1.correlator = cpu_to_be32(index);
1612         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1613         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1614         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1615
1616         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1617                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1618                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1619         }
1620
1621         if (skb->protocol == htons(ETH_P_IP)) {
1622                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1623                 proto = ip_hdr(skb)->protocol;
1624         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1625                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1626                 proto = ipv6_hdr(skb)->nexthdr;
1627         }
1628
1629         if (proto == IPPROTO_TCP)
1630                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1631         else if (proto == IPPROTO_UDP)
1632                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1633
1634         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1635                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1636                 hdrs += 2;
1637         }
1638         if (skb_is_gso(skb)) {
1639                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1640                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1641                 hdrs += 2;
1642         }
1643         /* determine if l2/3/4 headers are sent to firmware */
1644         if ((*hdrs >> 7) & 1) {
1645                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1646                 tx_crq.v1.n_crq_elem = num_entries;
1647                 tx_buff->num_entries = num_entries;
1648                 tx_buff->indir_arr[0] = tx_crq;
1649                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1650                                                     sizeof(tx_buff->indir_arr),
1651                                                     DMA_TO_DEVICE);
1652                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1653                         dev_kfree_skb_any(skb);
1654                         tx_buff->skb = NULL;
1655                         if (!firmware_has_feature(FW_FEATURE_CMO))
1656                                 dev_err(dev, "tx: unable to map descriptor array\n");
1657                         tx_map_failed++;
1658                         tx_dropped++;
1659                         ret = NETDEV_TX_OK;
1660                         goto tx_err_out;
1661                 }
1662                 lpar_rc = send_subcrq_indirect(adapter, handle,
1663                                                (u64)tx_buff->indir_dma,
1664                                                (u64)num_entries);
1665                 dma_unmap_single(dev, tx_buff->indir_dma,
1666                                  sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1667         } else {
1668                 tx_buff->num_entries = num_entries;
1669                 lpar_rc = send_subcrq(adapter, handle,
1670                                       &tx_crq);
1671         }
1672         if (lpar_rc != H_SUCCESS) {
1673                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1674                         dev_err_ratelimited(dev, "tx: send failed\n");
1675                 dev_kfree_skb_any(skb);
1676                 tx_buff->skb = NULL;
1677
1678                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1679                         /* Disable TX and report carrier off if queue is closed
1680                          * or pending failover.
1681                          * Firmware guarantees that a signal will be sent to the
1682                          * driver, triggering a reset or some other action.
1683                          */
1684                         netif_tx_stop_all_queues(netdev);
1685                         netif_carrier_off(netdev);
1686                 }
1687
1688                 tx_send_failed++;
1689                 tx_dropped++;
1690                 ret = NETDEV_TX_OK;
1691                 goto tx_err_out;
1692         }
1693
1694         if (atomic_add_return(num_entries, &tx_scrq->used)
1695                                         >= adapter->req_tx_entries_per_subcrq) {
1696                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1697                 netif_stop_subqueue(netdev, queue_num);
1698         }
1699
1700         tx_packets++;
1701         tx_bytes += skb->len;
1702         txq->trans_start = jiffies;
1703         ret = NETDEV_TX_OK;
1704         goto out;
1705
1706 tx_err_out:
1707         /* roll back consumer index and map array*/
1708         if (tx_pool->consumer_index == 0)
1709                 tx_pool->consumer_index =
1710                         tx_pool->num_buffers - 1;
1711         else
1712                 tx_pool->consumer_index--;
1713         tx_pool->free_map[tx_pool->consumer_index] = index;
1714 out:
1715         netdev->stats.tx_dropped += tx_dropped;
1716         netdev->stats.tx_bytes += tx_bytes;
1717         netdev->stats.tx_packets += tx_packets;
1718         adapter->tx_send_failed += tx_send_failed;
1719         adapter->tx_map_failed += tx_map_failed;
1720         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1721         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1722         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1723
1724         return ret;
1725 }
1726
1727 static void ibmvnic_set_multi(struct net_device *netdev)
1728 {
1729         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1730         struct netdev_hw_addr *ha;
1731         union ibmvnic_crq crq;
1732
1733         memset(&crq, 0, sizeof(crq));
1734         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1735         crq.request_capability.cmd = REQUEST_CAPABILITY;
1736
1737         if (netdev->flags & IFF_PROMISC) {
1738                 if (!adapter->promisc_supported)
1739                         return;
1740         } else {
1741                 if (netdev->flags & IFF_ALLMULTI) {
1742                         /* Accept all multicast */
1743                         memset(&crq, 0, sizeof(crq));
1744                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1745                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1746                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1747                         ibmvnic_send_crq(adapter, &crq);
1748                 } else if (netdev_mc_empty(netdev)) {
1749                         /* Reject all multicast */
1750                         memset(&crq, 0, sizeof(crq));
1751                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1752                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1753                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1754                         ibmvnic_send_crq(adapter, &crq);
1755                 } else {
1756                         /* Accept one or more multicast(s) */
1757                         netdev_for_each_mc_addr(ha, netdev) {
1758                                 memset(&crq, 0, sizeof(crq));
1759                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1760                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1761                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1762                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1763                                                 ha->addr);
1764                                 ibmvnic_send_crq(adapter, &crq);
1765                         }
1766                 }
1767         }
1768 }
1769
1770 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1771 {
1772         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1773         union ibmvnic_crq crq;
1774         int rc;
1775
1776         if (!is_valid_ether_addr(dev_addr)) {
1777                 rc = -EADDRNOTAVAIL;
1778                 goto err;
1779         }
1780
1781         memset(&crq, 0, sizeof(crq));
1782         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1783         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1784         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1785
1786         mutex_lock(&adapter->fw_lock);
1787         adapter->fw_done_rc = 0;
1788         reinit_completion(&adapter->fw_done);
1789
1790         rc = ibmvnic_send_crq(adapter, &crq);
1791         if (rc) {
1792                 rc = -EIO;
1793                 mutex_unlock(&adapter->fw_lock);
1794                 goto err;
1795         }
1796
1797         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1798         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1799         if (rc || adapter->fw_done_rc) {
1800                 rc = -EIO;
1801                 mutex_unlock(&adapter->fw_lock);
1802                 goto err;
1803         }
1804         mutex_unlock(&adapter->fw_lock);
1805         return 0;
1806 err:
1807         ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1808         return rc;
1809 }
1810
1811 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1812 {
1813         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1814         struct sockaddr *addr = p;
1815         int rc;
1816
1817         rc = 0;
1818         if (!is_valid_ether_addr(addr->sa_data))
1819                 return -EADDRNOTAVAIL;
1820
1821         if (adapter->state != VNIC_PROBED) {
1822                 ether_addr_copy(adapter->mac_addr, addr->sa_data);
1823                 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1824         }
1825
1826         return rc;
1827 }
1828
1829 /**
1830  * do_change_param_reset returns zero if we are able to keep processing reset
1831  * events, or non-zero if we hit a fatal error and must halt.
1832  */
1833 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1834                                  struct ibmvnic_rwi *rwi,
1835                                  u32 reset_state)
1836 {
1837         struct net_device *netdev = adapter->netdev;
1838         int i, rc;
1839
1840         netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1841                    rwi->reset_reason);
1842
1843         netif_carrier_off(netdev);
1844         adapter->reset_reason = rwi->reset_reason;
1845
1846         ibmvnic_cleanup(netdev);
1847
1848         if (reset_state == VNIC_OPEN) {
1849                 rc = __ibmvnic_close(netdev);
1850                 if (rc)
1851                         return rc;
1852         }
1853
1854         release_resources(adapter);
1855         release_sub_crqs(adapter, 1);
1856         release_crq_queue(adapter);
1857
1858         adapter->state = VNIC_PROBED;
1859
1860         rc = init_crq_queue(adapter);
1861
1862         if (rc) {
1863                 netdev_err(adapter->netdev,
1864                            "Couldn't initialize crq. rc=%d\n", rc);
1865                 return rc;
1866         }
1867
1868         rc = ibmvnic_reset_init(adapter, true);
1869         if (rc)
1870                 return IBMVNIC_INIT_FAILED;
1871
1872         /* If the adapter was in PROBE state prior to the reset,
1873          * exit here.
1874          */
1875         if (reset_state == VNIC_PROBED)
1876                 return 0;
1877
1878         rc = ibmvnic_login(netdev);
1879         if (rc) {
1880                 adapter->state = reset_state;
1881                 return rc;
1882         }
1883
1884         rc = init_resources(adapter);
1885         if (rc)
1886                 return rc;
1887
1888         ibmvnic_disable_irqs(adapter);
1889
1890         adapter->state = VNIC_CLOSED;
1891
1892         if (reset_state == VNIC_CLOSED)
1893                 return 0;
1894
1895         rc = __ibmvnic_open(netdev);
1896         if (rc)
1897                 return IBMVNIC_OPEN_FAILED;
1898
1899         /* refresh device's multicast list */
1900         ibmvnic_set_multi(netdev);
1901
1902         /* kick napi */
1903         for (i = 0; i < adapter->req_rx_queues; i++)
1904                 napi_schedule(&adapter->napi[i]);
1905
1906         return 0;
1907 }
1908
1909 /**
1910  * do_reset returns zero if we are able to keep processing reset events, or
1911  * non-zero if we hit a fatal error and must halt.
1912  */
1913 static int do_reset(struct ibmvnic_adapter *adapter,
1914                     struct ibmvnic_rwi *rwi, u32 reset_state)
1915 {
1916         u64 old_num_rx_queues, old_num_tx_queues;
1917         u64 old_num_rx_slots, old_num_tx_slots;
1918         struct net_device *netdev = adapter->netdev;
1919         int i, rc;
1920
1921         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1922                    rwi->reset_reason);
1923
1924         rtnl_lock();
1925
1926         netif_carrier_off(netdev);
1927         adapter->reset_reason = rwi->reset_reason;
1928
1929         old_num_rx_queues = adapter->req_rx_queues;
1930         old_num_tx_queues = adapter->req_tx_queues;
1931         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1932         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1933
1934         ibmvnic_cleanup(netdev);
1935
1936         if (reset_state == VNIC_OPEN &&
1937             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1938             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1939                 adapter->state = VNIC_CLOSING;
1940
1941                 /* Release the RTNL lock before link state change and
1942                  * re-acquire after the link state change to allow
1943                  * linkwatch_event to grab the RTNL lock and run during
1944                  * a reset.
1945                  */
1946                 rtnl_unlock();
1947                 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1948                 rtnl_lock();
1949                 if (rc)
1950                         goto out;
1951
1952                 if (adapter->state != VNIC_CLOSING) {
1953                         rc = -1;
1954                         goto out;
1955                 }
1956
1957                 adapter->state = VNIC_CLOSED;
1958         }
1959
1960         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1961                 /* remove the closed state so when we call open it appears
1962                  * we are coming from the probed state.
1963                  */
1964                 adapter->state = VNIC_PROBED;
1965
1966                 if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1967                         rc = ibmvnic_reenable_crq_queue(adapter);
1968                         release_sub_crqs(adapter, 1);
1969                 } else {
1970                         rc = ibmvnic_reset_crq(adapter);
1971                         if (rc == H_CLOSED || rc == H_SUCCESS) {
1972                                 rc = vio_enable_interrupts(adapter->vdev);
1973                                 if (rc)
1974                                         netdev_err(adapter->netdev,
1975                                                    "Reset failed to enable interrupts. rc=%d\n",
1976                                                    rc);
1977                         }
1978                 }
1979
1980                 if (rc) {
1981                         netdev_err(adapter->netdev,
1982                                    "Reset couldn't initialize crq. rc=%d\n", rc);
1983                         goto out;
1984                 }
1985
1986                 rc = ibmvnic_reset_init(adapter, true);
1987                 if (rc) {
1988                         rc = IBMVNIC_INIT_FAILED;
1989                         goto out;
1990                 }
1991
1992                 /* If the adapter was in PROBE state prior to the reset,
1993                  * exit here.
1994                  */
1995                 if (reset_state == VNIC_PROBED) {
1996                         rc = 0;
1997                         goto out;
1998                 }
1999
2000                 rc = ibmvnic_login(netdev);
2001                 if (rc) {
2002                         adapter->state = reset_state;
2003                         goto out;
2004                 }
2005
2006                 if (adapter->req_rx_queues != old_num_rx_queues ||
2007                     adapter->req_tx_queues != old_num_tx_queues ||
2008                     adapter->req_rx_add_entries_per_subcrq !=
2009                     old_num_rx_slots ||
2010                     adapter->req_tx_entries_per_subcrq !=
2011                     old_num_tx_slots ||
2012                     !adapter->rx_pool ||
2013                     !adapter->tso_pool ||
2014                     !adapter->tx_pool) {
2015                         release_rx_pools(adapter);
2016                         release_tx_pools(adapter);
2017                         release_napi(adapter);
2018                         release_vpd_data(adapter);
2019
2020                         rc = init_resources(adapter);
2021                         if (rc)
2022                                 goto out;
2023
2024                 } else {
2025                         rc = reset_tx_pools(adapter);
2026                         if (rc) {
2027                                 netdev_dbg(adapter->netdev, "reset tx pools failed (%d)\n",
2028                                                 rc);
2029                                 goto out;
2030                         }
2031
2032                         rc = reset_rx_pools(adapter);
2033                         if (rc) {
2034                                 netdev_dbg(adapter->netdev, "reset rx pools failed (%d)\n",
2035                                                 rc);
2036                                 goto out;
2037                         }
2038                 }
2039                 ibmvnic_disable_irqs(adapter);
2040         }
2041         adapter->state = VNIC_CLOSED;
2042
2043         if (reset_state == VNIC_CLOSED) {
2044                 rc = 0;
2045                 goto out;
2046         }
2047
2048         rc = __ibmvnic_open(netdev);
2049         if (rc) {
2050                 rc = IBMVNIC_OPEN_FAILED;
2051                 goto out;
2052         }
2053
2054         /* refresh device's multicast list */
2055         ibmvnic_set_multi(netdev);
2056
2057         /* kick napi */
2058         for (i = 0; i < adapter->req_rx_queues; i++)
2059                 napi_schedule(&adapter->napi[i]);
2060
2061         if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2062                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2063
2064         rc = 0;
2065
2066 out:
2067         rtnl_unlock();
2068
2069         return rc;
2070 }
2071
2072 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2073                          struct ibmvnic_rwi *rwi, u32 reset_state)
2074 {
2075         struct net_device *netdev = adapter->netdev;
2076         int rc;
2077
2078         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2079                    rwi->reset_reason);
2080
2081         netif_carrier_off(netdev);
2082         adapter->reset_reason = rwi->reset_reason;
2083
2084         ibmvnic_cleanup(netdev);
2085         release_resources(adapter);
2086         release_sub_crqs(adapter, 0);
2087         release_crq_queue(adapter);
2088
2089         /* remove the closed state so when we call open it appears
2090          * we are coming from the probed state.
2091          */
2092         adapter->state = VNIC_PROBED;
2093
2094         reinit_completion(&adapter->init_done);
2095         rc = init_crq_queue(adapter);
2096         if (rc) {
2097                 netdev_err(adapter->netdev,
2098                            "Couldn't initialize crq. rc=%d\n", rc);
2099                 return rc;
2100         }
2101
2102         rc = ibmvnic_reset_init(adapter, false);
2103         if (rc)
2104                 return rc;
2105
2106         /* If the adapter was in PROBE state prior to the reset,
2107          * exit here.
2108          */
2109         if (reset_state == VNIC_PROBED)
2110                 return 0;
2111
2112         rc = ibmvnic_login(netdev);
2113         if (rc) {
2114                 adapter->state = VNIC_PROBED;
2115                 return 0;
2116         }
2117
2118         rc = init_resources(adapter);
2119         if (rc)
2120                 return rc;
2121
2122         ibmvnic_disable_irqs(adapter);
2123         adapter->state = VNIC_CLOSED;
2124
2125         if (reset_state == VNIC_CLOSED)
2126                 return 0;
2127
2128         rc = __ibmvnic_open(netdev);
2129         if (rc)
2130                 return IBMVNIC_OPEN_FAILED;
2131
2132         return 0;
2133 }
2134
2135 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2136 {
2137         struct ibmvnic_rwi *rwi;
2138         unsigned long flags;
2139
2140         spin_lock_irqsave(&adapter->rwi_lock, flags);
2141
2142         if (!list_empty(&adapter->rwi_list)) {
2143                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2144                                        list);
2145                 list_del(&rwi->list);
2146         } else {
2147                 rwi = NULL;
2148         }
2149
2150         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2151         return rwi;
2152 }
2153
2154 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2155 {
2156         struct ibmvnic_rwi *rwi;
2157
2158         rwi = get_next_rwi(adapter);
2159         while (rwi) {
2160                 kfree(rwi);
2161                 rwi = get_next_rwi(adapter);
2162         }
2163 }
2164
2165 static void __ibmvnic_reset(struct work_struct *work)
2166 {
2167         struct ibmvnic_rwi *rwi;
2168         struct ibmvnic_adapter *adapter;
2169         bool saved_state = false;
2170         unsigned long flags;
2171         u32 reset_state;
2172         int rc = 0;
2173
2174         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2175
2176         if (test_and_set_bit_lock(0, &adapter->resetting)) {
2177                 schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2178                                       IBMVNIC_RESET_DELAY);
2179                 return;
2180         }
2181
2182         rwi = get_next_rwi(adapter);
2183         while (rwi) {
2184                 spin_lock_irqsave(&adapter->state_lock, flags);
2185
2186                 if (adapter->state == VNIC_REMOVING ||
2187                     adapter->state == VNIC_REMOVED) {
2188                         spin_unlock_irqrestore(&adapter->state_lock, flags);
2189                         kfree(rwi);
2190                         rc = EBUSY;
2191                         break;
2192                 }
2193
2194                 if (!saved_state) {
2195                         reset_state = adapter->state;
2196                         adapter->state = VNIC_RESETTING;
2197                         saved_state = true;
2198                 }
2199                 spin_unlock_irqrestore(&adapter->state_lock, flags);
2200
2201                 if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2202                         /* CHANGE_PARAM requestor holds rtnl_lock */
2203                         rc = do_change_param_reset(adapter, rwi, reset_state);
2204                 } else if (adapter->force_reset_recovery) {
2205                         /* Transport event occurred during previous reset */
2206                         if (adapter->wait_for_reset) {
2207                                 /* Previous was CHANGE_PARAM; caller locked */
2208                                 adapter->force_reset_recovery = false;
2209                                 rc = do_hard_reset(adapter, rwi, reset_state);
2210                         } else {
2211                                 rtnl_lock();
2212                                 adapter->force_reset_recovery = false;
2213                                 rc = do_hard_reset(adapter, rwi, reset_state);
2214                                 rtnl_unlock();
2215                         }
2216                 } else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2217                                 adapter->from_passive_init)) {
2218                         rc = do_reset(adapter, rwi, reset_state);
2219                 }
2220                 kfree(rwi);
2221                 if (rc == IBMVNIC_OPEN_FAILED) {
2222                         if (list_empty(&adapter->rwi_list))
2223                                 adapter->state = VNIC_CLOSED;
2224                         else
2225                                 adapter->state = reset_state;
2226                         rc = 0;
2227                 } else if (rc && rc != IBMVNIC_INIT_FAILED &&
2228                     !adapter->force_reset_recovery)
2229                         break;
2230
2231                 rwi = get_next_rwi(adapter);
2232
2233                 if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2234                             rwi->reset_reason == VNIC_RESET_MOBILITY))
2235                         adapter->force_reset_recovery = true;
2236         }
2237
2238         if (adapter->wait_for_reset) {
2239                 adapter->reset_done_rc = rc;
2240                 complete(&adapter->reset_done);
2241         }
2242
2243         if (rc) {
2244                 netdev_dbg(adapter->netdev, "Reset failed\n");
2245                 free_all_rwi(adapter);
2246         }
2247
2248         clear_bit_unlock(0, &adapter->resetting);
2249 }
2250
2251 static void __ibmvnic_delayed_reset(struct work_struct *work)
2252 {
2253         struct ibmvnic_adapter *adapter;
2254
2255         adapter = container_of(work, struct ibmvnic_adapter,
2256                                ibmvnic_delayed_reset.work);
2257         __ibmvnic_reset(&adapter->ibmvnic_reset);
2258 }
2259
2260 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2261                          enum ibmvnic_reset_reason reason)
2262 {
2263         struct list_head *entry, *tmp_entry;
2264         struct ibmvnic_rwi *rwi, *tmp;
2265         struct net_device *netdev = adapter->netdev;
2266         unsigned long flags;
2267         int ret;
2268
2269         if (adapter->state == VNIC_REMOVING ||
2270             adapter->state == VNIC_REMOVED ||
2271             adapter->failover_pending) {
2272                 ret = EBUSY;
2273                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2274                 goto err;
2275         }
2276
2277         if (adapter->state == VNIC_PROBING) {
2278                 netdev_warn(netdev, "Adapter reset during probe\n");
2279                 ret = adapter->init_done_rc = EAGAIN;
2280                 goto err;
2281         }
2282
2283         spin_lock_irqsave(&adapter->rwi_lock, flags);
2284
2285         list_for_each(entry, &adapter->rwi_list) {
2286                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2287                 if (tmp->reset_reason == reason) {
2288                         netdev_dbg(netdev, "Skipping matching reset\n");
2289                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2290                         ret = EBUSY;
2291                         goto err;
2292                 }
2293         }
2294
2295         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2296         if (!rwi) {
2297                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2298                 ibmvnic_close(netdev);
2299                 ret = ENOMEM;
2300                 goto err;
2301         }
2302         /* if we just received a transport event,
2303          * flush reset queue and process this reset
2304          */
2305         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2306                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2307                         list_del(entry);
2308         }
2309         rwi->reset_reason = reason;
2310         list_add_tail(&rwi->list, &adapter->rwi_list);
2311         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2312         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2313         schedule_work(&adapter->ibmvnic_reset);
2314
2315         return 0;
2316 err:
2317         return -ret;
2318 }
2319
2320 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2321 {
2322         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2323
2324         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2325 }
2326
2327 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2328                                   struct ibmvnic_rx_buff *rx_buff)
2329 {
2330         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2331
2332         rx_buff->skb = NULL;
2333
2334         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2335         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2336
2337         atomic_dec(&pool->available);
2338 }
2339
2340 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2341 {
2342         struct net_device *netdev = napi->dev;
2343         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2344         int scrq_num = (int)(napi - adapter->napi);
2345         int frames_processed = 0;
2346
2347 restart_poll:
2348         while (frames_processed < budget) {
2349                 struct sk_buff *skb;
2350                 struct ibmvnic_rx_buff *rx_buff;
2351                 union sub_crq *next;
2352                 u32 length;
2353                 u16 offset;
2354                 u8 flags = 0;
2355
2356                 if (unlikely(test_bit(0, &adapter->resetting) &&
2357                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2358                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2359                         napi_complete_done(napi, frames_processed);
2360                         return frames_processed;
2361                 }
2362
2363                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2364                         break;
2365                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2366                 rx_buff =
2367                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2368                                                           rx_comp.correlator);
2369                 /* do error checking */
2370                 if (next->rx_comp.rc) {
2371                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2372                                    be16_to_cpu(next->rx_comp.rc));
2373                         /* free the entry */
2374                         next->rx_comp.first = 0;
2375                         dev_kfree_skb_any(rx_buff->skb);
2376                         remove_buff_from_pool(adapter, rx_buff);
2377                         continue;
2378                 } else if (!rx_buff->skb) {
2379                         /* free the entry */
2380                         next->rx_comp.first = 0;
2381                         remove_buff_from_pool(adapter, rx_buff);
2382                         continue;
2383                 }
2384
2385                 length = be32_to_cpu(next->rx_comp.len);
2386                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2387                 flags = next->rx_comp.flags;
2388                 skb = rx_buff->skb;
2389                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2390                                         length);
2391
2392                 /* VLAN Header has been stripped by the system firmware and
2393                  * needs to be inserted by the driver
2394                  */
2395                 if (adapter->rx_vlan_header_insertion &&
2396                     (flags & IBMVNIC_VLAN_STRIPPED))
2397                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2398                                                ntohs(next->rx_comp.vlan_tci));
2399
2400                 /* free the entry */
2401                 next->rx_comp.first = 0;
2402                 remove_buff_from_pool(adapter, rx_buff);
2403
2404                 skb_put(skb, length);
2405                 skb->protocol = eth_type_trans(skb, netdev);
2406                 skb_record_rx_queue(skb, scrq_num);
2407
2408                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2409                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2410                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2411                 }
2412
2413                 length = skb->len;
2414                 napi_gro_receive(napi, skb); /* send it up */
2415                 netdev->stats.rx_packets++;
2416                 netdev->stats.rx_bytes += length;
2417                 adapter->rx_stats_buffers[scrq_num].packets++;
2418                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2419                 frames_processed++;
2420         }
2421
2422         if (adapter->state != VNIC_CLOSING)
2423                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2424
2425         if (frames_processed < budget) {
2426                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2427                 napi_complete_done(napi, frames_processed);
2428                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2429                     napi_reschedule(napi)) {
2430                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2431                         goto restart_poll;
2432                 }
2433         }
2434         return frames_processed;
2435 }
2436
2437 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2438 {
2439         int rc, ret;
2440
2441         adapter->fallback.mtu = adapter->req_mtu;
2442         adapter->fallback.rx_queues = adapter->req_rx_queues;
2443         adapter->fallback.tx_queues = adapter->req_tx_queues;
2444         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2445         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2446
2447         reinit_completion(&adapter->reset_done);
2448         adapter->wait_for_reset = true;
2449         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2450
2451         if (rc) {
2452                 ret = rc;
2453                 goto out;
2454         }
2455         rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2456         if (rc) {
2457                 ret = -ENODEV;
2458                 goto out;
2459         }
2460
2461         ret = 0;
2462         if (adapter->reset_done_rc) {
2463                 ret = -EIO;
2464                 adapter->desired.mtu = adapter->fallback.mtu;
2465                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2466                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2467                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2468                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2469
2470                 reinit_completion(&adapter->reset_done);
2471                 adapter->wait_for_reset = true;
2472                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2473                 if (rc) {
2474                         ret = rc;
2475                         goto out;
2476                 }
2477                 rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2478                                                  60000);
2479                 if (rc) {
2480                         ret = -ENODEV;
2481                         goto out;
2482                 }
2483         }
2484 out:
2485         adapter->wait_for_reset = false;
2486
2487         return ret;
2488 }
2489
2490 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2491 {
2492         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2493
2494         adapter->desired.mtu = new_mtu + ETH_HLEN;
2495
2496         return wait_for_reset(adapter);
2497 }
2498
2499 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2500                                                 struct net_device *dev,
2501                                                 netdev_features_t features)
2502 {
2503         /* Some backing hardware adapters can not
2504          * handle packets with a MSS less than 224
2505          * or with only one segment.
2506          */
2507         if (skb_is_gso(skb)) {
2508                 if (skb_shinfo(skb)->gso_size < 224 ||
2509                     skb_shinfo(skb)->gso_segs == 1)
2510                         features &= ~NETIF_F_GSO_MASK;
2511         }
2512
2513         return features;
2514 }
2515
2516 static const struct net_device_ops ibmvnic_netdev_ops = {
2517         .ndo_open               = ibmvnic_open,
2518         .ndo_stop               = ibmvnic_close,
2519         .ndo_start_xmit         = ibmvnic_xmit,
2520         .ndo_set_rx_mode        = ibmvnic_set_multi,
2521         .ndo_set_mac_address    = ibmvnic_set_mac,
2522         .ndo_validate_addr      = eth_validate_addr,
2523         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2524         .ndo_change_mtu         = ibmvnic_change_mtu,
2525         .ndo_features_check     = ibmvnic_features_check,
2526 };
2527
2528 /* ethtool functions */
2529
2530 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2531                                       struct ethtool_link_ksettings *cmd)
2532 {
2533         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2534         int rc;
2535
2536         rc = send_query_phys_parms(adapter);
2537         if (rc) {
2538                 adapter->speed = SPEED_UNKNOWN;
2539                 adapter->duplex = DUPLEX_UNKNOWN;
2540         }
2541         cmd->base.speed = adapter->speed;
2542         cmd->base.duplex = adapter->duplex;
2543         cmd->base.port = PORT_FIBRE;
2544         cmd->base.phy_address = 0;
2545         cmd->base.autoneg = AUTONEG_ENABLE;
2546
2547         return 0;
2548 }
2549
2550 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2551                                 struct ethtool_drvinfo *info)
2552 {
2553         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2554
2555         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2556         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2557         strlcpy(info->fw_version, adapter->fw_version,
2558                 sizeof(info->fw_version));
2559 }
2560
2561 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2562 {
2563         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2564
2565         return adapter->msg_enable;
2566 }
2567
2568 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2569 {
2570         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2571
2572         adapter->msg_enable = data;
2573 }
2574
2575 static u32 ibmvnic_get_link(struct net_device *netdev)
2576 {
2577         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2578
2579         /* Don't need to send a query because we request a logical link up at
2580          * init and then we wait for link state indications
2581          */
2582         return adapter->logical_link_state;
2583 }
2584
2585 static void ibmvnic_get_ringparam(struct net_device *netdev,
2586                                   struct ethtool_ringparam *ring)
2587 {
2588         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2589
2590         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2591                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2592                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2593         } else {
2594                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2595                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2596         }
2597         ring->rx_mini_max_pending = 0;
2598         ring->rx_jumbo_max_pending = 0;
2599         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2600         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2601         ring->rx_mini_pending = 0;
2602         ring->rx_jumbo_pending = 0;
2603 }
2604
2605 static int ibmvnic_set_ringparam(struct net_device *netdev,
2606                                  struct ethtool_ringparam *ring)
2607 {
2608         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2609         int ret;
2610
2611         ret = 0;
2612         adapter->desired.rx_entries = ring->rx_pending;
2613         adapter->desired.tx_entries = ring->tx_pending;
2614
2615         ret = wait_for_reset(adapter);
2616
2617         if (!ret &&
2618             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2619              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2620                 netdev_info(netdev,
2621                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2622                             ring->rx_pending, ring->tx_pending,
2623                             adapter->req_rx_add_entries_per_subcrq,
2624                             adapter->req_tx_entries_per_subcrq);
2625         return ret;
2626 }
2627
2628 static void ibmvnic_get_channels(struct net_device *netdev,
2629                                  struct ethtool_channels *channels)
2630 {
2631         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2632
2633         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2634                 channels->max_rx = adapter->max_rx_queues;
2635                 channels->max_tx = adapter->max_tx_queues;
2636         } else {
2637                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2638                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2639         }
2640
2641         channels->max_other = 0;
2642         channels->max_combined = 0;
2643         channels->rx_count = adapter->req_rx_queues;
2644         channels->tx_count = adapter->req_tx_queues;
2645         channels->other_count = 0;
2646         channels->combined_count = 0;
2647 }
2648
2649 static int ibmvnic_set_channels(struct net_device *netdev,
2650                                 struct ethtool_channels *channels)
2651 {
2652         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2653         int ret;
2654
2655         ret = 0;
2656         adapter->desired.rx_queues = channels->rx_count;
2657         adapter->desired.tx_queues = channels->tx_count;
2658
2659         ret = wait_for_reset(adapter);
2660
2661         if (!ret &&
2662             (adapter->req_rx_queues != channels->rx_count ||
2663              adapter->req_tx_queues != channels->tx_count))
2664                 netdev_info(netdev,
2665                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2666                             channels->rx_count, channels->tx_count,
2667                             adapter->req_rx_queues, adapter->req_tx_queues);
2668         return ret;
2669
2670 }
2671
2672 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2673 {
2674         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2675         int i;
2676
2677         switch (stringset) {
2678         case ETH_SS_STATS:
2679                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2680                                 i++, data += ETH_GSTRING_LEN)
2681                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2682
2683                 for (i = 0; i < adapter->req_tx_queues; i++) {
2684                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2685                         data += ETH_GSTRING_LEN;
2686
2687                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2688                         data += ETH_GSTRING_LEN;
2689
2690                         snprintf(data, ETH_GSTRING_LEN,
2691                                  "tx%d_dropped_packets", i);
2692                         data += ETH_GSTRING_LEN;
2693                 }
2694
2695                 for (i = 0; i < adapter->req_rx_queues; i++) {
2696                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2697                         data += ETH_GSTRING_LEN;
2698
2699                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2700                         data += ETH_GSTRING_LEN;
2701
2702                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2703                         data += ETH_GSTRING_LEN;
2704                 }
2705                 break;
2706
2707         case ETH_SS_PRIV_FLAGS:
2708                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2709                         strcpy(data + i * ETH_GSTRING_LEN,
2710                                ibmvnic_priv_flags[i]);
2711                 break;
2712         default:
2713                 return;
2714         }
2715 }
2716
2717 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2718 {
2719         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2720
2721         switch (sset) {
2722         case ETH_SS_STATS:
2723                 return ARRAY_SIZE(ibmvnic_stats) +
2724                        adapter->req_tx_queues * NUM_TX_STATS +
2725                        adapter->req_rx_queues * NUM_RX_STATS;
2726         case ETH_SS_PRIV_FLAGS:
2727                 return ARRAY_SIZE(ibmvnic_priv_flags);
2728         default:
2729                 return -EOPNOTSUPP;
2730         }
2731 }
2732
2733 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2734                                       struct ethtool_stats *stats, u64 *data)
2735 {
2736         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2737         union ibmvnic_crq crq;
2738         int i, j;
2739         int rc;
2740
2741         memset(&crq, 0, sizeof(crq));
2742         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2743         crq.request_statistics.cmd = REQUEST_STATISTICS;
2744         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2745         crq.request_statistics.len =
2746             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2747
2748         /* Wait for data to be written */
2749         reinit_completion(&adapter->stats_done);
2750         rc = ibmvnic_send_crq(adapter, &crq);
2751         if (rc)
2752                 return;
2753         rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2754         if (rc)
2755                 return;
2756
2757         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2758                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2759                                                 ibmvnic_stats[i].offset));
2760
2761         for (j = 0; j < adapter->req_tx_queues; j++) {
2762                 data[i] = adapter->tx_stats_buffers[j].packets;
2763                 i++;
2764                 data[i] = adapter->tx_stats_buffers[j].bytes;
2765                 i++;
2766                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2767                 i++;
2768         }
2769
2770         for (j = 0; j < adapter->req_rx_queues; j++) {
2771                 data[i] = adapter->rx_stats_buffers[j].packets;
2772                 i++;
2773                 data[i] = adapter->rx_stats_buffers[j].bytes;
2774                 i++;
2775                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2776                 i++;
2777         }
2778 }
2779
2780 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2781 {
2782         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2783
2784         return adapter->priv_flags;
2785 }
2786
2787 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2788 {
2789         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2790         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2791
2792         if (which_maxes)
2793                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2794         else
2795                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2796
2797         return 0;
2798 }
2799 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2800         .get_drvinfo            = ibmvnic_get_drvinfo,
2801         .get_msglevel           = ibmvnic_get_msglevel,
2802         .set_msglevel           = ibmvnic_set_msglevel,
2803         .get_link               = ibmvnic_get_link,
2804         .get_ringparam          = ibmvnic_get_ringparam,
2805         .set_ringparam          = ibmvnic_set_ringparam,
2806         .get_channels           = ibmvnic_get_channels,
2807         .set_channels           = ibmvnic_set_channels,
2808         .get_strings            = ibmvnic_get_strings,
2809         .get_sset_count         = ibmvnic_get_sset_count,
2810         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2811         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2812         .get_priv_flags         = ibmvnic_get_priv_flags,
2813         .set_priv_flags         = ibmvnic_set_priv_flags,
2814 };
2815
2816 /* Routines for managing CRQs/sCRQs  */
2817
2818 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2819                                    struct ibmvnic_sub_crq_queue *scrq)
2820 {
2821         int rc;
2822
2823         if (scrq->irq) {
2824                 free_irq(scrq->irq, scrq);
2825                 irq_dispose_mapping(scrq->irq);
2826                 scrq->irq = 0;
2827         }
2828
2829         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2830         atomic_set(&scrq->used, 0);
2831         scrq->cur = 0;
2832
2833         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2834                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2835         return rc;
2836 }
2837
2838 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2839 {
2840         int i, rc;
2841
2842         for (i = 0; i < adapter->req_tx_queues; i++) {
2843                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2844                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2845                 if (rc)
2846                         return rc;
2847         }
2848
2849         for (i = 0; i < adapter->req_rx_queues; i++) {
2850                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2851                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2852                 if (rc)
2853                         return rc;
2854         }
2855
2856         return rc;
2857 }
2858
2859 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2860                                   struct ibmvnic_sub_crq_queue *scrq,
2861                                   bool do_h_free)
2862 {
2863         struct device *dev = &adapter->vdev->dev;
2864         long rc;
2865
2866         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2867
2868         if (do_h_free) {
2869                 /* Close the sub-crqs */
2870                 do {
2871                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2872                                                 adapter->vdev->unit_address,
2873                                                 scrq->crq_num);
2874                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2875
2876                 if (rc) {
2877                         netdev_err(adapter->netdev,
2878                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2879                                    scrq->crq_num, rc);
2880                 }
2881         }
2882
2883         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2884                          DMA_BIDIRECTIONAL);
2885         free_pages((unsigned long)scrq->msgs, 2);
2886         kfree(scrq);
2887 }
2888
2889 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2890                                                         *adapter)
2891 {
2892         struct device *dev = &adapter->vdev->dev;
2893         struct ibmvnic_sub_crq_queue *scrq;
2894         int rc;
2895
2896         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2897         if (!scrq)
2898                 return NULL;
2899
2900         scrq->msgs =
2901                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2902         if (!scrq->msgs) {
2903                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2904                 goto zero_page_failed;
2905         }
2906
2907         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2908                                          DMA_BIDIRECTIONAL);
2909         if (dma_mapping_error(dev, scrq->msg_token)) {
2910                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2911                 goto map_failed;
2912         }
2913
2914         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2915                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2916
2917         if (rc == H_RESOURCE)
2918                 rc = ibmvnic_reset_crq(adapter);
2919
2920         if (rc == H_CLOSED) {
2921                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2922         } else if (rc) {
2923                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2924                 goto reg_failed;
2925         }
2926
2927         scrq->adapter = adapter;
2928         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2929         spin_lock_init(&scrq->lock);
2930
2931         netdev_dbg(adapter->netdev,
2932                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2933                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2934
2935         return scrq;
2936
2937 reg_failed:
2938         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2939                          DMA_BIDIRECTIONAL);
2940 map_failed:
2941         free_pages((unsigned long)scrq->msgs, 2);
2942 zero_page_failed:
2943         kfree(scrq);
2944
2945         return NULL;
2946 }
2947
2948 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2949 {
2950         int i;
2951
2952         if (adapter->tx_scrq) {
2953                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2954                         if (!adapter->tx_scrq[i])
2955                                 continue;
2956
2957                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2958                                    i);
2959                         if (adapter->tx_scrq[i]->irq) {
2960                                 free_irq(adapter->tx_scrq[i]->irq,
2961                                          adapter->tx_scrq[i]);
2962                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2963                                 adapter->tx_scrq[i]->irq = 0;
2964                         }
2965
2966                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2967                                               do_h_free);
2968                 }
2969
2970                 kfree(adapter->tx_scrq);
2971                 adapter->tx_scrq = NULL;
2972                 adapter->num_active_tx_scrqs = 0;
2973         }
2974
2975         if (adapter->rx_scrq) {
2976                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2977                         if (!adapter->rx_scrq[i])
2978                                 continue;
2979
2980                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2981                                    i);
2982                         if (adapter->rx_scrq[i]->irq) {
2983                                 free_irq(adapter->rx_scrq[i]->irq,
2984                                          adapter->rx_scrq[i]);
2985                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2986                                 adapter->rx_scrq[i]->irq = 0;
2987                         }
2988
2989                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2990                                               do_h_free);
2991                 }
2992
2993                 kfree(adapter->rx_scrq);
2994                 adapter->rx_scrq = NULL;
2995                 adapter->num_active_rx_scrqs = 0;
2996         }
2997 }
2998
2999 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
3000                             struct ibmvnic_sub_crq_queue *scrq)
3001 {
3002         struct device *dev = &adapter->vdev->dev;
3003         unsigned long rc;
3004
3005         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3006                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3007         if (rc)
3008                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
3009                         scrq->hw_irq, rc);
3010         return rc;
3011 }
3012
3013 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3014                            struct ibmvnic_sub_crq_queue *scrq)
3015 {
3016         struct device *dev = &adapter->vdev->dev;
3017         unsigned long rc;
3018
3019         if (scrq->hw_irq > 0x100000000ULL) {
3020                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3021                 return 1;
3022         }
3023
3024         if (test_bit(0, &adapter->resetting) &&
3025             adapter->reset_reason == VNIC_RESET_MOBILITY) {
3026                 u64 val = (0xff000000) | scrq->hw_irq;
3027
3028                 rc = plpar_hcall_norets(H_EOI, val);
3029                 /* H_EOI would fail with rc = H_FUNCTION when running
3030                  * in XIVE mode which is expected, but not an error.
3031                  */
3032                 if (rc && (rc != H_FUNCTION))
3033                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3034                                 val, rc);
3035         }
3036
3037         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3038                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3039         if (rc)
3040                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3041                         scrq->hw_irq, rc);
3042         return rc;
3043 }
3044
3045 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3046                                struct ibmvnic_sub_crq_queue *scrq)
3047 {
3048         struct device *dev = &adapter->vdev->dev;
3049         struct ibmvnic_tx_pool *tx_pool;
3050         struct ibmvnic_tx_buff *txbuff;
3051         union sub_crq *next;
3052         int index;
3053         int i, j;
3054
3055 restart_loop:
3056         while (pending_scrq(adapter, scrq)) {
3057                 unsigned int pool = scrq->pool_index;
3058                 int num_entries = 0;
3059
3060                 next = ibmvnic_next_scrq(adapter, scrq);
3061                 for (i = 0; i < next->tx_comp.num_comps; i++) {
3062                         if (next->tx_comp.rcs[i]) {
3063                                 dev_err(dev, "tx error %x\n",
3064                                         next->tx_comp.rcs[i]);
3065                                 continue;
3066                         }
3067                         index = be32_to_cpu(next->tx_comp.correlators[i]);
3068                         if (index & IBMVNIC_TSO_POOL_MASK) {
3069                                 tx_pool = &adapter->tso_pool[pool];
3070                                 index &= ~IBMVNIC_TSO_POOL_MASK;
3071                         } else {
3072                                 tx_pool = &adapter->tx_pool[pool];
3073                         }
3074
3075                         txbuff = &tx_pool->tx_buff[index];
3076
3077                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3078                                 if (!txbuff->data_dma[j])
3079                                         continue;
3080
3081                                 txbuff->data_dma[j] = 0;
3082                         }
3083
3084                         if (txbuff->last_frag) {
3085                                 dev_kfree_skb_any(txbuff->skb);
3086                                 txbuff->skb = NULL;
3087                         }
3088
3089                         num_entries += txbuff->num_entries;
3090
3091                         tx_pool->free_map[tx_pool->producer_index] = index;
3092                         tx_pool->producer_index =
3093                                 (tx_pool->producer_index + 1) %
3094                                         tx_pool->num_buffers;
3095                 }
3096                 /* remove tx_comp scrq*/
3097                 next->tx_comp.first = 0;
3098
3099                 if (atomic_sub_return(num_entries, &scrq->used) <=
3100                     (adapter->req_tx_entries_per_subcrq / 2) &&
3101                     __netif_subqueue_stopped(adapter->netdev,
3102                                              scrq->pool_index)) {
3103                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3104                         netdev_dbg(adapter->netdev, "Started queue %d\n",
3105                                    scrq->pool_index);
3106                 }
3107         }
3108
3109         enable_scrq_irq(adapter, scrq);
3110
3111         if (pending_scrq(adapter, scrq)) {
3112                 disable_scrq_irq(adapter, scrq);
3113                 goto restart_loop;
3114         }
3115
3116         return 0;
3117 }
3118
3119 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3120 {
3121         struct ibmvnic_sub_crq_queue *scrq = instance;
3122         struct ibmvnic_adapter *adapter = scrq->adapter;
3123
3124         disable_scrq_irq(adapter, scrq);
3125         ibmvnic_complete_tx(adapter, scrq);
3126
3127         return IRQ_HANDLED;
3128 }
3129
3130 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3131 {
3132         struct ibmvnic_sub_crq_queue *scrq = instance;
3133         struct ibmvnic_adapter *adapter = scrq->adapter;
3134
3135         /* When booting a kdump kernel we can hit pending interrupts
3136          * prior to completing driver initialization.
3137          */
3138         if (unlikely(adapter->state != VNIC_OPEN))
3139                 return IRQ_NONE;
3140
3141         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3142
3143         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3144                 disable_scrq_irq(adapter, scrq);
3145                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
3146         }
3147
3148         return IRQ_HANDLED;
3149 }
3150
3151 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3152 {
3153         struct device *dev = &adapter->vdev->dev;
3154         struct ibmvnic_sub_crq_queue *scrq;
3155         int i = 0, j = 0;
3156         int rc = 0;
3157
3158         for (i = 0; i < adapter->req_tx_queues; i++) {
3159                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3160                            i);
3161                 scrq = adapter->tx_scrq[i];
3162                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3163
3164                 if (!scrq->irq) {
3165                         rc = -EINVAL;
3166                         dev_err(dev, "Error mapping irq\n");
3167                         goto req_tx_irq_failed;
3168                 }
3169
3170                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3171                          adapter->vdev->unit_address, i);
3172                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3173                                  0, scrq->name, scrq);
3174
3175                 if (rc) {
3176                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3177                                 scrq->irq, rc);
3178                         irq_dispose_mapping(scrq->irq);
3179                         goto req_tx_irq_failed;
3180                 }
3181         }
3182
3183         for (i = 0; i < adapter->req_rx_queues; i++) {
3184                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3185                            i);
3186                 scrq = adapter->rx_scrq[i];
3187                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3188                 if (!scrq->irq) {
3189                         rc = -EINVAL;
3190                         dev_err(dev, "Error mapping irq\n");
3191                         goto req_rx_irq_failed;
3192                 }
3193                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3194                          adapter->vdev->unit_address, i);
3195                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3196                                  0, scrq->name, scrq);
3197                 if (rc) {
3198                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3199                                 scrq->irq, rc);
3200                         irq_dispose_mapping(scrq->irq);
3201                         goto req_rx_irq_failed;
3202                 }
3203         }
3204         return rc;
3205
3206 req_rx_irq_failed:
3207         for (j = 0; j < i; j++) {
3208                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3209                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3210         }
3211         i = adapter->req_tx_queues;
3212 req_tx_irq_failed:
3213         for (j = 0; j < i; j++) {
3214                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3215                 irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3216         }
3217         release_sub_crqs(adapter, 1);
3218         return rc;
3219 }
3220
3221 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3222 {
3223         struct device *dev = &adapter->vdev->dev;
3224         struct ibmvnic_sub_crq_queue **allqueues;
3225         int registered_queues = 0;
3226         int total_queues;
3227         int more = 0;
3228         int i;
3229
3230         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3231
3232         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3233         if (!allqueues)
3234                 return -1;
3235
3236         for (i = 0; i < total_queues; i++) {
3237                 allqueues[i] = init_sub_crq_queue(adapter);
3238                 if (!allqueues[i]) {
3239                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3240                         break;
3241                 }
3242                 registered_queues++;
3243         }
3244
3245         /* Make sure we were able to register the minimum number of queues */
3246         if (registered_queues <
3247             adapter->min_tx_queues + adapter->min_rx_queues) {
3248                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3249                 goto tx_failed;
3250         }
3251
3252         /* Distribute the failed allocated queues*/
3253         for (i = 0; i < total_queues - registered_queues + more ; i++) {
3254                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3255                 switch (i % 3) {
3256                 case 0:
3257                         if (adapter->req_rx_queues > adapter->min_rx_queues)
3258                                 adapter->req_rx_queues--;
3259                         else
3260                                 more++;
3261                         break;
3262                 case 1:
3263                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3264                                 adapter->req_tx_queues--;
3265                         else
3266                                 more++;
3267                         break;
3268                 }
3269         }
3270
3271         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3272                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3273         if (!adapter->tx_scrq)
3274                 goto tx_failed;
3275
3276         for (i = 0; i < adapter->req_tx_queues; i++) {
3277                 adapter->tx_scrq[i] = allqueues[i];
3278                 adapter->tx_scrq[i]->pool_index = i;
3279                 adapter->num_active_tx_scrqs++;
3280         }
3281
3282         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3283                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3284         if (!adapter->rx_scrq)
3285                 goto rx_failed;
3286
3287         for (i = 0; i < adapter->req_rx_queues; i++) {
3288                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3289                 adapter->rx_scrq[i]->scrq_num = i;
3290                 adapter->num_active_rx_scrqs++;
3291         }
3292
3293         kfree(allqueues);
3294         return 0;
3295
3296 rx_failed:
3297         kfree(adapter->tx_scrq);
3298         adapter->tx_scrq = NULL;
3299 tx_failed:
3300         for (i = 0; i < registered_queues; i++)
3301                 release_sub_crq_queue(adapter, allqueues[i], 1);
3302         kfree(allqueues);
3303         return -1;
3304 }
3305
3306 static void send_request_cap(struct ibmvnic_adapter *adapter, int retry)
3307 {
3308         struct device *dev = &adapter->vdev->dev;
3309         union ibmvnic_crq crq;
3310         int max_entries;
3311
3312         if (!retry) {
3313                 /* Sub-CRQ entries are 32 byte long */
3314                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3315
3316                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3317                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3318                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3319                         return;
3320                 }
3321
3322                 if (adapter->desired.mtu)
3323                         adapter->req_mtu = adapter->desired.mtu;
3324                 else
3325                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3326
3327                 if (!adapter->desired.tx_entries)
3328                         adapter->desired.tx_entries =
3329                                         adapter->max_tx_entries_per_subcrq;
3330                 if (!adapter->desired.rx_entries)
3331                         adapter->desired.rx_entries =
3332                                         adapter->max_rx_add_entries_per_subcrq;
3333
3334                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3335                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3336
3337                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3338                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3339                         adapter->desired.tx_entries = max_entries;
3340                 }
3341
3342                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3343                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3344                         adapter->desired.rx_entries = max_entries;
3345                 }
3346
3347                 if (adapter->desired.tx_entries)
3348                         adapter->req_tx_entries_per_subcrq =
3349                                         adapter->desired.tx_entries;
3350                 else
3351                         adapter->req_tx_entries_per_subcrq =
3352                                         adapter->max_tx_entries_per_subcrq;
3353
3354                 if (adapter->desired.rx_entries)
3355                         adapter->req_rx_add_entries_per_subcrq =
3356                                         adapter->desired.rx_entries;
3357                 else
3358                         adapter->req_rx_add_entries_per_subcrq =
3359                                         adapter->max_rx_add_entries_per_subcrq;
3360
3361                 if (adapter->desired.tx_queues)
3362                         adapter->req_tx_queues =
3363                                         adapter->desired.tx_queues;
3364                 else
3365                         adapter->req_tx_queues =
3366                                         adapter->opt_tx_comp_sub_queues;
3367
3368                 if (adapter->desired.rx_queues)
3369                         adapter->req_rx_queues =
3370                                         adapter->desired.rx_queues;
3371                 else
3372                         adapter->req_rx_queues =
3373                                         adapter->opt_rx_comp_queues;
3374
3375                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3376         }
3377
3378         memset(&crq, 0, sizeof(crq));
3379         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3380         crq.request_capability.cmd = REQUEST_CAPABILITY;
3381
3382         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3383         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3384         atomic_inc(&adapter->running_cap_crqs);
3385         ibmvnic_send_crq(adapter, &crq);
3386
3387         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3388         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3389         atomic_inc(&adapter->running_cap_crqs);
3390         ibmvnic_send_crq(adapter, &crq);
3391
3392         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3393         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3394         atomic_inc(&adapter->running_cap_crqs);
3395         ibmvnic_send_crq(adapter, &crq);
3396
3397         crq.request_capability.capability =
3398             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3399         crq.request_capability.number =
3400             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3401         atomic_inc(&adapter->running_cap_crqs);
3402         ibmvnic_send_crq(adapter, &crq);
3403
3404         crq.request_capability.capability =
3405             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3406         crq.request_capability.number =
3407             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3408         atomic_inc(&adapter->running_cap_crqs);
3409         ibmvnic_send_crq(adapter, &crq);
3410
3411         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3412         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3413         atomic_inc(&adapter->running_cap_crqs);
3414         ibmvnic_send_crq(adapter, &crq);
3415
3416         if (adapter->netdev->flags & IFF_PROMISC) {
3417                 if (adapter->promisc_supported) {
3418                         crq.request_capability.capability =
3419                             cpu_to_be16(PROMISC_REQUESTED);
3420                         crq.request_capability.number = cpu_to_be64(1);
3421                         atomic_inc(&adapter->running_cap_crqs);
3422                         ibmvnic_send_crq(adapter, &crq);
3423                 }
3424         } else {
3425                 crq.request_capability.capability =
3426                     cpu_to_be16(PROMISC_REQUESTED);
3427                 crq.request_capability.number = cpu_to_be64(0);
3428                 atomic_inc(&adapter->running_cap_crqs);
3429                 ibmvnic_send_crq(adapter, &crq);
3430         }
3431 }
3432
3433 static int pending_scrq(struct ibmvnic_adapter *adapter,
3434                         struct ibmvnic_sub_crq_queue *scrq)
3435 {
3436         union sub_crq *entry = &scrq->msgs[scrq->cur];
3437
3438         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3439                 return 1;
3440         else
3441                 return 0;
3442 }
3443
3444 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3445                                         struct ibmvnic_sub_crq_queue *scrq)
3446 {
3447         union sub_crq *entry;
3448         unsigned long flags;
3449
3450         spin_lock_irqsave(&scrq->lock, flags);
3451         entry = &scrq->msgs[scrq->cur];
3452         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3453                 if (++scrq->cur == scrq->size)
3454                         scrq->cur = 0;
3455         } else {
3456                 entry = NULL;
3457         }
3458         spin_unlock_irqrestore(&scrq->lock, flags);
3459
3460         return entry;
3461 }
3462
3463 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3464 {
3465         struct ibmvnic_crq_queue *queue = &adapter->crq;
3466         union ibmvnic_crq *crq;
3467
3468         crq = &queue->msgs[queue->cur];
3469         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3470                 if (++queue->cur == queue->size)
3471                         queue->cur = 0;
3472         } else {
3473                 crq = NULL;
3474         }
3475
3476         return crq;
3477 }
3478
3479 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3480 {
3481         switch (rc) {
3482         case H_PARAMETER:
3483                 dev_warn_ratelimited(dev,
3484                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3485                                      func, rc);
3486                 break;
3487         case H_CLOSED:
3488                 dev_warn_ratelimited(dev,
3489                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3490                                      func, rc);
3491                 break;
3492         default:
3493                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3494                 break;
3495         }
3496 }
3497
3498 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3499                        union sub_crq *sub_crq)
3500 {
3501         unsigned int ua = adapter->vdev->unit_address;
3502         struct device *dev = &adapter->vdev->dev;
3503         u64 *u64_crq = (u64 *)sub_crq;
3504         int rc;
3505
3506         netdev_dbg(adapter->netdev,
3507                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3508                    (unsigned long int)cpu_to_be64(remote_handle),
3509                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3510                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3511                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3512                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3513
3514         /* Make sure the hypervisor sees the complete request */
3515         mb();
3516
3517         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3518                                 cpu_to_be64(remote_handle),
3519                                 cpu_to_be64(u64_crq[0]),
3520                                 cpu_to_be64(u64_crq[1]),
3521                                 cpu_to_be64(u64_crq[2]),
3522                                 cpu_to_be64(u64_crq[3]));
3523
3524         if (rc)
3525                 print_subcrq_error(dev, rc, __func__);
3526
3527         return rc;
3528 }
3529
3530 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3531                                 u64 remote_handle, u64 ioba, u64 num_entries)
3532 {
3533         unsigned int ua = adapter->vdev->unit_address;
3534         struct device *dev = &adapter->vdev->dev;
3535         int rc;
3536
3537         /* Make sure the hypervisor sees the complete request */
3538         mb();
3539         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3540                                 cpu_to_be64(remote_handle),
3541                                 ioba, num_entries);
3542
3543         if (rc)
3544                 print_subcrq_error(dev, rc, __func__);
3545
3546         return rc;
3547 }
3548
3549 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3550                             union ibmvnic_crq *crq)
3551 {
3552         unsigned int ua = adapter->vdev->unit_address;
3553         struct device *dev = &adapter->vdev->dev;
3554         u64 *u64_crq = (u64 *)crq;
3555         int rc;
3556
3557         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3558                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3559                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3560
3561         if (!adapter->crq.active &&
3562             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3563                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3564                 return -EINVAL;
3565         }
3566
3567         /* Make sure the hypervisor sees the complete request */
3568         mb();
3569
3570         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3571                                 cpu_to_be64(u64_crq[0]),
3572                                 cpu_to_be64(u64_crq[1]));
3573
3574         if (rc) {
3575                 if (rc == H_CLOSED) {
3576                         dev_warn(dev, "CRQ Queue closed\n");
3577                         /* do not reset, report the fail, wait for passive init from server */
3578                 }
3579
3580                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3581         }
3582
3583         return rc;
3584 }
3585
3586 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3587 {
3588         struct device *dev = &adapter->vdev->dev;
3589         union ibmvnic_crq crq;
3590         int retries = 100;
3591         int rc;
3592
3593         memset(&crq, 0, sizeof(crq));
3594         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3595         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3596         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3597
3598         do {
3599                 rc = ibmvnic_send_crq(adapter, &crq);
3600                 if (rc != H_CLOSED)
3601                         break;
3602                 retries--;
3603                 msleep(50);
3604
3605         } while (retries > 0);
3606
3607         if (rc) {
3608                 dev_err(dev, "Failed to send init request, rc = %d\n", rc);
3609                 return rc;
3610         }
3611
3612         return 0;
3613 }
3614
3615 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3616 {
3617         union ibmvnic_crq crq;
3618
3619         memset(&crq, 0, sizeof(crq));
3620         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3621         crq.version_exchange.cmd = VERSION_EXCHANGE;
3622         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3623
3624         return ibmvnic_send_crq(adapter, &crq);
3625 }
3626
3627 struct vnic_login_client_data {
3628         u8      type;
3629         __be16  len;
3630         char    name[];
3631 } __packed;
3632
3633 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3634 {
3635         int len;
3636
3637         /* Calculate the amount of buffer space needed for the
3638          * vnic client data in the login buffer. There are four entries,
3639          * OS name, LPAR name, device name, and a null last entry.
3640          */
3641         len = 4 * sizeof(struct vnic_login_client_data);
3642         len += 6; /* "Linux" plus NULL */
3643         len += strlen(utsname()->nodename) + 1;
3644         len += strlen(adapter->netdev->name) + 1;
3645
3646         return len;
3647 }
3648
3649 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3650                                  struct vnic_login_client_data *vlcd)
3651 {
3652         const char *os_name = "Linux";
3653         int len;
3654
3655         /* Type 1 - LPAR OS */
3656         vlcd->type = 1;
3657         len = strlen(os_name) + 1;
3658         vlcd->len = cpu_to_be16(len);
3659         strncpy(vlcd->name, os_name, len);
3660         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3661
3662         /* Type 2 - LPAR name */
3663         vlcd->type = 2;
3664         len = strlen(utsname()->nodename) + 1;
3665         vlcd->len = cpu_to_be16(len);
3666         strncpy(vlcd->name, utsname()->nodename, len);
3667         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3668
3669         /* Type 3 - device name */
3670         vlcd->type = 3;
3671         len = strlen(adapter->netdev->name) + 1;
3672         vlcd->len = cpu_to_be16(len);
3673         strncpy(vlcd->name, adapter->netdev->name, len);
3674 }
3675
3676 static int send_login(struct ibmvnic_adapter *adapter)
3677 {
3678         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3679         struct ibmvnic_login_buffer *login_buffer;
3680         struct device *dev = &adapter->vdev->dev;
3681         dma_addr_t rsp_buffer_token;
3682         dma_addr_t buffer_token;
3683         size_t rsp_buffer_size;
3684         union ibmvnic_crq crq;
3685         size_t buffer_size;
3686         __be64 *tx_list_p;
3687         __be64 *rx_list_p;
3688         int client_data_len;
3689         struct vnic_login_client_data *vlcd;
3690         int i;
3691
3692         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3693                 netdev_err(adapter->netdev,
3694                            "RX or TX queues are not allocated, device login failed\n");
3695                 return -1;
3696         }
3697
3698         release_login_rsp_buffer(adapter);
3699         client_data_len = vnic_client_data_len(adapter);
3700
3701         buffer_size =
3702             sizeof(struct ibmvnic_login_buffer) +
3703             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3704             client_data_len;
3705
3706         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3707         if (!login_buffer)
3708                 goto buf_alloc_failed;
3709
3710         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3711                                       DMA_TO_DEVICE);
3712         if (dma_mapping_error(dev, buffer_token)) {
3713                 dev_err(dev, "Couldn't map login buffer\n");
3714                 goto buf_map_failed;
3715         }
3716
3717         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3718                           sizeof(u64) * adapter->req_tx_queues +
3719                           sizeof(u64) * adapter->req_rx_queues +
3720                           sizeof(u64) * adapter->req_rx_queues +
3721                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3722
3723         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3724         if (!login_rsp_buffer)
3725                 goto buf_rsp_alloc_failed;
3726
3727         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3728                                           rsp_buffer_size, DMA_FROM_DEVICE);
3729         if (dma_mapping_error(dev, rsp_buffer_token)) {
3730                 dev_err(dev, "Couldn't map login rsp buffer\n");
3731                 goto buf_rsp_map_failed;
3732         }
3733
3734         adapter->login_buf = login_buffer;
3735         adapter->login_buf_token = buffer_token;
3736         adapter->login_buf_sz = buffer_size;
3737         adapter->login_rsp_buf = login_rsp_buffer;
3738         adapter->login_rsp_buf_token = rsp_buffer_token;
3739         adapter->login_rsp_buf_sz = rsp_buffer_size;
3740
3741         login_buffer->len = cpu_to_be32(buffer_size);
3742         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3743         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3744         login_buffer->off_txcomp_subcrqs =
3745             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3746         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3747         login_buffer->off_rxcomp_subcrqs =
3748             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3749                         sizeof(u64) * adapter->req_tx_queues);
3750         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3751         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3752
3753         tx_list_p = (__be64 *)((char *)login_buffer +
3754                                       sizeof(struct ibmvnic_login_buffer));
3755         rx_list_p = (__be64 *)((char *)login_buffer +
3756                                       sizeof(struct ibmvnic_login_buffer) +
3757                                       sizeof(u64) * adapter->req_tx_queues);
3758
3759         for (i = 0; i < adapter->req_tx_queues; i++) {
3760                 if (adapter->tx_scrq[i]) {
3761                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3762                                                    crq_num);
3763                 }
3764         }
3765
3766         for (i = 0; i < adapter->req_rx_queues; i++) {
3767                 if (adapter->rx_scrq[i]) {
3768                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3769                                                    crq_num);
3770                 }
3771         }
3772
3773         /* Insert vNIC login client data */
3774         vlcd = (struct vnic_login_client_data *)
3775                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3776         login_buffer->client_data_offset =
3777                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3778         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3779
3780         vnic_add_client_data(adapter, vlcd);
3781
3782         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3783         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3784                 netdev_dbg(adapter->netdev, "%016lx\n",
3785                            ((unsigned long int *)(adapter->login_buf))[i]);
3786         }
3787
3788         memset(&crq, 0, sizeof(crq));
3789         crq.login.first = IBMVNIC_CRQ_CMD;
3790         crq.login.cmd = LOGIN;
3791         crq.login.ioba = cpu_to_be32(buffer_token);
3792         crq.login.len = cpu_to_be32(buffer_size);
3793         ibmvnic_send_crq(adapter, &crq);
3794
3795         return 0;
3796
3797 buf_rsp_map_failed:
3798         kfree(login_rsp_buffer);
3799 buf_rsp_alloc_failed:
3800         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3801 buf_map_failed:
3802         kfree(login_buffer);
3803 buf_alloc_failed:
3804         return -1;
3805 }
3806
3807 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3808                             u32 len, u8 map_id)
3809 {
3810         union ibmvnic_crq crq;
3811
3812         memset(&crq, 0, sizeof(crq));
3813         crq.request_map.first = IBMVNIC_CRQ_CMD;
3814         crq.request_map.cmd = REQUEST_MAP;
3815         crq.request_map.map_id = map_id;
3816         crq.request_map.ioba = cpu_to_be32(addr);
3817         crq.request_map.len = cpu_to_be32(len);
3818         return ibmvnic_send_crq(adapter, &crq);
3819 }
3820
3821 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3822 {
3823         union ibmvnic_crq crq;
3824
3825         memset(&crq, 0, sizeof(crq));
3826         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3827         crq.request_unmap.cmd = REQUEST_UNMAP;
3828         crq.request_unmap.map_id = map_id;
3829         return ibmvnic_send_crq(adapter, &crq);
3830 }
3831
3832 static void send_query_map(struct ibmvnic_adapter *adapter)
3833 {
3834         union ibmvnic_crq crq;
3835
3836         memset(&crq, 0, sizeof(crq));
3837         crq.query_map.first = IBMVNIC_CRQ_CMD;
3838         crq.query_map.cmd = QUERY_MAP;
3839         ibmvnic_send_crq(adapter, &crq);
3840 }
3841
3842 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3843 static void send_query_cap(struct ibmvnic_adapter *adapter)
3844 {
3845         union ibmvnic_crq crq;
3846
3847         atomic_set(&adapter->running_cap_crqs, 0);
3848         memset(&crq, 0, sizeof(crq));
3849         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3850         crq.query_capability.cmd = QUERY_CAPABILITY;
3851
3852         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3853         atomic_inc(&adapter->running_cap_crqs);
3854         ibmvnic_send_crq(adapter, &crq);
3855
3856         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3857         atomic_inc(&adapter->running_cap_crqs);
3858         ibmvnic_send_crq(adapter, &crq);
3859
3860         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3861         atomic_inc(&adapter->running_cap_crqs);
3862         ibmvnic_send_crq(adapter, &crq);
3863
3864         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3865         atomic_inc(&adapter->running_cap_crqs);
3866         ibmvnic_send_crq(adapter, &crq);
3867
3868         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3869         atomic_inc(&adapter->running_cap_crqs);
3870         ibmvnic_send_crq(adapter, &crq);
3871
3872         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3873         atomic_inc(&adapter->running_cap_crqs);
3874         ibmvnic_send_crq(adapter, &crq);
3875
3876         crq.query_capability.capability =
3877             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3878         atomic_inc(&adapter->running_cap_crqs);
3879         ibmvnic_send_crq(adapter, &crq);
3880
3881         crq.query_capability.capability =
3882             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3883         atomic_inc(&adapter->running_cap_crqs);
3884         ibmvnic_send_crq(adapter, &crq);
3885
3886         crq.query_capability.capability =
3887             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3888         atomic_inc(&adapter->running_cap_crqs);
3889         ibmvnic_send_crq(adapter, &crq);
3890
3891         crq.query_capability.capability =
3892             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3893         atomic_inc(&adapter->running_cap_crqs);
3894         ibmvnic_send_crq(adapter, &crq);
3895
3896         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3897         atomic_inc(&adapter->running_cap_crqs);
3898         ibmvnic_send_crq(adapter, &crq);
3899
3900         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3901         atomic_inc(&adapter->running_cap_crqs);
3902         ibmvnic_send_crq(adapter, &crq);
3903
3904         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3905         atomic_inc(&adapter->running_cap_crqs);
3906         ibmvnic_send_crq(adapter, &crq);
3907
3908         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3909         atomic_inc(&adapter->running_cap_crqs);
3910         ibmvnic_send_crq(adapter, &crq);
3911
3912         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3913         atomic_inc(&adapter->running_cap_crqs);
3914         ibmvnic_send_crq(adapter, &crq);
3915
3916         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3917         atomic_inc(&adapter->running_cap_crqs);
3918         ibmvnic_send_crq(adapter, &crq);
3919
3920         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3921         atomic_inc(&adapter->running_cap_crqs);
3922         ibmvnic_send_crq(adapter, &crq);
3923
3924         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3925         atomic_inc(&adapter->running_cap_crqs);
3926         ibmvnic_send_crq(adapter, &crq);
3927
3928         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3929         atomic_inc(&adapter->running_cap_crqs);
3930         ibmvnic_send_crq(adapter, &crq);
3931
3932         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3933         atomic_inc(&adapter->running_cap_crqs);
3934         ibmvnic_send_crq(adapter, &crq);
3935
3936         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3937         atomic_inc(&adapter->running_cap_crqs);
3938         ibmvnic_send_crq(adapter, &crq);
3939
3940         crq.query_capability.capability =
3941                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3942         atomic_inc(&adapter->running_cap_crqs);
3943         ibmvnic_send_crq(adapter, &crq);
3944
3945         crq.query_capability.capability =
3946                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3947         atomic_inc(&adapter->running_cap_crqs);
3948         ibmvnic_send_crq(adapter, &crq);
3949
3950         crq.query_capability.capability =
3951                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3952         atomic_inc(&adapter->running_cap_crqs);
3953         ibmvnic_send_crq(adapter, &crq);
3954
3955         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3956         atomic_inc(&adapter->running_cap_crqs);
3957         ibmvnic_send_crq(adapter, &crq);
3958 }
3959
3960 static void send_query_ip_offload(struct ibmvnic_adapter *adapter)
3961 {
3962         int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3963         struct device *dev = &adapter->vdev->dev;
3964         union ibmvnic_crq crq;
3965
3966         adapter->ip_offload_tok =
3967                 dma_map_single(dev,
3968                                &adapter->ip_offload_buf,
3969                                buf_sz,
3970                                DMA_FROM_DEVICE);
3971
3972         if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3973                 if (!firmware_has_feature(FW_FEATURE_CMO))
3974                         dev_err(dev, "Couldn't map offload buffer\n");
3975                 return;
3976         }
3977
3978         memset(&crq, 0, sizeof(crq));
3979         crq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3980         crq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3981         crq.query_ip_offload.len = cpu_to_be32(buf_sz);
3982         crq.query_ip_offload.ioba =
3983             cpu_to_be32(adapter->ip_offload_tok);
3984
3985         ibmvnic_send_crq(adapter, &crq);
3986 }
3987
3988 static void send_control_ip_offload(struct ibmvnic_adapter *adapter)
3989 {
3990         struct ibmvnic_control_ip_offload_buffer *ctrl_buf = &adapter->ip_offload_ctrl;
3991         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3992         struct device *dev = &adapter->vdev->dev;
3993         netdev_features_t old_hw_features = 0;
3994         union ibmvnic_crq crq;
3995
3996         adapter->ip_offload_ctrl_tok =
3997                 dma_map_single(dev,
3998                                ctrl_buf,
3999                                sizeof(adapter->ip_offload_ctrl),
4000                                DMA_TO_DEVICE);
4001
4002         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4003                 dev_err(dev, "Couldn't map ip offload control buffer\n");
4004                 return;
4005         }
4006
4007         ctrl_buf->len = cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4008         ctrl_buf->version = cpu_to_be32(INITIAL_VERSION_IOB);
4009         ctrl_buf->ipv4_chksum = buf->ipv4_chksum;
4010         ctrl_buf->ipv6_chksum = buf->ipv6_chksum;
4011         ctrl_buf->tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4012         ctrl_buf->udp_ipv4_chksum = buf->udp_ipv4_chksum;
4013         ctrl_buf->tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4014         ctrl_buf->udp_ipv6_chksum = buf->udp_ipv6_chksum;
4015         ctrl_buf->large_tx_ipv4 = buf->large_tx_ipv4;
4016         ctrl_buf->large_tx_ipv6 = buf->large_tx_ipv6;
4017
4018         /* large_rx disabled for now, additional features needed */
4019         ctrl_buf->large_rx_ipv4 = 0;
4020         ctrl_buf->large_rx_ipv6 = 0;
4021
4022         if (adapter->state != VNIC_PROBING) {
4023                 old_hw_features = adapter->netdev->hw_features;
4024                 adapter->netdev->hw_features = 0;
4025         }
4026
4027         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4028
4029         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4030                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4031
4032         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4033                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4034
4035         if ((adapter->netdev->features &
4036             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4037                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4038
4039         if (buf->large_tx_ipv4)
4040                 adapter->netdev->hw_features |= NETIF_F_TSO;
4041         if (buf->large_tx_ipv6)
4042                 adapter->netdev->hw_features |= NETIF_F_TSO6;
4043
4044         if (adapter->state == VNIC_PROBING) {
4045                 adapter->netdev->features |= adapter->netdev->hw_features;
4046         } else if (old_hw_features != adapter->netdev->hw_features) {
4047                 netdev_features_t tmp = 0;
4048
4049                 /* disable features no longer supported */
4050                 adapter->netdev->features &= adapter->netdev->hw_features;
4051                 /* turn on features now supported if previously enabled */
4052                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4053                         adapter->netdev->hw_features;
4054                 adapter->netdev->features |=
4055                                 tmp & adapter->netdev->wanted_features;
4056         }
4057
4058         memset(&crq, 0, sizeof(crq));
4059         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4060         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4061         crq.control_ip_offload.len =
4062             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4063         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4064         ibmvnic_send_crq(adapter, &crq);
4065 }
4066
4067 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
4068                                 struct ibmvnic_adapter *adapter)
4069 {
4070         struct device *dev = &adapter->vdev->dev;
4071
4072         if (crq->get_vpd_size_rsp.rc.code) {
4073                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
4074                         crq->get_vpd_size_rsp.rc.code);
4075                 complete(&adapter->fw_done);
4076                 return;
4077         }
4078
4079         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
4080         complete(&adapter->fw_done);
4081 }
4082
4083 static void handle_vpd_rsp(union ibmvnic_crq *crq,
4084                            struct ibmvnic_adapter *adapter)
4085 {
4086         struct device *dev = &adapter->vdev->dev;
4087         unsigned char *substr = NULL;
4088         u8 fw_level_len = 0;
4089
4090         memset(adapter->fw_version, 0, 32);
4091
4092         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
4093                          DMA_FROM_DEVICE);
4094
4095         if (crq->get_vpd_rsp.rc.code) {
4096                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
4097                         crq->get_vpd_rsp.rc.code);
4098                 goto complete;
4099         }
4100
4101         /* get the position of the firmware version info
4102          * located after the ASCII 'RM' substring in the buffer
4103          */
4104         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
4105         if (!substr) {
4106                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
4107                 goto complete;
4108         }
4109
4110         /* get length of firmware level ASCII substring */
4111         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
4112                 fw_level_len = *(substr + 2);
4113         } else {
4114                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
4115                 goto complete;
4116         }
4117
4118         /* copy firmware version string from vpd into adapter */
4119         if ((substr + 3 + fw_level_len) <
4120             (adapter->vpd->buff + adapter->vpd->len)) {
4121                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
4122         } else {
4123                 dev_info(dev, "FW substr extrapolated VPD buff\n");
4124         }
4125
4126 complete:
4127         if (adapter->fw_version[0] == '\0')
4128                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
4129         complete(&adapter->fw_done);
4130 }
4131
4132 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4133 {
4134         struct device *dev = &adapter->vdev->dev;
4135         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4136         int i;
4137
4138         dma_unmap_single(dev, adapter->ip_offload_tok,
4139                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4140
4141         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4142         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4143                 netdev_dbg(adapter->netdev, "%016lx\n",
4144                            ((unsigned long int *)(buf))[i]);
4145
4146         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4147         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4148         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4149                    buf->tcp_ipv4_chksum);
4150         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4151                    buf->tcp_ipv6_chksum);
4152         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4153                    buf->udp_ipv4_chksum);
4154         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4155                    buf->udp_ipv6_chksum);
4156         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4157                    buf->large_tx_ipv4);
4158         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4159                    buf->large_tx_ipv6);
4160         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4161                    buf->large_rx_ipv4);
4162         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4163                    buf->large_rx_ipv6);
4164         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4165                    buf->max_ipv4_header_size);
4166         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4167                    buf->max_ipv6_header_size);
4168         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4169                    buf->max_tcp_header_size);
4170         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4171                    buf->max_udp_header_size);
4172         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4173                    buf->max_large_tx_size);
4174         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4175                    buf->max_large_rx_size);
4176         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4177                    buf->ipv6_extension_header);
4178         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4179                    buf->tcp_pseudosum_req);
4180         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4181                    buf->num_ipv6_ext_headers);
4182         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4183                    buf->off_ipv6_ext_headers);
4184
4185         send_control_ip_offload(adapter);
4186 }
4187
4188 static const char *ibmvnic_fw_err_cause(u16 cause)
4189 {
4190         switch (cause) {
4191         case ADAPTER_PROBLEM:
4192                 return "adapter problem";
4193         case BUS_PROBLEM:
4194                 return "bus problem";
4195         case FW_PROBLEM:
4196                 return "firmware problem";
4197         case DD_PROBLEM:
4198                 return "device driver problem";
4199         case EEH_RECOVERY:
4200                 return "EEH recovery";
4201         case FW_UPDATED:
4202                 return "firmware updated";
4203         case LOW_MEMORY:
4204                 return "low Memory";
4205         default:
4206                 return "unknown";
4207         }
4208 }
4209
4210 static void handle_error_indication(union ibmvnic_crq *crq,
4211                                     struct ibmvnic_adapter *adapter)
4212 {
4213         struct device *dev = &adapter->vdev->dev;
4214         u16 cause;
4215
4216         cause = be16_to_cpu(crq->error_indication.error_cause);
4217
4218         dev_warn_ratelimited(dev,
4219                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
4220                              crq->error_indication.flags
4221                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4222                              ibmvnic_fw_err_cause(cause));
4223
4224         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4225                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4226         else
4227                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4228 }
4229
4230 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4231                                  struct ibmvnic_adapter *adapter)
4232 {
4233         struct net_device *netdev = adapter->netdev;
4234         struct device *dev = &adapter->vdev->dev;
4235         long rc;
4236
4237         rc = crq->change_mac_addr_rsp.rc.code;
4238         if (rc) {
4239                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4240                 goto out;
4241         }
4242         /* crq->change_mac_addr.mac_addr is the requested one
4243          * crq->change_mac_addr_rsp.mac_addr is the returned valid one.
4244          */
4245         ether_addr_copy(netdev->dev_addr,
4246                         &crq->change_mac_addr_rsp.mac_addr[0]);
4247         ether_addr_copy(adapter->mac_addr,
4248                         &crq->change_mac_addr_rsp.mac_addr[0]);
4249 out:
4250         complete(&adapter->fw_done);
4251         return rc;
4252 }
4253
4254 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4255                                    struct ibmvnic_adapter *adapter)
4256 {
4257         struct device *dev = &adapter->vdev->dev;
4258         u64 *req_value;
4259         char *name;
4260
4261         atomic_dec(&adapter->running_cap_crqs);
4262         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4263         case REQ_TX_QUEUES:
4264                 req_value = &adapter->req_tx_queues;
4265                 name = "tx";
4266                 break;
4267         case REQ_RX_QUEUES:
4268                 req_value = &adapter->req_rx_queues;
4269                 name = "rx";
4270                 break;
4271         case REQ_RX_ADD_QUEUES:
4272                 req_value = &adapter->req_rx_add_queues;
4273                 name = "rx_add";
4274                 break;
4275         case REQ_TX_ENTRIES_PER_SUBCRQ:
4276                 req_value = &adapter->req_tx_entries_per_subcrq;
4277                 name = "tx_entries_per_subcrq";
4278                 break;
4279         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4280                 req_value = &adapter->req_rx_add_entries_per_subcrq;
4281                 name = "rx_add_entries_per_subcrq";
4282                 break;
4283         case REQ_MTU:
4284                 req_value = &adapter->req_mtu;
4285                 name = "mtu";
4286                 break;
4287         case PROMISC_REQUESTED:
4288                 req_value = &adapter->promisc;
4289                 name = "promisc";
4290                 break;
4291         default:
4292                 dev_err(dev, "Got invalid cap request rsp %d\n",
4293                         crq->request_capability.capability);
4294                 return;
4295         }
4296
4297         switch (crq->request_capability_rsp.rc.code) {
4298         case SUCCESS:
4299                 break;
4300         case PARTIALSUCCESS:
4301                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4302                          *req_value,
4303                          (long int)be64_to_cpu(crq->request_capability_rsp.
4304                                                number), name);
4305
4306                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4307                     REQ_MTU) {
4308                         pr_err("mtu of %llu is not supported. Reverting.\n",
4309                                *req_value);
4310                         *req_value = adapter->fallback.mtu;
4311                 } else {
4312                         *req_value =
4313                                 be64_to_cpu(crq->request_capability_rsp.number);
4314                 }
4315
4316                 send_request_cap(adapter, 1);
4317                 return;
4318         default:
4319                 dev_err(dev, "Error %d in request cap rsp\n",
4320                         crq->request_capability_rsp.rc.code);
4321                 return;
4322         }
4323
4324         /* Done receiving requested capabilities, query IP offload support */
4325         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4326                 adapter->wait_capability = false;
4327                 send_query_ip_offload(adapter);
4328         }
4329 }
4330
4331 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4332                             struct ibmvnic_adapter *adapter)
4333 {
4334         struct device *dev = &adapter->vdev->dev;
4335         struct net_device *netdev = adapter->netdev;
4336         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4337         struct ibmvnic_login_buffer *login = adapter->login_buf;
4338         u64 *tx_handle_array;
4339         u64 *rx_handle_array;
4340         int num_tx_pools;
4341         int num_rx_pools;
4342         u64 *size_array;
4343         int i;
4344
4345         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4346                          DMA_TO_DEVICE);
4347         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4348                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4349
4350         /* If the number of queues requested can't be allocated by the
4351          * server, the login response will return with code 1. We will need
4352          * to resend the login buffer with fewer queues requested.
4353          */
4354         if (login_rsp_crq->generic.rc.code) {
4355                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4356                 complete(&adapter->init_done);
4357                 return 0;
4358         }
4359
4360         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4361
4362         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4363         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4364                 netdev_dbg(adapter->netdev, "%016lx\n",
4365                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4366         }
4367
4368         /* Sanity checks */
4369         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4370             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4371              adapter->req_rx_add_queues !=
4372              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4373                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4374                 ibmvnic_remove(adapter->vdev);
4375                 return -EIO;
4376         }
4377         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4378                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
4379         /* variable buffer sizes are not supported, so just read the
4380          * first entry.
4381          */
4382         adapter->cur_rx_buf_sz = be64_to_cpu(size_array[0]);
4383
4384         num_tx_pools = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
4385         num_rx_pools = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4386
4387         tx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4388                                   be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
4389         rx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4390                                   be32_to_cpu(adapter->login_rsp_buf->off_rxadd_subcrqs));
4391
4392         for (i = 0; i < num_tx_pools; i++)
4393                 adapter->tx_scrq[i]->handle = tx_handle_array[i];
4394
4395         for (i = 0; i < num_rx_pools; i++)
4396                 adapter->rx_scrq[i]->handle = rx_handle_array[i];
4397
4398         adapter->num_active_tx_scrqs = num_tx_pools;
4399         adapter->num_active_rx_scrqs = num_rx_pools;
4400         release_login_rsp_buffer(adapter);
4401         release_login_buffer(adapter);
4402         complete(&adapter->init_done);
4403
4404         return 0;
4405 }
4406
4407 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4408                                      struct ibmvnic_adapter *adapter)
4409 {
4410         struct device *dev = &adapter->vdev->dev;
4411         long rc;
4412
4413         rc = crq->request_unmap_rsp.rc.code;
4414         if (rc)
4415                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4416 }
4417
4418 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4419                                  struct ibmvnic_adapter *adapter)
4420 {
4421         struct net_device *netdev = adapter->netdev;
4422         struct device *dev = &adapter->vdev->dev;
4423         long rc;
4424
4425         rc = crq->query_map_rsp.rc.code;
4426         if (rc) {
4427                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4428                 return;
4429         }
4430         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4431                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4432                    crq->query_map_rsp.free_pages);
4433 }
4434
4435 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4436                                  struct ibmvnic_adapter *adapter)
4437 {
4438         struct net_device *netdev = adapter->netdev;
4439         struct device *dev = &adapter->vdev->dev;
4440         long rc;
4441
4442         atomic_dec(&adapter->running_cap_crqs);
4443         netdev_dbg(netdev, "Outstanding queries: %d\n",
4444                    atomic_read(&adapter->running_cap_crqs));
4445         rc = crq->query_capability.rc.code;
4446         if (rc) {
4447                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4448                 goto out;
4449         }
4450
4451         switch (be16_to_cpu(crq->query_capability.capability)) {
4452         case MIN_TX_QUEUES:
4453                 adapter->min_tx_queues =
4454                     be64_to_cpu(crq->query_capability.number);
4455                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4456                            adapter->min_tx_queues);
4457                 break;
4458         case MIN_RX_QUEUES:
4459                 adapter->min_rx_queues =
4460                     be64_to_cpu(crq->query_capability.number);
4461                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4462                            adapter->min_rx_queues);
4463                 break;
4464         case MIN_RX_ADD_QUEUES:
4465                 adapter->min_rx_add_queues =
4466                     be64_to_cpu(crq->query_capability.number);
4467                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4468                            adapter->min_rx_add_queues);
4469                 break;
4470         case MAX_TX_QUEUES:
4471                 adapter->max_tx_queues =
4472                     be64_to_cpu(crq->query_capability.number);
4473                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4474                            adapter->max_tx_queues);
4475                 break;
4476         case MAX_RX_QUEUES:
4477                 adapter->max_rx_queues =
4478                     be64_to_cpu(crq->query_capability.number);
4479                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4480                            adapter->max_rx_queues);
4481                 break;
4482         case MAX_RX_ADD_QUEUES:
4483                 adapter->max_rx_add_queues =
4484                     be64_to_cpu(crq->query_capability.number);
4485                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4486                            adapter->max_rx_add_queues);
4487                 break;
4488         case MIN_TX_ENTRIES_PER_SUBCRQ:
4489                 adapter->min_tx_entries_per_subcrq =
4490                     be64_to_cpu(crq->query_capability.number);
4491                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4492                            adapter->min_tx_entries_per_subcrq);
4493                 break;
4494         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4495                 adapter->min_rx_add_entries_per_subcrq =
4496                     be64_to_cpu(crq->query_capability.number);
4497                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4498                            adapter->min_rx_add_entries_per_subcrq);
4499                 break;
4500         case MAX_TX_ENTRIES_PER_SUBCRQ:
4501                 adapter->max_tx_entries_per_subcrq =
4502                     be64_to_cpu(crq->query_capability.number);
4503                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4504                            adapter->max_tx_entries_per_subcrq);
4505                 break;
4506         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4507                 adapter->max_rx_add_entries_per_subcrq =
4508                     be64_to_cpu(crq->query_capability.number);
4509                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4510                            adapter->max_rx_add_entries_per_subcrq);
4511                 break;
4512         case TCP_IP_OFFLOAD:
4513                 adapter->tcp_ip_offload =
4514                     be64_to_cpu(crq->query_capability.number);
4515                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4516                            adapter->tcp_ip_offload);
4517                 break;
4518         case PROMISC_SUPPORTED:
4519                 adapter->promisc_supported =
4520                     be64_to_cpu(crq->query_capability.number);
4521                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4522                            adapter->promisc_supported);
4523                 break;
4524         case MIN_MTU:
4525                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4526                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4527                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4528                 break;
4529         case MAX_MTU:
4530                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4531                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4532                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4533                 break;
4534         case MAX_MULTICAST_FILTERS:
4535                 adapter->max_multicast_filters =
4536                     be64_to_cpu(crq->query_capability.number);
4537                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4538                            adapter->max_multicast_filters);
4539                 break;
4540         case VLAN_HEADER_INSERTION:
4541                 adapter->vlan_header_insertion =
4542                     be64_to_cpu(crq->query_capability.number);
4543                 if (adapter->vlan_header_insertion)
4544                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4545                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4546                            adapter->vlan_header_insertion);
4547                 break;
4548         case RX_VLAN_HEADER_INSERTION:
4549                 adapter->rx_vlan_header_insertion =
4550                     be64_to_cpu(crq->query_capability.number);
4551                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4552                            adapter->rx_vlan_header_insertion);
4553                 break;
4554         case MAX_TX_SG_ENTRIES:
4555                 adapter->max_tx_sg_entries =
4556                     be64_to_cpu(crq->query_capability.number);
4557                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4558                            adapter->max_tx_sg_entries);
4559                 break;
4560         case RX_SG_SUPPORTED:
4561                 adapter->rx_sg_supported =
4562                     be64_to_cpu(crq->query_capability.number);
4563                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4564                            adapter->rx_sg_supported);
4565                 break;
4566         case OPT_TX_COMP_SUB_QUEUES:
4567                 adapter->opt_tx_comp_sub_queues =
4568                     be64_to_cpu(crq->query_capability.number);
4569                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4570                            adapter->opt_tx_comp_sub_queues);
4571                 break;
4572         case OPT_RX_COMP_QUEUES:
4573                 adapter->opt_rx_comp_queues =
4574                     be64_to_cpu(crq->query_capability.number);
4575                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4576                            adapter->opt_rx_comp_queues);
4577                 break;
4578         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4579                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4580                     be64_to_cpu(crq->query_capability.number);
4581                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4582                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4583                 break;
4584         case OPT_TX_ENTRIES_PER_SUBCRQ:
4585                 adapter->opt_tx_entries_per_subcrq =
4586                     be64_to_cpu(crq->query_capability.number);
4587                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4588                            adapter->opt_tx_entries_per_subcrq);
4589                 break;
4590         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4591                 adapter->opt_rxba_entries_per_subcrq =
4592                     be64_to_cpu(crq->query_capability.number);
4593                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4594                            adapter->opt_rxba_entries_per_subcrq);
4595                 break;
4596         case TX_RX_DESC_REQ:
4597                 adapter->tx_rx_desc_req = crq->query_capability.number;
4598                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4599                            adapter->tx_rx_desc_req);
4600                 break;
4601
4602         default:
4603                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4604                            crq->query_capability.capability);
4605         }
4606
4607 out:
4608         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4609                 adapter->wait_capability = false;
4610                 send_request_cap(adapter, 0);
4611         }
4612 }
4613
4614 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4615 {
4616         union ibmvnic_crq crq;
4617         int rc;
4618
4619         memset(&crq, 0, sizeof(crq));
4620         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4621         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4622
4623         mutex_lock(&adapter->fw_lock);
4624         adapter->fw_done_rc = 0;
4625         reinit_completion(&adapter->fw_done);
4626
4627         rc = ibmvnic_send_crq(adapter, &crq);
4628         if (rc) {
4629                 mutex_unlock(&adapter->fw_lock);
4630                 return rc;
4631         }
4632
4633         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4634         if (rc) {
4635                 mutex_unlock(&adapter->fw_lock);
4636                 return rc;
4637         }
4638
4639         mutex_unlock(&adapter->fw_lock);
4640         return adapter->fw_done_rc ? -EIO : 0;
4641 }
4642
4643 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4644                                        struct ibmvnic_adapter *adapter)
4645 {
4646         struct net_device *netdev = adapter->netdev;
4647         int rc;
4648         __be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4649
4650         rc = crq->query_phys_parms_rsp.rc.code;
4651         if (rc) {
4652                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4653                 return rc;
4654         }
4655         switch (rspeed) {
4656         case IBMVNIC_10MBPS:
4657                 adapter->speed = SPEED_10;
4658                 break;
4659         case IBMVNIC_100MBPS:
4660                 adapter->speed = SPEED_100;
4661                 break;
4662         case IBMVNIC_1GBPS:
4663                 adapter->speed = SPEED_1000;
4664                 break;
4665         case IBMVNIC_10GBPS:
4666                 adapter->speed = SPEED_10000;
4667                 break;
4668         case IBMVNIC_25GBPS:
4669                 adapter->speed = SPEED_25000;
4670                 break;
4671         case IBMVNIC_40GBPS:
4672                 adapter->speed = SPEED_40000;
4673                 break;
4674         case IBMVNIC_50GBPS:
4675                 adapter->speed = SPEED_50000;
4676                 break;
4677         case IBMVNIC_100GBPS:
4678                 adapter->speed = SPEED_100000;
4679                 break;
4680         case IBMVNIC_200GBPS:
4681                 adapter->speed = SPEED_200000;
4682                 break;
4683         default:
4684                 if (netif_carrier_ok(netdev))
4685                         netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4686                 adapter->speed = SPEED_UNKNOWN;
4687         }
4688         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4689                 adapter->duplex = DUPLEX_FULL;
4690         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4691                 adapter->duplex = DUPLEX_HALF;
4692         else
4693                 adapter->duplex = DUPLEX_UNKNOWN;
4694
4695         return rc;
4696 }
4697
4698 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4699                                struct ibmvnic_adapter *adapter)
4700 {
4701         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4702         struct net_device *netdev = adapter->netdev;
4703         struct device *dev = &adapter->vdev->dev;
4704         u64 *u64_crq = (u64 *)crq;
4705         long rc;
4706
4707         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4708                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4709                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4710         switch (gen_crq->first) {
4711         case IBMVNIC_CRQ_INIT_RSP:
4712                 switch (gen_crq->cmd) {
4713                 case IBMVNIC_CRQ_INIT:
4714                         dev_info(dev, "Partner initialized\n");
4715                         adapter->from_passive_init = true;
4716                         adapter->failover_pending = false;
4717                         if (!completion_done(&adapter->init_done)) {
4718                                 complete(&adapter->init_done);
4719                                 adapter->init_done_rc = -EIO;
4720                         }
4721                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4722                         break;
4723                 case IBMVNIC_CRQ_INIT_COMPLETE:
4724                         dev_info(dev, "Partner initialization complete\n");
4725                         adapter->crq.active = true;
4726                         send_version_xchg(adapter);
4727                         break;
4728                 default:
4729                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4730                 }
4731                 return;
4732         case IBMVNIC_CRQ_XPORT_EVENT:
4733                 netif_carrier_off(netdev);
4734                 adapter->crq.active = false;
4735                 /* terminate any thread waiting for a response
4736                  * from the device
4737                  */
4738                 if (!completion_done(&adapter->fw_done)) {
4739                         adapter->fw_done_rc = -EIO;
4740                         complete(&adapter->fw_done);
4741                 }
4742                 if (!completion_done(&adapter->stats_done))
4743                         complete(&adapter->stats_done);
4744                 if (test_bit(0, &adapter->resetting))
4745                         adapter->force_reset_recovery = true;
4746                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4747                         dev_info(dev, "Migrated, re-enabling adapter\n");
4748                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4749                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4750                         dev_info(dev, "Backing device failover detected\n");
4751                         adapter->failover_pending = true;
4752                 } else {
4753                         /* The adapter lost the connection */
4754                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4755                                 gen_crq->cmd);
4756                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4757                 }
4758                 return;
4759         case IBMVNIC_CRQ_CMD_RSP:
4760                 break;
4761         default:
4762                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4763                         gen_crq->first);
4764                 return;
4765         }
4766
4767         switch (gen_crq->cmd) {
4768         case VERSION_EXCHANGE_RSP:
4769                 rc = crq->version_exchange_rsp.rc.code;
4770                 if (rc) {
4771                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4772                         break;
4773                 }
4774                 ibmvnic_version =
4775                             be16_to_cpu(crq->version_exchange_rsp.version);
4776                 dev_info(dev, "Partner protocol version is %d\n",
4777                          ibmvnic_version);
4778                 send_query_cap(adapter);
4779                 break;
4780         case QUERY_CAPABILITY_RSP:
4781                 handle_query_cap_rsp(crq, adapter);
4782                 break;
4783         case QUERY_MAP_RSP:
4784                 handle_query_map_rsp(crq, adapter);
4785                 break;
4786         case REQUEST_MAP_RSP:
4787                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4788                 complete(&adapter->fw_done);
4789                 break;
4790         case REQUEST_UNMAP_RSP:
4791                 handle_request_unmap_rsp(crq, adapter);
4792                 break;
4793         case REQUEST_CAPABILITY_RSP:
4794                 handle_request_cap_rsp(crq, adapter);
4795                 break;
4796         case LOGIN_RSP:
4797                 netdev_dbg(netdev, "Got Login Response\n");
4798                 handle_login_rsp(crq, adapter);
4799                 break;
4800         case LOGICAL_LINK_STATE_RSP:
4801                 netdev_dbg(netdev,
4802                            "Got Logical Link State Response, state: %d rc: %d\n",
4803                            crq->logical_link_state_rsp.link_state,
4804                            crq->logical_link_state_rsp.rc.code);
4805                 adapter->logical_link_state =
4806                     crq->logical_link_state_rsp.link_state;
4807                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4808                 complete(&adapter->init_done);
4809                 break;
4810         case LINK_STATE_INDICATION:
4811                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4812                 adapter->phys_link_state =
4813                     crq->link_state_indication.phys_link_state;
4814                 adapter->logical_link_state =
4815                     crq->link_state_indication.logical_link_state;
4816                 if (adapter->phys_link_state && adapter->logical_link_state)
4817                         netif_carrier_on(netdev);
4818                 else
4819                         netif_carrier_off(netdev);
4820                 break;
4821         case CHANGE_MAC_ADDR_RSP:
4822                 netdev_dbg(netdev, "Got MAC address change Response\n");
4823                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4824                 break;
4825         case ERROR_INDICATION:
4826                 netdev_dbg(netdev, "Got Error Indication\n");
4827                 handle_error_indication(crq, adapter);
4828                 break;
4829         case REQUEST_STATISTICS_RSP:
4830                 netdev_dbg(netdev, "Got Statistics Response\n");
4831                 complete(&adapter->stats_done);
4832                 break;
4833         case QUERY_IP_OFFLOAD_RSP:
4834                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4835                 handle_query_ip_offload_rsp(adapter);
4836                 break;
4837         case MULTICAST_CTRL_RSP:
4838                 netdev_dbg(netdev, "Got multicast control Response\n");
4839                 break;
4840         case CONTROL_IP_OFFLOAD_RSP:
4841                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4842                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4843                                  sizeof(adapter->ip_offload_ctrl),
4844                                  DMA_TO_DEVICE);
4845                 complete(&adapter->init_done);
4846                 break;
4847         case COLLECT_FW_TRACE_RSP:
4848                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4849                 complete(&adapter->fw_done);
4850                 break;
4851         case GET_VPD_SIZE_RSP:
4852                 handle_vpd_size_rsp(crq, adapter);
4853                 break;
4854         case GET_VPD_RSP:
4855                 handle_vpd_rsp(crq, adapter);
4856                 break;
4857         case QUERY_PHYS_PARMS_RSP:
4858                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4859                 complete(&adapter->fw_done);
4860                 break;
4861         default:
4862                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4863                            gen_crq->cmd);
4864         }
4865 }
4866
4867 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4868 {
4869         struct ibmvnic_adapter *adapter = instance;
4870
4871         tasklet_schedule(&adapter->tasklet);
4872         return IRQ_HANDLED;
4873 }
4874
4875 static void ibmvnic_tasklet(struct tasklet_struct *t)
4876 {
4877         struct ibmvnic_adapter *adapter = from_tasklet(adapter, t, tasklet);
4878         struct ibmvnic_crq_queue *queue = &adapter->crq;
4879         union ibmvnic_crq *crq;
4880         unsigned long flags;
4881         bool done = false;
4882
4883         spin_lock_irqsave(&queue->lock, flags);
4884         while (!done) {
4885                 /* Pull all the valid messages off the CRQ */
4886                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4887                         ibmvnic_handle_crq(crq, adapter);
4888                         crq->generic.first = 0;
4889                 }
4890
4891                 /* remain in tasklet until all
4892                  * capabilities responses are received
4893                  */
4894                 if (!adapter->wait_capability)
4895                         done = true;
4896         }
4897         /* if capabilities CRQ's were sent in this tasklet, the following
4898          * tasklet must wait until all responses are received
4899          */
4900         if (atomic_read(&adapter->running_cap_crqs) != 0)
4901                 adapter->wait_capability = true;
4902         spin_unlock_irqrestore(&queue->lock, flags);
4903 }
4904
4905 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4906 {
4907         struct vio_dev *vdev = adapter->vdev;
4908         int rc;
4909
4910         do {
4911                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4912         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4913
4914         if (rc)
4915                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4916
4917         return rc;
4918 }
4919
4920 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4921 {
4922         struct ibmvnic_crq_queue *crq = &adapter->crq;
4923         struct device *dev = &adapter->vdev->dev;
4924         struct vio_dev *vdev = adapter->vdev;
4925         int rc;
4926
4927         /* Close the CRQ */
4928         do {
4929                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4930         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4931
4932         /* Clean out the queue */
4933         memset(crq->msgs, 0, PAGE_SIZE);
4934         crq->cur = 0;
4935         crq->active = false;
4936
4937         /* And re-open it again */
4938         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4939                                 crq->msg_token, PAGE_SIZE);
4940
4941         if (rc == H_CLOSED)
4942                 /* Adapter is good, but other end is not ready */
4943                 dev_warn(dev, "Partner adapter not ready\n");
4944         else if (rc != 0)
4945                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4946
4947         return rc;
4948 }
4949
4950 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4951 {
4952         struct ibmvnic_crq_queue *crq = &adapter->crq;
4953         struct vio_dev *vdev = adapter->vdev;
4954         long rc;
4955
4956         if (!crq->msgs)
4957                 return;
4958
4959         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4960         free_irq(vdev->irq, adapter);
4961         tasklet_kill(&adapter->tasklet);
4962         do {
4963                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4964         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4965
4966         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4967                          DMA_BIDIRECTIONAL);
4968         free_page((unsigned long)crq->msgs);
4969         crq->msgs = NULL;
4970         crq->active = false;
4971 }
4972
4973 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4974 {
4975         struct ibmvnic_crq_queue *crq = &adapter->crq;
4976         struct device *dev = &adapter->vdev->dev;
4977         struct vio_dev *vdev = adapter->vdev;
4978         int rc, retrc = -ENOMEM;
4979
4980         if (crq->msgs)
4981                 return 0;
4982
4983         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4984         /* Should we allocate more than one page? */
4985
4986         if (!crq->msgs)
4987                 return -ENOMEM;
4988
4989         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4990         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4991                                         DMA_BIDIRECTIONAL);
4992         if (dma_mapping_error(dev, crq->msg_token))
4993                 goto map_failed;
4994
4995         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4996                                 crq->msg_token, PAGE_SIZE);
4997
4998         if (rc == H_RESOURCE)
4999                 /* maybe kexecing and resource is busy. try a reset */
5000                 rc = ibmvnic_reset_crq(adapter);
5001         retrc = rc;
5002
5003         if (rc == H_CLOSED) {
5004                 dev_warn(dev, "Partner adapter not ready\n");
5005         } else if (rc) {
5006                 dev_warn(dev, "Error %d opening adapter\n", rc);
5007                 goto reg_crq_failed;
5008         }
5009
5010         retrc = 0;
5011
5012         tasklet_setup(&adapter->tasklet, (void *)ibmvnic_tasklet);
5013
5014         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
5015         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
5016                  adapter->vdev->unit_address);
5017         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
5018         if (rc) {
5019                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
5020                         vdev->irq, rc);
5021                 goto req_irq_failed;
5022         }
5023
5024         rc = vio_enable_interrupts(vdev);
5025         if (rc) {
5026                 dev_err(dev, "Error %d enabling interrupts\n", rc);
5027                 goto req_irq_failed;
5028         }
5029
5030         crq->cur = 0;
5031         spin_lock_init(&crq->lock);
5032
5033         return retrc;
5034
5035 req_irq_failed:
5036         tasklet_kill(&adapter->tasklet);
5037         do {
5038                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5039         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5040 reg_crq_failed:
5041         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
5042 map_failed:
5043         free_page((unsigned long)crq->msgs);
5044         crq->msgs = NULL;
5045         return retrc;
5046 }
5047
5048 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter, bool reset)
5049 {
5050         struct device *dev = &adapter->vdev->dev;
5051         unsigned long timeout = msecs_to_jiffies(30000);
5052         u64 old_num_rx_queues, old_num_tx_queues;
5053         int rc;
5054
5055         adapter->from_passive_init = false;
5056
5057         if (reset) {
5058                 old_num_rx_queues = adapter->req_rx_queues;
5059                 old_num_tx_queues = adapter->req_tx_queues;
5060                 reinit_completion(&adapter->init_done);
5061         }
5062
5063         adapter->init_done_rc = 0;
5064         rc = ibmvnic_send_crq_init(adapter);
5065         if (rc) {
5066                 dev_err(dev, "Send crq init failed with error %d\n", rc);
5067                 return rc;
5068         }
5069
5070         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5071                 dev_err(dev, "Initialization sequence timed out\n");
5072                 return -1;
5073         }
5074
5075         if (adapter->init_done_rc) {
5076                 release_crq_queue(adapter);
5077                 return adapter->init_done_rc;
5078         }
5079
5080         if (adapter->from_passive_init) {
5081                 adapter->state = VNIC_OPEN;
5082                 adapter->from_passive_init = false;
5083                 return -1;
5084         }
5085
5086         if (reset &&
5087             test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5088             adapter->reset_reason != VNIC_RESET_MOBILITY) {
5089                 if (adapter->req_rx_queues != old_num_rx_queues ||
5090                     adapter->req_tx_queues != old_num_tx_queues) {
5091                         release_sub_crqs(adapter, 0);
5092                         rc = init_sub_crqs(adapter);
5093                 } else {
5094                         rc = reset_sub_crq_queues(adapter);
5095                 }
5096         } else {
5097                 rc = init_sub_crqs(adapter);
5098         }
5099
5100         if (rc) {
5101                 dev_err(dev, "Initialization of sub crqs failed\n");
5102                 release_crq_queue(adapter);
5103                 return rc;
5104         }
5105
5106         rc = init_sub_crq_irqs(adapter);
5107         if (rc) {
5108                 dev_err(dev, "Failed to initialize sub crq irqs\n");
5109                 release_crq_queue(adapter);
5110         }
5111
5112         return rc;
5113 }
5114
5115 static struct device_attribute dev_attr_failover;
5116
5117 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5118 {
5119         struct ibmvnic_adapter *adapter;
5120         struct net_device *netdev;
5121         unsigned char *mac_addr_p;
5122         int rc;
5123
5124         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5125                 dev->unit_address);
5126
5127         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5128                                                         VETH_MAC_ADDR, NULL);
5129         if (!mac_addr_p) {
5130                 dev_err(&dev->dev,
5131                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5132                         __FILE__, __LINE__);
5133                 return 0;
5134         }
5135
5136         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5137                                    IBMVNIC_MAX_QUEUES);
5138         if (!netdev)
5139                 return -ENOMEM;
5140
5141         adapter = netdev_priv(netdev);
5142         adapter->state = VNIC_PROBING;
5143         dev_set_drvdata(&dev->dev, netdev);
5144         adapter->vdev = dev;
5145         adapter->netdev = netdev;
5146
5147         ether_addr_copy(adapter->mac_addr, mac_addr_p);
5148         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5149         netdev->irq = dev->irq;
5150         netdev->netdev_ops = &ibmvnic_netdev_ops;
5151         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5152         SET_NETDEV_DEV(netdev, &dev->dev);
5153
5154         spin_lock_init(&adapter->stats_lock);
5155
5156         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5157         INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5158                           __ibmvnic_delayed_reset);
5159         INIT_LIST_HEAD(&adapter->rwi_list);
5160         spin_lock_init(&adapter->rwi_lock);
5161         spin_lock_init(&adapter->state_lock);
5162         mutex_init(&adapter->fw_lock);
5163         init_completion(&adapter->init_done);
5164         init_completion(&adapter->fw_done);
5165         init_completion(&adapter->reset_done);
5166         init_completion(&adapter->stats_done);
5167         clear_bit(0, &adapter->resetting);
5168
5169         do {
5170                 rc = init_crq_queue(adapter);
5171                 if (rc) {
5172                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5173                                 rc);
5174                         goto ibmvnic_init_fail;
5175                 }
5176
5177                 rc = ibmvnic_reset_init(adapter, false);
5178                 if (rc && rc != EAGAIN)
5179                         goto ibmvnic_init_fail;
5180         } while (rc == EAGAIN);
5181
5182         rc = init_stats_buffers(adapter);
5183         if (rc)
5184                 goto ibmvnic_init_fail;
5185
5186         rc = init_stats_token(adapter);
5187         if (rc)
5188                 goto ibmvnic_stats_fail;
5189
5190         netdev->mtu = adapter->req_mtu - ETH_HLEN;
5191         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5192         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5193
5194         rc = device_create_file(&dev->dev, &dev_attr_failover);
5195         if (rc)
5196                 goto ibmvnic_dev_file_err;
5197
5198         netif_carrier_off(netdev);
5199         rc = register_netdev(netdev);
5200         if (rc) {
5201                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5202                 goto ibmvnic_register_fail;
5203         }
5204         dev_info(&dev->dev, "ibmvnic registered\n");
5205
5206         adapter->state = VNIC_PROBED;
5207
5208         adapter->wait_for_reset = false;
5209
5210         return 0;
5211
5212 ibmvnic_register_fail:
5213         device_remove_file(&dev->dev, &dev_attr_failover);
5214
5215 ibmvnic_dev_file_err:
5216         release_stats_token(adapter);
5217
5218 ibmvnic_stats_fail:
5219         release_stats_buffers(adapter);
5220
5221 ibmvnic_init_fail:
5222         release_sub_crqs(adapter, 1);
5223         release_crq_queue(adapter);
5224         mutex_destroy(&adapter->fw_lock);
5225         free_netdev(netdev);
5226
5227         return rc;
5228 }
5229
5230 static int ibmvnic_remove(struct vio_dev *dev)
5231 {
5232         struct net_device *netdev = dev_get_drvdata(&dev->dev);
5233         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5234         unsigned long flags;
5235
5236         spin_lock_irqsave(&adapter->state_lock, flags);
5237         if (adapter->state == VNIC_RESETTING) {
5238                 spin_unlock_irqrestore(&adapter->state_lock, flags);
5239                 return -EBUSY;
5240         }
5241
5242         adapter->state = VNIC_REMOVING;
5243         spin_unlock_irqrestore(&adapter->state_lock, flags);
5244
5245         flush_work(&adapter->ibmvnic_reset);
5246         flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5247
5248         rtnl_lock();
5249         unregister_netdevice(netdev);
5250
5251         release_resources(adapter);
5252         release_sub_crqs(adapter, 1);
5253         release_crq_queue(adapter);
5254
5255         release_stats_token(adapter);
5256         release_stats_buffers(adapter);
5257
5258         adapter->state = VNIC_REMOVED;
5259
5260         rtnl_unlock();
5261         mutex_destroy(&adapter->fw_lock);
5262         device_remove_file(&dev->dev, &dev_attr_failover);
5263         free_netdev(netdev);
5264         dev_set_drvdata(&dev->dev, NULL);
5265
5266         return 0;
5267 }
5268
5269 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5270                               const char *buf, size_t count)
5271 {
5272         struct net_device *netdev = dev_get_drvdata(dev);
5273         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5274         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5275         __be64 session_token;
5276         long rc;
5277
5278         if (!sysfs_streq(buf, "1"))
5279                 return -EINVAL;
5280
5281         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5282                          H_GET_SESSION_TOKEN, 0, 0, 0);
5283         if (rc) {
5284                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5285                            rc);
5286                 return -EINVAL;
5287         }
5288
5289         session_token = (__be64)retbuf[0];
5290         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5291                    be64_to_cpu(session_token));
5292         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5293                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
5294         if (rc) {
5295                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5296                            rc);
5297                 return -EINVAL;
5298         }
5299
5300         return count;
5301 }
5302
5303 static DEVICE_ATTR_WO(failover);
5304
5305 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5306 {
5307         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5308         struct ibmvnic_adapter *adapter;
5309         struct iommu_table *tbl;
5310         unsigned long ret = 0;
5311         int i;
5312
5313         tbl = get_iommu_table_base(&vdev->dev);
5314
5315         /* netdev inits at probe time along with the structures we need below*/
5316         if (!netdev)
5317                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5318
5319         adapter = netdev_priv(netdev);
5320
5321         ret += PAGE_SIZE; /* the crq message queue */
5322         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5323
5324         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5325                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
5326
5327         for (i = 0; i < adapter->num_active_rx_pools; i++)
5328                 ret += adapter->rx_pool[i].size *
5329                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5330
5331         return ret;
5332 }
5333
5334 static int ibmvnic_resume(struct device *dev)
5335 {
5336         struct net_device *netdev = dev_get_drvdata(dev);
5337         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5338
5339         if (adapter->state != VNIC_OPEN)
5340                 return 0;
5341
5342         tasklet_schedule(&adapter->tasklet);
5343
5344         return 0;
5345 }
5346
5347 static const struct vio_device_id ibmvnic_device_table[] = {
5348         {"network", "IBM,vnic"},
5349         {"", "" }
5350 };
5351 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5352
5353 static const struct dev_pm_ops ibmvnic_pm_ops = {
5354         .resume = ibmvnic_resume
5355 };
5356
5357 static struct vio_driver ibmvnic_driver = {
5358         .id_table       = ibmvnic_device_table,
5359         .probe          = ibmvnic_probe,
5360         .remove         = ibmvnic_remove,
5361         .get_desired_dma = ibmvnic_get_desired_dma,
5362         .name           = ibmvnic_driver_name,
5363         .pm             = &ibmvnic_pm_ops,
5364 };
5365
5366 /* module functions */
5367 static int __init ibmvnic_module_init(void)
5368 {
5369         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5370                 IBMVNIC_DRIVER_VERSION);
5371
5372         return vio_register_driver(&ibmvnic_driver);
5373 }
5374
5375 static void __exit ibmvnic_module_exit(void)
5376 {
5377         vio_unregister_driver(&ibmvnic_driver);
5378 }
5379
5380 module_init(ibmvnic_module_init);
5381 module_exit(ibmvnic_module_exit);