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                         goto out;
1189
1190                 rc = init_resources(adapter);
1191                 if (rc) {
1192                         netdev_err(netdev, "failed to initialize resources\n");
1193                         release_resources(adapter);
1194                         goto out;
1195                 }
1196         }
1197
1198         rc = __ibmvnic_open(netdev);
1199
1200 out:
1201         /*
1202          * If open fails due to a pending failover, set device state and
1203          * return. Device operation will be handled by reset routine.
1204          */
1205         if (rc && adapter->failover_pending) {
1206                 adapter->state = VNIC_OPEN;
1207                 rc = 0;
1208         }
1209         return rc;
1210 }
1211
1212 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1213 {
1214         struct ibmvnic_rx_pool *rx_pool;
1215         struct ibmvnic_rx_buff *rx_buff;
1216         u64 rx_entries;
1217         int rx_scrqs;
1218         int i, j;
1219
1220         if (!adapter->rx_pool)
1221                 return;
1222
1223         rx_scrqs = adapter->num_active_rx_pools;
1224         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1225
1226         /* Free any remaining skbs in the rx buffer pools */
1227         for (i = 0; i < rx_scrqs; i++) {
1228                 rx_pool = &adapter->rx_pool[i];
1229                 if (!rx_pool || !rx_pool->rx_buff)
1230                         continue;
1231
1232                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1233                 for (j = 0; j < rx_entries; j++) {
1234                         rx_buff = &rx_pool->rx_buff[j];
1235                         if (rx_buff && rx_buff->skb) {
1236                                 dev_kfree_skb_any(rx_buff->skb);
1237                                 rx_buff->skb = NULL;
1238                         }
1239                 }
1240         }
1241 }
1242
1243 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1244                               struct ibmvnic_tx_pool *tx_pool)
1245 {
1246         struct ibmvnic_tx_buff *tx_buff;
1247         u64 tx_entries;
1248         int i;
1249
1250         if (!tx_pool || !tx_pool->tx_buff)
1251                 return;
1252
1253         tx_entries = tx_pool->num_buffers;
1254
1255         for (i = 0; i < tx_entries; i++) {
1256                 tx_buff = &tx_pool->tx_buff[i];
1257                 if (tx_buff && tx_buff->skb) {
1258                         dev_kfree_skb_any(tx_buff->skb);
1259                         tx_buff->skb = NULL;
1260                 }
1261         }
1262 }
1263
1264 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1265 {
1266         int tx_scrqs;
1267         int i;
1268
1269         if (!adapter->tx_pool || !adapter->tso_pool)
1270                 return;
1271
1272         tx_scrqs = adapter->num_active_tx_pools;
1273
1274         /* Free any remaining skbs in the tx buffer pools */
1275         for (i = 0; i < tx_scrqs; i++) {
1276                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1277                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1278                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1279         }
1280 }
1281
1282 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1283 {
1284         struct net_device *netdev = adapter->netdev;
1285         int i;
1286
1287         if (adapter->tx_scrq) {
1288                 for (i = 0; i < adapter->req_tx_queues; i++)
1289                         if (adapter->tx_scrq[i]->irq) {
1290                                 netdev_dbg(netdev,
1291                                            "Disabling tx_scrq[%d] irq\n", i);
1292                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1293                                 disable_irq(adapter->tx_scrq[i]->irq);
1294                         }
1295         }
1296
1297         if (adapter->rx_scrq) {
1298                 for (i = 0; i < adapter->req_rx_queues; i++) {
1299                         if (adapter->rx_scrq[i]->irq) {
1300                                 netdev_dbg(netdev,
1301                                            "Disabling rx_scrq[%d] irq\n", i);
1302                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1303                                 disable_irq(adapter->rx_scrq[i]->irq);
1304                         }
1305                 }
1306         }
1307 }
1308
1309 static void ibmvnic_cleanup(struct net_device *netdev)
1310 {
1311         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1312
1313         /* ensure that transmissions are stopped if called by do_reset */
1314         if (test_bit(0, &adapter->resetting))
1315                 netif_tx_disable(netdev);
1316         else
1317                 netif_tx_stop_all_queues(netdev);
1318
1319         ibmvnic_napi_disable(adapter);
1320         ibmvnic_disable_irqs(adapter);
1321
1322         clean_rx_pools(adapter);
1323         clean_tx_pools(adapter);
1324 }
1325
1326 static int __ibmvnic_close(struct net_device *netdev)
1327 {
1328         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1329         int rc = 0;
1330
1331         adapter->state = VNIC_CLOSING;
1332         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1333         if (rc)
1334                 return rc;
1335         adapter->state = VNIC_CLOSED;
1336         return 0;
1337 }
1338
1339 static int ibmvnic_close(struct net_device *netdev)
1340 {
1341         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1342         int rc;
1343
1344         /* If device failover is pending, just set device state and return.
1345          * Device operation will be handled by reset routine.
1346          */
1347         if (adapter->failover_pending) {
1348                 adapter->state = VNIC_CLOSED;
1349                 return 0;
1350         }
1351
1352         rc = __ibmvnic_close(netdev);
1353         ibmvnic_cleanup(netdev);
1354
1355         return rc;
1356 }
1357
1358 /**
1359  * build_hdr_data - creates L2/L3/L4 header data buffer
1360  * @hdr_field - bitfield determining needed headers
1361  * @skb - socket buffer
1362  * @hdr_len - array of header lengths
1363  * @tot_len - total length of data
1364  *
1365  * Reads hdr_field to determine which headers are needed by firmware.
1366  * Builds a buffer containing these headers.  Saves individual header
1367  * lengths and total buffer length to be used to build descriptors.
1368  */
1369 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1370                           int *hdr_len, u8 *hdr_data)
1371 {
1372         int len = 0;
1373         u8 *hdr;
1374
1375         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1376                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1377         else
1378                 hdr_len[0] = sizeof(struct ethhdr);
1379
1380         if (skb->protocol == htons(ETH_P_IP)) {
1381                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1382                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1383                         hdr_len[2] = tcp_hdrlen(skb);
1384                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1385                         hdr_len[2] = sizeof(struct udphdr);
1386         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1387                 hdr_len[1] = sizeof(struct ipv6hdr);
1388                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1389                         hdr_len[2] = tcp_hdrlen(skb);
1390                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1391                         hdr_len[2] = sizeof(struct udphdr);
1392         } else if (skb->protocol == htons(ETH_P_ARP)) {
1393                 hdr_len[1] = arp_hdr_len(skb->dev);
1394                 hdr_len[2] = 0;
1395         }
1396
1397         memset(hdr_data, 0, 120);
1398         if ((hdr_field >> 6) & 1) {
1399                 hdr = skb_mac_header(skb);
1400                 memcpy(hdr_data, hdr, hdr_len[0]);
1401                 len += hdr_len[0];
1402         }
1403
1404         if ((hdr_field >> 5) & 1) {
1405                 hdr = skb_network_header(skb);
1406                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1407                 len += hdr_len[1];
1408         }
1409
1410         if ((hdr_field >> 4) & 1) {
1411                 hdr = skb_transport_header(skb);
1412                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1413                 len += hdr_len[2];
1414         }
1415         return len;
1416 }
1417
1418 /**
1419  * create_hdr_descs - create header and header extension descriptors
1420  * @hdr_field - bitfield determining needed headers
1421  * @data - buffer containing header data
1422  * @len - length of data buffer
1423  * @hdr_len - array of individual header lengths
1424  * @scrq_arr - descriptor array
1425  *
1426  * Creates header and, if needed, header extension descriptors and
1427  * places them in a descriptor array, scrq_arr
1428  */
1429
1430 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1431                             union sub_crq *scrq_arr)
1432 {
1433         union sub_crq hdr_desc;
1434         int tmp_len = len;
1435         int num_descs = 0;
1436         u8 *data, *cur;
1437         int tmp;
1438
1439         while (tmp_len > 0) {
1440                 cur = hdr_data + len - tmp_len;
1441
1442                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1443                 if (cur != hdr_data) {
1444                         data = hdr_desc.hdr_ext.data;
1445                         tmp = tmp_len > 29 ? 29 : tmp_len;
1446                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1447                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1448                         hdr_desc.hdr_ext.len = tmp;
1449                 } else {
1450                         data = hdr_desc.hdr.data;
1451                         tmp = tmp_len > 24 ? 24 : tmp_len;
1452                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1453                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1454                         hdr_desc.hdr.len = tmp;
1455                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1456                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1457                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1458                         hdr_desc.hdr.flag = hdr_field << 1;
1459                 }
1460                 memcpy(data, cur, tmp);
1461                 tmp_len -= tmp;
1462                 *scrq_arr = hdr_desc;
1463                 scrq_arr++;
1464                 num_descs++;
1465         }
1466
1467         return num_descs;
1468 }
1469
1470 /**
1471  * build_hdr_descs_arr - build a header descriptor array
1472  * @skb - socket buffer
1473  * @num_entries - number of descriptors to be sent
1474  * @subcrq - first TX descriptor
1475  * @hdr_field - bit field determining which headers will be sent
1476  *
1477  * This function will build a TX descriptor array with applicable
1478  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1479  */
1480
1481 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1482                                 int *num_entries, u8 hdr_field)
1483 {
1484         int hdr_len[3] = {0, 0, 0};
1485         int tot_len;
1486         u8 *hdr_data = txbuff->hdr_data;
1487
1488         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1489                                  txbuff->hdr_data);
1490         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1491                          txbuff->indir_arr + 1);
1492 }
1493
1494 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1495                                     struct net_device *netdev)
1496 {
1497         /* For some backing devices, mishandling of small packets
1498          * can result in a loss of connection or TX stall. Device
1499          * architects recommend that no packet should be smaller
1500          * than the minimum MTU value provided to the driver, so
1501          * pad any packets to that length
1502          */
1503         if (skb->len < netdev->min_mtu)
1504                 return skb_put_padto(skb, netdev->min_mtu);
1505
1506         return 0;
1507 }
1508
1509 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1510 {
1511         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1512         int queue_num = skb_get_queue_mapping(skb);
1513         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1514         struct device *dev = &adapter->vdev->dev;
1515         struct ibmvnic_tx_buff *tx_buff = NULL;
1516         struct ibmvnic_sub_crq_queue *tx_scrq;
1517         struct ibmvnic_tx_pool *tx_pool;
1518         unsigned int tx_send_failed = 0;
1519         unsigned int tx_map_failed = 0;
1520         unsigned int tx_dropped = 0;
1521         unsigned int tx_packets = 0;
1522         unsigned int tx_bytes = 0;
1523         dma_addr_t data_dma_addr;
1524         struct netdev_queue *txq;
1525         unsigned long lpar_rc;
1526         union sub_crq tx_crq;
1527         unsigned int offset;
1528         int num_entries = 1;
1529         unsigned char *dst;
1530         int index = 0;
1531         u8 proto = 0;
1532         u64 handle;
1533         netdev_tx_t ret = NETDEV_TX_OK;
1534
1535         if (test_bit(0, &adapter->resetting)) {
1536                 if (!netif_subqueue_stopped(netdev, skb))
1537                         netif_stop_subqueue(netdev, queue_num);
1538                 dev_kfree_skb_any(skb);
1539
1540                 tx_send_failed++;
1541                 tx_dropped++;
1542                 ret = NETDEV_TX_OK;
1543                 goto out;
1544         }
1545
1546         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1547                 tx_dropped++;
1548                 tx_send_failed++;
1549                 ret = NETDEV_TX_OK;
1550                 goto out;
1551         }
1552         if (skb_is_gso(skb))
1553                 tx_pool = &adapter->tso_pool[queue_num];
1554         else
1555                 tx_pool = &adapter->tx_pool[queue_num];
1556
1557         tx_scrq = adapter->tx_scrq[queue_num];
1558         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1559         handle = tx_scrq->handle;
1560
1561         index = tx_pool->free_map[tx_pool->consumer_index];
1562
1563         if (index == IBMVNIC_INVALID_MAP) {
1564                 dev_kfree_skb_any(skb);
1565                 tx_send_failed++;
1566                 tx_dropped++;
1567                 ret = NETDEV_TX_OK;
1568                 goto out;
1569         }
1570
1571         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1572
1573         offset = index * tx_pool->buf_size;
1574         dst = tx_pool->long_term_buff.buff + offset;
1575         memset(dst, 0, tx_pool->buf_size);
1576         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1577
1578         if (skb_shinfo(skb)->nr_frags) {
1579                 int cur, i;
1580
1581                 /* Copy the head */
1582                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1583                 cur = skb_headlen(skb);
1584
1585                 /* Copy the frags */
1586                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1587                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1588
1589                         memcpy(dst + cur,
1590                                page_address(skb_frag_page(frag)) +
1591                                skb_frag_off(frag), skb_frag_size(frag));
1592                         cur += skb_frag_size(frag);
1593                 }
1594         } else {
1595                 skb_copy_from_linear_data(skb, dst, skb->len);
1596         }
1597
1598         tx_pool->consumer_index =
1599             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1600
1601         tx_buff = &tx_pool->tx_buff[index];
1602         tx_buff->skb = skb;
1603         tx_buff->data_dma[0] = data_dma_addr;
1604         tx_buff->data_len[0] = skb->len;
1605         tx_buff->index = index;
1606         tx_buff->pool_index = queue_num;
1607         tx_buff->last_frag = true;
1608
1609         memset(&tx_crq, 0, sizeof(tx_crq));
1610         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1611         tx_crq.v1.type = IBMVNIC_TX_DESC;
1612         tx_crq.v1.n_crq_elem = 1;
1613         tx_crq.v1.n_sge = 1;
1614         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1615
1616         if (skb_is_gso(skb))
1617                 tx_crq.v1.correlator =
1618                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1619         else
1620                 tx_crq.v1.correlator = cpu_to_be32(index);
1621         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1622         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1623         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1624
1625         if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1626                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1627                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1628         }
1629
1630         if (skb->protocol == htons(ETH_P_IP)) {
1631                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1632                 proto = ip_hdr(skb)->protocol;
1633         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1634                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1635                 proto = ipv6_hdr(skb)->nexthdr;
1636         }
1637
1638         if (proto == IPPROTO_TCP)
1639                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1640         else if (proto == IPPROTO_UDP)
1641                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1642
1643         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1644                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1645                 hdrs += 2;
1646         }
1647         if (skb_is_gso(skb)) {
1648                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1649                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1650                 hdrs += 2;
1651         }
1652         /* determine if l2/3/4 headers are sent to firmware */
1653         if ((*hdrs >> 7) & 1) {
1654                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1655                 tx_crq.v1.n_crq_elem = num_entries;
1656                 tx_buff->num_entries = num_entries;
1657                 tx_buff->indir_arr[0] = tx_crq;
1658                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1659                                                     sizeof(tx_buff->indir_arr),
1660                                                     DMA_TO_DEVICE);
1661                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1662                         dev_kfree_skb_any(skb);
1663                         tx_buff->skb = NULL;
1664                         if (!firmware_has_feature(FW_FEATURE_CMO))
1665                                 dev_err(dev, "tx: unable to map descriptor array\n");
1666                         tx_map_failed++;
1667                         tx_dropped++;
1668                         ret = NETDEV_TX_OK;
1669                         goto tx_err_out;
1670                 }
1671                 lpar_rc = send_subcrq_indirect(adapter, handle,
1672                                                (u64)tx_buff->indir_dma,
1673                                                (u64)num_entries);
1674                 dma_unmap_single(dev, tx_buff->indir_dma,
1675                                  sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1676         } else {
1677                 tx_buff->num_entries = num_entries;
1678                 lpar_rc = send_subcrq(adapter, handle,
1679                                       &tx_crq);
1680         }
1681         if (lpar_rc != H_SUCCESS) {
1682                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1683                         dev_err_ratelimited(dev, "tx: send failed\n");
1684                 dev_kfree_skb_any(skb);
1685                 tx_buff->skb = NULL;
1686
1687                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1688                         /* Disable TX and report carrier off if queue is closed
1689                          * or pending failover.
1690                          * Firmware guarantees that a signal will be sent to the
1691                          * driver, triggering a reset or some other action.
1692                          */
1693                         netif_tx_stop_all_queues(netdev);
1694                         netif_carrier_off(netdev);
1695                 }
1696
1697                 tx_send_failed++;
1698                 tx_dropped++;
1699                 ret = NETDEV_TX_OK;
1700                 goto tx_err_out;
1701         }
1702
1703         if (atomic_add_return(num_entries, &tx_scrq->used)
1704                                         >= adapter->req_tx_entries_per_subcrq) {
1705                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1706                 netif_stop_subqueue(netdev, queue_num);
1707         }
1708
1709         tx_packets++;
1710         tx_bytes += skb->len;
1711         txq->trans_start = jiffies;
1712         ret = NETDEV_TX_OK;
1713         goto out;
1714
1715 tx_err_out:
1716         /* roll back consumer index and map array*/
1717         if (tx_pool->consumer_index == 0)
1718                 tx_pool->consumer_index =
1719                         tx_pool->num_buffers - 1;
1720         else
1721                 tx_pool->consumer_index--;
1722         tx_pool->free_map[tx_pool->consumer_index] = index;
1723 out:
1724         netdev->stats.tx_dropped += tx_dropped;
1725         netdev->stats.tx_bytes += tx_bytes;
1726         netdev->stats.tx_packets += tx_packets;
1727         adapter->tx_send_failed += tx_send_failed;
1728         adapter->tx_map_failed += tx_map_failed;
1729         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1730         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1731         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1732
1733         return ret;
1734 }
1735
1736 static void ibmvnic_set_multi(struct net_device *netdev)
1737 {
1738         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1739         struct netdev_hw_addr *ha;
1740         union ibmvnic_crq crq;
1741
1742         memset(&crq, 0, sizeof(crq));
1743         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1744         crq.request_capability.cmd = REQUEST_CAPABILITY;
1745
1746         if (netdev->flags & IFF_PROMISC) {
1747                 if (!adapter->promisc_supported)
1748                         return;
1749         } else {
1750                 if (netdev->flags & IFF_ALLMULTI) {
1751                         /* Accept all multicast */
1752                         memset(&crq, 0, sizeof(crq));
1753                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1754                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1755                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1756                         ibmvnic_send_crq(adapter, &crq);
1757                 } else if (netdev_mc_empty(netdev)) {
1758                         /* Reject all multicast */
1759                         memset(&crq, 0, sizeof(crq));
1760                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1761                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1762                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1763                         ibmvnic_send_crq(adapter, &crq);
1764                 } else {
1765                         /* Accept one or more multicast(s) */
1766                         netdev_for_each_mc_addr(ha, netdev) {
1767                                 memset(&crq, 0, sizeof(crq));
1768                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1769                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1770                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1771                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1772                                                 ha->addr);
1773                                 ibmvnic_send_crq(adapter, &crq);
1774                         }
1775                 }
1776         }
1777 }
1778
1779 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1780 {
1781         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1782         union ibmvnic_crq crq;
1783         int rc;
1784
1785         if (!is_valid_ether_addr(dev_addr)) {
1786                 rc = -EADDRNOTAVAIL;
1787                 goto err;
1788         }
1789
1790         memset(&crq, 0, sizeof(crq));
1791         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1792         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1793         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1794
1795         mutex_lock(&adapter->fw_lock);
1796         adapter->fw_done_rc = 0;
1797         reinit_completion(&adapter->fw_done);
1798
1799         rc = ibmvnic_send_crq(adapter, &crq);
1800         if (rc) {
1801                 rc = -EIO;
1802                 mutex_unlock(&adapter->fw_lock);
1803                 goto err;
1804         }
1805
1806         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1807         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1808         if (rc || adapter->fw_done_rc) {
1809                 rc = -EIO;
1810                 mutex_unlock(&adapter->fw_lock);
1811                 goto err;
1812         }
1813         mutex_unlock(&adapter->fw_lock);
1814         return 0;
1815 err:
1816         ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1817         return rc;
1818 }
1819
1820 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1821 {
1822         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1823         struct sockaddr *addr = p;
1824         int rc;
1825
1826         rc = 0;
1827         if (!is_valid_ether_addr(addr->sa_data))
1828                 return -EADDRNOTAVAIL;
1829
1830         if (adapter->state != VNIC_PROBED) {
1831                 ether_addr_copy(adapter->mac_addr, addr->sa_data);
1832                 rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1833         }
1834
1835         return rc;
1836 }
1837
1838 /**
1839  * do_change_param_reset returns zero if we are able to keep processing reset
1840  * events, or non-zero if we hit a fatal error and must halt.
1841  */
1842 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1843                                  struct ibmvnic_rwi *rwi,
1844                                  u32 reset_state)
1845 {
1846         struct net_device *netdev = adapter->netdev;
1847         int i, rc;
1848
1849         netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1850                    rwi->reset_reason);
1851
1852         netif_carrier_off(netdev);
1853         adapter->reset_reason = rwi->reset_reason;
1854
1855         ibmvnic_cleanup(netdev);
1856
1857         if (reset_state == VNIC_OPEN) {
1858                 rc = __ibmvnic_close(netdev);
1859                 if (rc)
1860                         goto out;
1861         }
1862
1863         release_resources(adapter);
1864         release_sub_crqs(adapter, 1);
1865         release_crq_queue(adapter);
1866
1867         adapter->state = VNIC_PROBED;
1868
1869         rc = init_crq_queue(adapter);
1870
1871         if (rc) {
1872                 netdev_err(adapter->netdev,
1873                            "Couldn't initialize crq. rc=%d\n", rc);
1874                 return rc;
1875         }
1876
1877         rc = ibmvnic_reset_init(adapter, true);
1878         if (rc) {
1879                 rc = IBMVNIC_INIT_FAILED;
1880                 goto out;
1881         }
1882
1883         /* If the adapter was in PROBE state prior to the reset,
1884          * exit here.
1885          */
1886         if (reset_state == VNIC_PROBED)
1887                 goto out;
1888
1889         rc = ibmvnic_login(netdev);
1890         if (rc) {
1891                 goto out;
1892         }
1893
1894         rc = init_resources(adapter);
1895         if (rc)
1896                 goto out;
1897
1898         ibmvnic_disable_irqs(adapter);
1899
1900         adapter->state = VNIC_CLOSED;
1901
1902         if (reset_state == VNIC_CLOSED)
1903                 return 0;
1904
1905         rc = __ibmvnic_open(netdev);
1906         if (rc) {
1907                 rc = IBMVNIC_OPEN_FAILED;
1908                 goto out;
1909         }
1910
1911         /* refresh device's multicast list */
1912         ibmvnic_set_multi(netdev);
1913
1914         /* kick napi */
1915         for (i = 0; i < adapter->req_rx_queues; i++)
1916                 napi_schedule(&adapter->napi[i]);
1917
1918 out:
1919         if (rc)
1920                 adapter->state = reset_state;
1921         return rc;
1922 }
1923
1924 /**
1925  * do_reset returns zero if we are able to keep processing reset events, or
1926  * non-zero if we hit a fatal error and must halt.
1927  */
1928 static int do_reset(struct ibmvnic_adapter *adapter,
1929                     struct ibmvnic_rwi *rwi, u32 reset_state)
1930 {
1931         u64 old_num_rx_queues, old_num_tx_queues;
1932         u64 old_num_rx_slots, old_num_tx_slots;
1933         struct net_device *netdev = adapter->netdev;
1934         int i, rc;
1935
1936         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1937                    rwi->reset_reason);
1938
1939         rtnl_lock();
1940         /*
1941          * Now that we have the rtnl lock, clear any pending failover.
1942          * This will ensure ibmvnic_open() has either completed or will
1943          * block until failover is complete.
1944          */
1945         if (rwi->reset_reason == VNIC_RESET_FAILOVER)
1946                 adapter->failover_pending = false;
1947
1948         netif_carrier_off(netdev);
1949         adapter->reset_reason = rwi->reset_reason;
1950
1951         old_num_rx_queues = adapter->req_rx_queues;
1952         old_num_tx_queues = adapter->req_tx_queues;
1953         old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1954         old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1955
1956         ibmvnic_cleanup(netdev);
1957
1958         if (reset_state == VNIC_OPEN &&
1959             adapter->reset_reason != VNIC_RESET_MOBILITY &&
1960             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1961                 adapter->state = VNIC_CLOSING;
1962
1963                 /* Release the RTNL lock before link state change and
1964                  * re-acquire after the link state change to allow
1965                  * linkwatch_event to grab the RTNL lock and run during
1966                  * a reset.
1967                  */
1968                 rtnl_unlock();
1969                 rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1970                 rtnl_lock();
1971                 if (rc)
1972                         goto out;
1973
1974                 if (adapter->state != VNIC_CLOSING) {
1975                         rc = -1;
1976                         goto out;
1977                 }
1978
1979                 adapter->state = VNIC_CLOSED;
1980         }
1981
1982         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1983                 /* remove the closed state so when we call open it appears
1984                  * we are coming from the probed state.
1985                  */
1986                 adapter->state = VNIC_PROBED;
1987
1988                 if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1989                         rc = ibmvnic_reenable_crq_queue(adapter);
1990                         release_sub_crqs(adapter, 1);
1991                 } else {
1992                         rc = ibmvnic_reset_crq(adapter);
1993                         if (rc == H_CLOSED || rc == H_SUCCESS) {
1994                                 rc = vio_enable_interrupts(adapter->vdev);
1995                                 if (rc)
1996                                         netdev_err(adapter->netdev,
1997                                                    "Reset failed to enable interrupts. rc=%d\n",
1998                                                    rc);
1999                         }
2000                 }
2001
2002                 if (rc) {
2003                         netdev_err(adapter->netdev,
2004                                    "Reset couldn't initialize crq. rc=%d\n", rc);
2005                         goto out;
2006                 }
2007
2008                 rc = ibmvnic_reset_init(adapter, true);
2009                 if (rc) {
2010                         rc = IBMVNIC_INIT_FAILED;
2011                         goto out;
2012                 }
2013
2014                 /* If the adapter was in PROBE state prior to the reset,
2015                  * exit here.
2016                  */
2017                 if (reset_state == VNIC_PROBED) {
2018                         rc = 0;
2019                         goto out;
2020                 }
2021
2022                 rc = ibmvnic_login(netdev);
2023                 if (rc) {
2024                         goto out;
2025                 }
2026
2027                 if (adapter->req_rx_queues != old_num_rx_queues ||
2028                     adapter->req_tx_queues != old_num_tx_queues ||
2029                     adapter->req_rx_add_entries_per_subcrq !=
2030                     old_num_rx_slots ||
2031                     adapter->req_tx_entries_per_subcrq !=
2032                     old_num_tx_slots ||
2033                     !adapter->rx_pool ||
2034                     !adapter->tso_pool ||
2035                     !adapter->tx_pool) {
2036                         release_rx_pools(adapter);
2037                         release_tx_pools(adapter);
2038                         release_napi(adapter);
2039                         release_vpd_data(adapter);
2040
2041                         rc = init_resources(adapter);
2042                         if (rc)
2043                                 goto out;
2044
2045                 } else {
2046                         rc = reset_tx_pools(adapter);
2047                         if (rc) {
2048                                 netdev_dbg(adapter->netdev, "reset tx pools failed (%d)\n",
2049                                                 rc);
2050                                 goto out;
2051                         }
2052
2053                         rc = reset_rx_pools(adapter);
2054                         if (rc) {
2055                                 netdev_dbg(adapter->netdev, "reset rx pools failed (%d)\n",
2056                                                 rc);
2057                                 goto out;
2058                         }
2059                 }
2060                 ibmvnic_disable_irqs(adapter);
2061         }
2062         adapter->state = VNIC_CLOSED;
2063
2064         if (reset_state == VNIC_CLOSED) {
2065                 rc = 0;
2066                 goto out;
2067         }
2068
2069         rc = __ibmvnic_open(netdev);
2070         if (rc) {
2071                 rc = IBMVNIC_OPEN_FAILED;
2072                 goto out;
2073         }
2074
2075         /* refresh device's multicast list */
2076         ibmvnic_set_multi(netdev);
2077
2078         /* kick napi */
2079         for (i = 0; i < adapter->req_rx_queues; i++)
2080                 napi_schedule(&adapter->napi[i]);
2081
2082         if (adapter->reset_reason == VNIC_RESET_FAILOVER ||
2083             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2084                 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2085                 call_netdevice_notifiers(NETDEV_RESEND_IGMP, netdev);
2086         }
2087
2088         rc = 0;
2089
2090 out:
2091         /* restore the adapter state if reset failed */
2092         if (rc)
2093                 adapter->state = reset_state;
2094         rtnl_unlock();
2095
2096         return rc;
2097 }
2098
2099 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2100                          struct ibmvnic_rwi *rwi, u32 reset_state)
2101 {
2102         struct net_device *netdev = adapter->netdev;
2103         int rc;
2104
2105         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2106                    rwi->reset_reason);
2107
2108         netif_carrier_off(netdev);
2109         adapter->reset_reason = rwi->reset_reason;
2110
2111         ibmvnic_cleanup(netdev);
2112         release_resources(adapter);
2113         release_sub_crqs(adapter, 0);
2114         release_crq_queue(adapter);
2115
2116         /* remove the closed state so when we call open it appears
2117          * we are coming from the probed state.
2118          */
2119         adapter->state = VNIC_PROBED;
2120
2121         reinit_completion(&adapter->init_done);
2122         rc = init_crq_queue(adapter);
2123         if (rc) {
2124                 netdev_err(adapter->netdev,
2125                            "Couldn't initialize crq. rc=%d\n", rc);
2126                 goto out;
2127         }
2128
2129         rc = ibmvnic_reset_init(adapter, false);
2130         if (rc)
2131                 goto out;
2132
2133         /* If the adapter was in PROBE state prior to the reset,
2134          * exit here.
2135          */
2136         if (reset_state == VNIC_PROBED)
2137                 goto out;
2138
2139         rc = ibmvnic_login(netdev);
2140         if (rc)
2141                 goto out;
2142
2143         rc = init_resources(adapter);
2144         if (rc)
2145                 goto out;
2146
2147         ibmvnic_disable_irqs(adapter);
2148         adapter->state = VNIC_CLOSED;
2149
2150         if (reset_state == VNIC_CLOSED)
2151                 goto out;
2152
2153         rc = __ibmvnic_open(netdev);
2154         if (rc) {
2155                 rc = IBMVNIC_OPEN_FAILED;
2156                 goto out;
2157         }
2158
2159         call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2160         call_netdevice_notifiers(NETDEV_RESEND_IGMP, netdev);
2161 out:
2162         /* restore adapter state if reset failed */
2163         if (rc)
2164                 adapter->state = reset_state;
2165         return rc;
2166 }
2167
2168 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2169 {
2170         struct ibmvnic_rwi *rwi;
2171         unsigned long flags;
2172
2173         spin_lock_irqsave(&adapter->rwi_lock, flags);
2174
2175         if (!list_empty(&adapter->rwi_list)) {
2176                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2177                                        list);
2178                 list_del(&rwi->list);
2179         } else {
2180                 rwi = NULL;
2181         }
2182
2183         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2184         return rwi;
2185 }
2186
2187 static void __ibmvnic_reset(struct work_struct *work)
2188 {
2189         struct ibmvnic_rwi *rwi;
2190         struct ibmvnic_adapter *adapter;
2191         bool saved_state = false;
2192         unsigned long flags;
2193         u32 reset_state;
2194         int rc = 0;
2195
2196         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2197
2198         if (test_and_set_bit_lock(0, &adapter->resetting)) {
2199                 schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2200                                       IBMVNIC_RESET_DELAY);
2201                 return;
2202         }
2203
2204         rwi = get_next_rwi(adapter);
2205         while (rwi) {
2206                 spin_lock_irqsave(&adapter->state_lock, flags);
2207
2208                 if (adapter->state == VNIC_REMOVING ||
2209                     adapter->state == VNIC_REMOVED) {
2210                         spin_unlock_irqrestore(&adapter->state_lock, flags);
2211                         kfree(rwi);
2212                         rc = EBUSY;
2213                         break;
2214                 }
2215
2216                 if (!saved_state) {
2217                         reset_state = adapter->state;
2218                         saved_state = true;
2219                 }
2220                 spin_unlock_irqrestore(&adapter->state_lock, flags);
2221
2222                 if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2223                         /* CHANGE_PARAM requestor holds rtnl_lock */
2224                         rc = do_change_param_reset(adapter, rwi, reset_state);
2225                 } else if (adapter->force_reset_recovery) {
2226                         /*
2227                          * Since we are doing a hard reset now, clear the
2228                          * failover_pending flag so we don't ignore any
2229                          * future MOBILITY or other resets.
2230                          */
2231                         adapter->failover_pending = false;
2232
2233                         /* Transport event occurred during previous reset */
2234                         if (adapter->wait_for_reset) {
2235                                 /* Previous was CHANGE_PARAM; caller locked */
2236                                 adapter->force_reset_recovery = false;
2237                                 rc = do_hard_reset(adapter, rwi, reset_state);
2238                         } else {
2239                                 rtnl_lock();
2240                                 adapter->force_reset_recovery = false;
2241                                 rc = do_hard_reset(adapter, rwi, reset_state);
2242                                 rtnl_unlock();
2243                         }
2244                 } else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2245                                 adapter->from_passive_init)) {
2246                         rc = do_reset(adapter, rwi, reset_state);
2247                 }
2248                 kfree(rwi);
2249
2250                 if (rc)
2251                         netdev_dbg(adapter->netdev, "Reset failed, rc=%d\n", rc);
2252
2253                 rwi = get_next_rwi(adapter);
2254
2255                 if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2256                             rwi->reset_reason == VNIC_RESET_MOBILITY))
2257                         adapter->force_reset_recovery = true;
2258         }
2259
2260         if (adapter->wait_for_reset) {
2261                 adapter->reset_done_rc = rc;
2262                 complete(&adapter->reset_done);
2263         }
2264
2265         clear_bit_unlock(0, &adapter->resetting);
2266 }
2267
2268 static void __ibmvnic_delayed_reset(struct work_struct *work)
2269 {
2270         struct ibmvnic_adapter *adapter;
2271
2272         adapter = container_of(work, struct ibmvnic_adapter,
2273                                ibmvnic_delayed_reset.work);
2274         __ibmvnic_reset(&adapter->ibmvnic_reset);
2275 }
2276
2277 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2278                          enum ibmvnic_reset_reason reason)
2279 {
2280         struct list_head *entry, *tmp_entry;
2281         struct ibmvnic_rwi *rwi, *tmp;
2282         struct net_device *netdev = adapter->netdev;
2283         unsigned long flags;
2284         int ret;
2285
2286         /*
2287          * If failover is pending don't schedule any other reset.
2288          * Instead let the failover complete. If there is already a
2289          * a failover reset scheduled, we will detect and drop the
2290          * duplicate reset when walking the ->rwi_list below.
2291          */
2292         if (adapter->state == VNIC_REMOVING ||
2293             adapter->state == VNIC_REMOVED ||
2294             (adapter->failover_pending && reason != VNIC_RESET_FAILOVER)) {
2295                 ret = EBUSY;
2296                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2297                 goto err;
2298         }
2299
2300         if (adapter->state == VNIC_PROBING) {
2301                 netdev_warn(netdev, "Adapter reset during probe\n");
2302                 ret = adapter->init_done_rc = EAGAIN;
2303                 goto err;
2304         }
2305
2306         spin_lock_irqsave(&adapter->rwi_lock, flags);
2307
2308         list_for_each(entry, &adapter->rwi_list) {
2309                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2310                 if (tmp->reset_reason == reason) {
2311                         netdev_dbg(netdev, "Skipping matching reset\n");
2312                         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2313                         ret = EBUSY;
2314                         goto err;
2315                 }
2316         }
2317
2318         rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2319         if (!rwi) {
2320                 spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2321                 ibmvnic_close(netdev);
2322                 ret = ENOMEM;
2323                 goto err;
2324         }
2325         /* if we just received a transport event,
2326          * flush reset queue and process this reset
2327          */
2328         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2329                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2330                         list_del(entry);
2331         }
2332         rwi->reset_reason = reason;
2333         list_add_tail(&rwi->list, &adapter->rwi_list);
2334         spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2335         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2336         schedule_work(&adapter->ibmvnic_reset);
2337
2338         return 0;
2339 err:
2340         return -ret;
2341 }
2342
2343 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2344 {
2345         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2346
2347         if (test_bit(0, &adapter->resetting)) {
2348                 netdev_err(adapter->netdev,
2349                            "Adapter is resetting, skip timeout reset\n");
2350                 return;
2351         }
2352
2353         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2354 }
2355
2356 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2357                                   struct ibmvnic_rx_buff *rx_buff)
2358 {
2359         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2360
2361         rx_buff->skb = NULL;
2362
2363         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2364         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2365
2366         atomic_dec(&pool->available);
2367 }
2368
2369 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2370 {
2371         struct net_device *netdev = napi->dev;
2372         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2373         int scrq_num = (int)(napi - adapter->napi);
2374         int frames_processed = 0;
2375
2376 restart_poll:
2377         while (frames_processed < budget) {
2378                 struct sk_buff *skb;
2379                 struct ibmvnic_rx_buff *rx_buff;
2380                 union sub_crq *next;
2381                 u32 length;
2382                 u16 offset;
2383                 u8 flags = 0;
2384
2385                 if (unlikely(test_bit(0, &adapter->resetting) &&
2386                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2387                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2388                         napi_complete_done(napi, frames_processed);
2389                         return frames_processed;
2390                 }
2391
2392                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2393                         break;
2394                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2395                 rx_buff =
2396                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2397                                                           rx_comp.correlator);
2398                 /* do error checking */
2399                 if (next->rx_comp.rc) {
2400                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2401                                    be16_to_cpu(next->rx_comp.rc));
2402                         /* free the entry */
2403                         next->rx_comp.first = 0;
2404                         dev_kfree_skb_any(rx_buff->skb);
2405                         remove_buff_from_pool(adapter, rx_buff);
2406                         continue;
2407                 } else if (!rx_buff->skb) {
2408                         /* free the entry */
2409                         next->rx_comp.first = 0;
2410                         remove_buff_from_pool(adapter, rx_buff);
2411                         continue;
2412                 }
2413
2414                 length = be32_to_cpu(next->rx_comp.len);
2415                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2416                 flags = next->rx_comp.flags;
2417                 skb = rx_buff->skb;
2418                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2419                                         length);
2420
2421                 /* VLAN Header has been stripped by the system firmware and
2422                  * needs to be inserted by the driver
2423                  */
2424                 if (adapter->rx_vlan_header_insertion &&
2425                     (flags & IBMVNIC_VLAN_STRIPPED))
2426                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2427                                                ntohs(next->rx_comp.vlan_tci));
2428
2429                 /* free the entry */
2430                 next->rx_comp.first = 0;
2431                 remove_buff_from_pool(adapter, rx_buff);
2432
2433                 skb_put(skb, length);
2434                 skb->protocol = eth_type_trans(skb, netdev);
2435                 skb_record_rx_queue(skb, scrq_num);
2436
2437                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2438                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2439                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2440                 }
2441
2442                 length = skb->len;
2443                 napi_gro_receive(napi, skb); /* send it up */
2444                 netdev->stats.rx_packets++;
2445                 netdev->stats.rx_bytes += length;
2446                 adapter->rx_stats_buffers[scrq_num].packets++;
2447                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2448                 frames_processed++;
2449         }
2450
2451         if (adapter->state != VNIC_CLOSING)
2452                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2453
2454         if (frames_processed < budget) {
2455                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2456                 napi_complete_done(napi, frames_processed);
2457                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2458                     napi_reschedule(napi)) {
2459                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2460                         goto restart_poll;
2461                 }
2462         }
2463         return frames_processed;
2464 }
2465
2466 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2467 {
2468         int rc, ret;
2469
2470         adapter->fallback.mtu = adapter->req_mtu;
2471         adapter->fallback.rx_queues = adapter->req_rx_queues;
2472         adapter->fallback.tx_queues = adapter->req_tx_queues;
2473         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2474         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2475
2476         reinit_completion(&adapter->reset_done);
2477         adapter->wait_for_reset = true;
2478         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2479
2480         if (rc) {
2481                 ret = rc;
2482                 goto out;
2483         }
2484         rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2485         if (rc) {
2486                 ret = -ENODEV;
2487                 goto out;
2488         }
2489
2490         ret = 0;
2491         if (adapter->reset_done_rc) {
2492                 ret = -EIO;
2493                 adapter->desired.mtu = adapter->fallback.mtu;
2494                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2495                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2496                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2497                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2498
2499                 reinit_completion(&adapter->reset_done);
2500                 adapter->wait_for_reset = true;
2501                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2502                 if (rc) {
2503                         ret = rc;
2504                         goto out;
2505                 }
2506                 rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2507                                                  60000);
2508                 if (rc) {
2509                         ret = -ENODEV;
2510                         goto out;
2511                 }
2512         }
2513 out:
2514         adapter->wait_for_reset = false;
2515
2516         return ret;
2517 }
2518
2519 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2520 {
2521         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2522
2523         adapter->desired.mtu = new_mtu + ETH_HLEN;
2524
2525         return wait_for_reset(adapter);
2526 }
2527
2528 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2529                                                 struct net_device *dev,
2530                                                 netdev_features_t features)
2531 {
2532         /* Some backing hardware adapters can not
2533          * handle packets with a MSS less than 224
2534          * or with only one segment.
2535          */
2536         if (skb_is_gso(skb)) {
2537                 if (skb_shinfo(skb)->gso_size < 224 ||
2538                     skb_shinfo(skb)->gso_segs == 1)
2539                         features &= ~NETIF_F_GSO_MASK;
2540         }
2541
2542         return features;
2543 }
2544
2545 static const struct net_device_ops ibmvnic_netdev_ops = {
2546         .ndo_open               = ibmvnic_open,
2547         .ndo_stop               = ibmvnic_close,
2548         .ndo_start_xmit         = ibmvnic_xmit,
2549         .ndo_set_rx_mode        = ibmvnic_set_multi,
2550         .ndo_set_mac_address    = ibmvnic_set_mac,
2551         .ndo_validate_addr      = eth_validate_addr,
2552         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2553         .ndo_change_mtu         = ibmvnic_change_mtu,
2554         .ndo_features_check     = ibmvnic_features_check,
2555 };
2556
2557 /* ethtool functions */
2558
2559 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2560                                       struct ethtool_link_ksettings *cmd)
2561 {
2562         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2563         int rc;
2564
2565         rc = send_query_phys_parms(adapter);
2566         if (rc) {
2567                 adapter->speed = SPEED_UNKNOWN;
2568                 adapter->duplex = DUPLEX_UNKNOWN;
2569         }
2570         cmd->base.speed = adapter->speed;
2571         cmd->base.duplex = adapter->duplex;
2572         cmd->base.port = PORT_FIBRE;
2573         cmd->base.phy_address = 0;
2574         cmd->base.autoneg = AUTONEG_ENABLE;
2575
2576         return 0;
2577 }
2578
2579 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2580                                 struct ethtool_drvinfo *info)
2581 {
2582         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2583
2584         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2585         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2586         strlcpy(info->fw_version, adapter->fw_version,
2587                 sizeof(info->fw_version));
2588 }
2589
2590 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2591 {
2592         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2593
2594         return adapter->msg_enable;
2595 }
2596
2597 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2598 {
2599         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2600
2601         adapter->msg_enable = data;
2602 }
2603
2604 static u32 ibmvnic_get_link(struct net_device *netdev)
2605 {
2606         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2607
2608         /* Don't need to send a query because we request a logical link up at
2609          * init and then we wait for link state indications
2610          */
2611         return adapter->logical_link_state;
2612 }
2613
2614 static void ibmvnic_get_ringparam(struct net_device *netdev,
2615                                   struct ethtool_ringparam *ring)
2616 {
2617         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2618
2619         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2620                 ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2621                 ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2622         } else {
2623                 ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2624                 ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2625         }
2626         ring->rx_mini_max_pending = 0;
2627         ring->rx_jumbo_max_pending = 0;
2628         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2629         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2630         ring->rx_mini_pending = 0;
2631         ring->rx_jumbo_pending = 0;
2632 }
2633
2634 static int ibmvnic_set_ringparam(struct net_device *netdev,
2635                                  struct ethtool_ringparam *ring)
2636 {
2637         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2638         int ret;
2639
2640         ret = 0;
2641         adapter->desired.rx_entries = ring->rx_pending;
2642         adapter->desired.tx_entries = ring->tx_pending;
2643
2644         ret = wait_for_reset(adapter);
2645
2646         if (!ret &&
2647             (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2648              adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2649                 netdev_info(netdev,
2650                             "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2651                             ring->rx_pending, ring->tx_pending,
2652                             adapter->req_rx_add_entries_per_subcrq,
2653                             adapter->req_tx_entries_per_subcrq);
2654         return ret;
2655 }
2656
2657 static void ibmvnic_get_channels(struct net_device *netdev,
2658                                  struct ethtool_channels *channels)
2659 {
2660         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2661
2662         if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2663                 channels->max_rx = adapter->max_rx_queues;
2664                 channels->max_tx = adapter->max_tx_queues;
2665         } else {
2666                 channels->max_rx = IBMVNIC_MAX_QUEUES;
2667                 channels->max_tx = IBMVNIC_MAX_QUEUES;
2668         }
2669
2670         channels->max_other = 0;
2671         channels->max_combined = 0;
2672         channels->rx_count = adapter->req_rx_queues;
2673         channels->tx_count = adapter->req_tx_queues;
2674         channels->other_count = 0;
2675         channels->combined_count = 0;
2676 }
2677
2678 static int ibmvnic_set_channels(struct net_device *netdev,
2679                                 struct ethtool_channels *channels)
2680 {
2681         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2682         int ret;
2683
2684         ret = 0;
2685         adapter->desired.rx_queues = channels->rx_count;
2686         adapter->desired.tx_queues = channels->tx_count;
2687
2688         ret = wait_for_reset(adapter);
2689
2690         if (!ret &&
2691             (adapter->req_rx_queues != channels->rx_count ||
2692              adapter->req_tx_queues != channels->tx_count))
2693                 netdev_info(netdev,
2694                             "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2695                             channels->rx_count, channels->tx_count,
2696                             adapter->req_rx_queues, adapter->req_tx_queues);
2697         return ret;
2698
2699 }
2700
2701 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2702 {
2703         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2704         int i;
2705
2706         switch (stringset) {
2707         case ETH_SS_STATS:
2708                 for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2709                                 i++, data += ETH_GSTRING_LEN)
2710                         memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2711
2712                 for (i = 0; i < adapter->req_tx_queues; i++) {
2713                         snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2714                         data += ETH_GSTRING_LEN;
2715
2716                         snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2717                         data += ETH_GSTRING_LEN;
2718
2719                         snprintf(data, ETH_GSTRING_LEN,
2720                                  "tx%d_dropped_packets", i);
2721                         data += ETH_GSTRING_LEN;
2722                 }
2723
2724                 for (i = 0; i < adapter->req_rx_queues; i++) {
2725                         snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2726                         data += ETH_GSTRING_LEN;
2727
2728                         snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2729                         data += ETH_GSTRING_LEN;
2730
2731                         snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2732                         data += ETH_GSTRING_LEN;
2733                 }
2734                 break;
2735
2736         case ETH_SS_PRIV_FLAGS:
2737                 for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2738                         strcpy(data + i * ETH_GSTRING_LEN,
2739                                ibmvnic_priv_flags[i]);
2740                 break;
2741         default:
2742                 return;
2743         }
2744 }
2745
2746 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2747 {
2748         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2749
2750         switch (sset) {
2751         case ETH_SS_STATS:
2752                 return ARRAY_SIZE(ibmvnic_stats) +
2753                        adapter->req_tx_queues * NUM_TX_STATS +
2754                        adapter->req_rx_queues * NUM_RX_STATS;
2755         case ETH_SS_PRIV_FLAGS:
2756                 return ARRAY_SIZE(ibmvnic_priv_flags);
2757         default:
2758                 return -EOPNOTSUPP;
2759         }
2760 }
2761
2762 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2763                                       struct ethtool_stats *stats, u64 *data)
2764 {
2765         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2766         union ibmvnic_crq crq;
2767         int i, j;
2768         int rc;
2769
2770         memset(&crq, 0, sizeof(crq));
2771         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2772         crq.request_statistics.cmd = REQUEST_STATISTICS;
2773         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2774         crq.request_statistics.len =
2775             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2776
2777         /* Wait for data to be written */
2778         reinit_completion(&adapter->stats_done);
2779         rc = ibmvnic_send_crq(adapter, &crq);
2780         if (rc)
2781                 return;
2782         rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2783         if (rc)
2784                 return;
2785
2786         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2787                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2788                                                 ibmvnic_stats[i].offset));
2789
2790         for (j = 0; j < adapter->req_tx_queues; j++) {
2791                 data[i] = adapter->tx_stats_buffers[j].packets;
2792                 i++;
2793                 data[i] = adapter->tx_stats_buffers[j].bytes;
2794                 i++;
2795                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2796                 i++;
2797         }
2798
2799         for (j = 0; j < adapter->req_rx_queues; j++) {
2800                 data[i] = adapter->rx_stats_buffers[j].packets;
2801                 i++;
2802                 data[i] = adapter->rx_stats_buffers[j].bytes;
2803                 i++;
2804                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2805                 i++;
2806         }
2807 }
2808
2809 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2810 {
2811         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2812
2813         return adapter->priv_flags;
2814 }
2815
2816 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2817 {
2818         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2819         bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2820
2821         if (which_maxes)
2822                 adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2823         else
2824                 adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2825
2826         return 0;
2827 }
2828 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2829         .get_drvinfo            = ibmvnic_get_drvinfo,
2830         .get_msglevel           = ibmvnic_get_msglevel,
2831         .set_msglevel           = ibmvnic_set_msglevel,
2832         .get_link               = ibmvnic_get_link,
2833         .get_ringparam          = ibmvnic_get_ringparam,
2834         .set_ringparam          = ibmvnic_set_ringparam,
2835         .get_channels           = ibmvnic_get_channels,
2836         .set_channels           = ibmvnic_set_channels,
2837         .get_strings            = ibmvnic_get_strings,
2838         .get_sset_count         = ibmvnic_get_sset_count,
2839         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2840         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2841         .get_priv_flags         = ibmvnic_get_priv_flags,
2842         .set_priv_flags         = ibmvnic_set_priv_flags,
2843 };
2844
2845 /* Routines for managing CRQs/sCRQs  */
2846
2847 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2848                                    struct ibmvnic_sub_crq_queue *scrq)
2849 {
2850         int rc;
2851
2852         if (!scrq) {
2853                 netdev_dbg(adapter->netdev,
2854                            "Invalid scrq reset. irq (%d) or msgs (%p).\n",
2855                            scrq->irq, scrq->msgs);
2856                 return -EINVAL;
2857         }
2858
2859         if (scrq->irq) {
2860                 free_irq(scrq->irq, scrq);
2861                 irq_dispose_mapping(scrq->irq);
2862                 scrq->irq = 0;
2863         }
2864         if (scrq->msgs) {
2865                 memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2866                 atomic_set(&scrq->used, 0);
2867                 scrq->cur = 0;
2868         } else {
2869                 netdev_dbg(adapter->netdev, "Invalid scrq reset\n");
2870                 return -EINVAL;
2871         }
2872
2873         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2874                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2875         return rc;
2876 }
2877
2878 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2879 {
2880         int i, rc;
2881
2882         if (!adapter->tx_scrq || !adapter->rx_scrq)
2883                 return -EINVAL;
2884
2885         for (i = 0; i < adapter->req_tx_queues; i++) {
2886                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2887                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2888                 if (rc)
2889                         return rc;
2890         }
2891
2892         for (i = 0; i < adapter->req_rx_queues; i++) {
2893                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2894                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2895                 if (rc)
2896                         return rc;
2897         }
2898
2899         return rc;
2900 }
2901
2902 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2903                                   struct ibmvnic_sub_crq_queue *scrq,
2904                                   bool do_h_free)
2905 {
2906         struct device *dev = &adapter->vdev->dev;
2907         long rc;
2908
2909         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2910
2911         if (do_h_free) {
2912                 /* Close the sub-crqs */
2913                 do {
2914                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2915                                                 adapter->vdev->unit_address,
2916                                                 scrq->crq_num);
2917                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2918
2919                 if (rc) {
2920                         netdev_err(adapter->netdev,
2921                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2922                                    scrq->crq_num, rc);
2923                 }
2924         }
2925
2926         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2927                          DMA_BIDIRECTIONAL);
2928         free_pages((unsigned long)scrq->msgs, 2);
2929         kfree(scrq);
2930 }
2931
2932 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2933                                                         *adapter)
2934 {
2935         struct device *dev = &adapter->vdev->dev;
2936         struct ibmvnic_sub_crq_queue *scrq;
2937         int rc;
2938
2939         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2940         if (!scrq)
2941                 return NULL;
2942
2943         scrq->msgs =
2944                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2945         if (!scrq->msgs) {
2946                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2947                 goto zero_page_failed;
2948         }
2949
2950         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2951                                          DMA_BIDIRECTIONAL);
2952         if (dma_mapping_error(dev, scrq->msg_token)) {
2953                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2954                 goto map_failed;
2955         }
2956
2957         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2958                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2959
2960         if (rc == H_RESOURCE)
2961                 rc = ibmvnic_reset_crq(adapter);
2962
2963         if (rc == H_CLOSED) {
2964                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2965         } else if (rc) {
2966                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2967                 goto reg_failed;
2968         }
2969
2970         scrq->adapter = adapter;
2971         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2972         spin_lock_init(&scrq->lock);
2973
2974         netdev_dbg(adapter->netdev,
2975                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2976                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2977
2978         return scrq;
2979
2980 reg_failed:
2981         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2982                          DMA_BIDIRECTIONAL);
2983 map_failed:
2984         free_pages((unsigned long)scrq->msgs, 2);
2985 zero_page_failed:
2986         kfree(scrq);
2987
2988         return NULL;
2989 }
2990
2991 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2992 {
2993         int i;
2994
2995         if (adapter->tx_scrq) {
2996                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2997                         if (!adapter->tx_scrq[i])
2998                                 continue;
2999
3000                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
3001                                    i);
3002                         if (adapter->tx_scrq[i]->irq) {
3003                                 free_irq(adapter->tx_scrq[i]->irq,
3004                                          adapter->tx_scrq[i]);
3005                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
3006                                 adapter->tx_scrq[i]->irq = 0;
3007                         }
3008
3009                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
3010                                               do_h_free);
3011                 }
3012
3013                 kfree(adapter->tx_scrq);
3014                 adapter->tx_scrq = NULL;
3015                 adapter->num_active_tx_scrqs = 0;
3016         }
3017
3018         if (adapter->rx_scrq) {
3019                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
3020                         if (!adapter->rx_scrq[i])
3021                                 continue;
3022
3023                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
3024                                    i);
3025                         if (adapter->rx_scrq[i]->irq) {
3026                                 free_irq(adapter->rx_scrq[i]->irq,
3027                                          adapter->rx_scrq[i]);
3028                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
3029                                 adapter->rx_scrq[i]->irq = 0;
3030                         }
3031
3032                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
3033                                               do_h_free);
3034                 }
3035
3036                 kfree(adapter->rx_scrq);
3037                 adapter->rx_scrq = NULL;
3038                 adapter->num_active_rx_scrqs = 0;
3039         }
3040 }
3041
3042 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
3043                             struct ibmvnic_sub_crq_queue *scrq)
3044 {
3045         struct device *dev = &adapter->vdev->dev;
3046         unsigned long rc;
3047
3048         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3049                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3050         if (rc)
3051                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
3052                         scrq->hw_irq, rc);
3053         return rc;
3054 }
3055
3056 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3057                            struct ibmvnic_sub_crq_queue *scrq)
3058 {
3059         struct device *dev = &adapter->vdev->dev;
3060         unsigned long rc;
3061
3062         if (scrq->hw_irq > 0x100000000ULL) {
3063                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3064                 return 1;
3065         }
3066
3067         if (test_bit(0, &adapter->resetting) &&
3068             adapter->reset_reason == VNIC_RESET_MOBILITY) {
3069                 u64 val = (0xff000000) | scrq->hw_irq;
3070
3071                 rc = plpar_hcall_norets(H_EOI, val);
3072                 /* H_EOI would fail with rc = H_FUNCTION when running
3073                  * in XIVE mode which is expected, but not an error.
3074                  */
3075                 if (rc && (rc != H_FUNCTION))
3076                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3077                                 val, rc);
3078         }
3079
3080         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3081                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3082         if (rc)
3083                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3084                         scrq->hw_irq, rc);
3085         return rc;
3086 }
3087
3088 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3089                                struct ibmvnic_sub_crq_queue *scrq)
3090 {
3091         struct device *dev = &adapter->vdev->dev;
3092         struct ibmvnic_tx_pool *tx_pool;
3093         struct ibmvnic_tx_buff *txbuff;
3094         union sub_crq *next;
3095         int index;
3096         int i, j;
3097
3098 restart_loop:
3099         while (pending_scrq(adapter, scrq)) {
3100                 unsigned int pool = scrq->pool_index;
3101                 int num_entries = 0;
3102
3103                 next = ibmvnic_next_scrq(adapter, scrq);
3104                 for (i = 0; i < next->tx_comp.num_comps; i++) {
3105                         if (next->tx_comp.rcs[i]) {
3106                                 dev_err(dev, "tx error %x\n",
3107                                         next->tx_comp.rcs[i]);
3108                                 continue;
3109                         }
3110                         index = be32_to_cpu(next->tx_comp.correlators[i]);
3111                         if (index & IBMVNIC_TSO_POOL_MASK) {
3112                                 tx_pool = &adapter->tso_pool[pool];
3113                                 index &= ~IBMVNIC_TSO_POOL_MASK;
3114                         } else {
3115                                 tx_pool = &adapter->tx_pool[pool];
3116                         }
3117
3118                         txbuff = &tx_pool->tx_buff[index];
3119
3120                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3121                                 if (!txbuff->data_dma[j])
3122                                         continue;
3123
3124                                 txbuff->data_dma[j] = 0;
3125                         }
3126
3127                         if (txbuff->last_frag) {
3128                                 dev_kfree_skb_any(txbuff->skb);
3129                                 txbuff->skb = NULL;
3130                         }
3131
3132                         num_entries += txbuff->num_entries;
3133
3134                         tx_pool->free_map[tx_pool->producer_index] = index;
3135                         tx_pool->producer_index =
3136                                 (tx_pool->producer_index + 1) %
3137                                         tx_pool->num_buffers;
3138                 }
3139                 /* remove tx_comp scrq*/
3140                 next->tx_comp.first = 0;
3141
3142                 if (atomic_sub_return(num_entries, &scrq->used) <=
3143                     (adapter->req_tx_entries_per_subcrq / 2) &&
3144                     __netif_subqueue_stopped(adapter->netdev,
3145                                              scrq->pool_index)) {
3146                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3147                         netdev_dbg(adapter->netdev, "Started queue %d\n",
3148                                    scrq->pool_index);
3149                 }
3150         }
3151
3152         enable_scrq_irq(adapter, scrq);
3153
3154         if (pending_scrq(adapter, scrq)) {
3155                 disable_scrq_irq(adapter, scrq);
3156                 goto restart_loop;
3157         }
3158
3159         return 0;
3160 }
3161
3162 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3163 {
3164         struct ibmvnic_sub_crq_queue *scrq = instance;
3165         struct ibmvnic_adapter *adapter = scrq->adapter;
3166
3167         disable_scrq_irq(adapter, scrq);
3168         ibmvnic_complete_tx(adapter, scrq);
3169
3170         return IRQ_HANDLED;
3171 }
3172
3173 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3174 {
3175         struct ibmvnic_sub_crq_queue *scrq = instance;
3176         struct ibmvnic_adapter *adapter = scrq->adapter;
3177
3178         /* When booting a kdump kernel we can hit pending interrupts
3179          * prior to completing driver initialization.
3180          */
3181         if (unlikely(adapter->state != VNIC_OPEN))
3182                 return IRQ_NONE;
3183
3184         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3185
3186         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3187                 disable_scrq_irq(adapter, scrq);
3188                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
3189         }
3190
3191         return IRQ_HANDLED;
3192 }
3193
3194 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3195 {
3196         struct device *dev = &adapter->vdev->dev;
3197         struct ibmvnic_sub_crq_queue *scrq;
3198         int i = 0, j = 0;
3199         int rc = 0;
3200
3201         for (i = 0; i < adapter->req_tx_queues; i++) {
3202                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3203                            i);
3204                 scrq = adapter->tx_scrq[i];
3205                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3206
3207                 if (!scrq->irq) {
3208                         rc = -EINVAL;
3209                         dev_err(dev, "Error mapping irq\n");
3210                         goto req_tx_irq_failed;
3211                 }
3212
3213                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3214                          adapter->vdev->unit_address, i);
3215                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3216                                  0, scrq->name, scrq);
3217
3218                 if (rc) {
3219                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3220                                 scrq->irq, rc);
3221                         irq_dispose_mapping(scrq->irq);
3222                         goto req_tx_irq_failed;
3223                 }
3224         }
3225
3226         for (i = 0; i < adapter->req_rx_queues; i++) {
3227                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3228                            i);
3229                 scrq = adapter->rx_scrq[i];
3230                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3231                 if (!scrq->irq) {
3232                         rc = -EINVAL;
3233                         dev_err(dev, "Error mapping irq\n");
3234                         goto req_rx_irq_failed;
3235                 }
3236                 snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3237                          adapter->vdev->unit_address, i);
3238                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3239                                  0, scrq->name, scrq);
3240                 if (rc) {
3241                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3242                                 scrq->irq, rc);
3243                         irq_dispose_mapping(scrq->irq);
3244                         goto req_rx_irq_failed;
3245                 }
3246         }
3247         return rc;
3248
3249 req_rx_irq_failed:
3250         for (j = 0; j < i; j++) {
3251                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3252                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3253         }
3254         i = adapter->req_tx_queues;
3255 req_tx_irq_failed:
3256         for (j = 0; j < i; j++) {
3257                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3258                 irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3259         }
3260         release_sub_crqs(adapter, 1);
3261         return rc;
3262 }
3263
3264 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3265 {
3266         struct device *dev = &adapter->vdev->dev;
3267         struct ibmvnic_sub_crq_queue **allqueues;
3268         int registered_queues = 0;
3269         int total_queues;
3270         int more = 0;
3271         int i;
3272
3273         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3274
3275         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3276         if (!allqueues)
3277                 return -1;
3278
3279         for (i = 0; i < total_queues; i++) {
3280                 allqueues[i] = init_sub_crq_queue(adapter);
3281                 if (!allqueues[i]) {
3282                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3283                         break;
3284                 }
3285                 registered_queues++;
3286         }
3287
3288         /* Make sure we were able to register the minimum number of queues */
3289         if (registered_queues <
3290             adapter->min_tx_queues + adapter->min_rx_queues) {
3291                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3292                 goto tx_failed;
3293         }
3294
3295         /* Distribute the failed allocated queues*/
3296         for (i = 0; i < total_queues - registered_queues + more ; i++) {
3297                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3298                 switch (i % 3) {
3299                 case 0:
3300                         if (adapter->req_rx_queues > adapter->min_rx_queues)
3301                                 adapter->req_rx_queues--;
3302                         else
3303                                 more++;
3304                         break;
3305                 case 1:
3306                         if (adapter->req_tx_queues > adapter->min_tx_queues)
3307                                 adapter->req_tx_queues--;
3308                         else
3309                                 more++;
3310                         break;
3311                 }
3312         }
3313
3314         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3315                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
3316         if (!adapter->tx_scrq)
3317                 goto tx_failed;
3318
3319         for (i = 0; i < adapter->req_tx_queues; i++) {
3320                 adapter->tx_scrq[i] = allqueues[i];
3321                 adapter->tx_scrq[i]->pool_index = i;
3322                 adapter->num_active_tx_scrqs++;
3323         }
3324
3325         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3326                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3327         if (!adapter->rx_scrq)
3328                 goto rx_failed;
3329
3330         for (i = 0; i < adapter->req_rx_queues; i++) {
3331                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3332                 adapter->rx_scrq[i]->scrq_num = i;
3333                 adapter->num_active_rx_scrqs++;
3334         }
3335
3336         kfree(allqueues);
3337         return 0;
3338
3339 rx_failed:
3340         kfree(adapter->tx_scrq);
3341         adapter->tx_scrq = NULL;
3342 tx_failed:
3343         for (i = 0; i < registered_queues; i++)
3344                 release_sub_crq_queue(adapter, allqueues[i], 1);
3345         kfree(allqueues);
3346         return -1;
3347 }
3348
3349 static void send_request_cap(struct ibmvnic_adapter *adapter, int retry)
3350 {
3351         struct device *dev = &adapter->vdev->dev;
3352         union ibmvnic_crq crq;
3353         int max_entries;
3354
3355         if (!retry) {
3356                 /* Sub-CRQ entries are 32 byte long */
3357                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3358
3359                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3360                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3361                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3362                         return;
3363                 }
3364
3365                 if (adapter->desired.mtu)
3366                         adapter->req_mtu = adapter->desired.mtu;
3367                 else
3368                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3369
3370                 if (!adapter->desired.tx_entries)
3371                         adapter->desired.tx_entries =
3372                                         adapter->max_tx_entries_per_subcrq;
3373                 if (!adapter->desired.rx_entries)
3374                         adapter->desired.rx_entries =
3375                                         adapter->max_rx_add_entries_per_subcrq;
3376
3377                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3378                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3379
3380                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3381                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3382                         adapter->desired.tx_entries = max_entries;
3383                 }
3384
3385                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3386                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3387                         adapter->desired.rx_entries = max_entries;
3388                 }
3389
3390                 if (adapter->desired.tx_entries)
3391                         adapter->req_tx_entries_per_subcrq =
3392                                         adapter->desired.tx_entries;
3393                 else
3394                         adapter->req_tx_entries_per_subcrq =
3395                                         adapter->max_tx_entries_per_subcrq;
3396
3397                 if (adapter->desired.rx_entries)
3398                         adapter->req_rx_add_entries_per_subcrq =
3399                                         adapter->desired.rx_entries;
3400                 else
3401                         adapter->req_rx_add_entries_per_subcrq =
3402                                         adapter->max_rx_add_entries_per_subcrq;
3403
3404                 if (adapter->desired.tx_queues)
3405                         adapter->req_tx_queues =
3406                                         adapter->desired.tx_queues;
3407                 else
3408                         adapter->req_tx_queues =
3409                                         adapter->opt_tx_comp_sub_queues;
3410
3411                 if (adapter->desired.rx_queues)
3412                         adapter->req_rx_queues =
3413                                         adapter->desired.rx_queues;
3414                 else
3415                         adapter->req_rx_queues =
3416                                         adapter->opt_rx_comp_queues;
3417
3418                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3419         }
3420
3421         memset(&crq, 0, sizeof(crq));
3422         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3423         crq.request_capability.cmd = REQUEST_CAPABILITY;
3424
3425         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3426         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3427         atomic_inc(&adapter->running_cap_crqs);
3428         ibmvnic_send_crq(adapter, &crq);
3429
3430         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3431         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3432         atomic_inc(&adapter->running_cap_crqs);
3433         ibmvnic_send_crq(adapter, &crq);
3434
3435         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3436         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3437         atomic_inc(&adapter->running_cap_crqs);
3438         ibmvnic_send_crq(adapter, &crq);
3439
3440         crq.request_capability.capability =
3441             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3442         crq.request_capability.number =
3443             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3444         atomic_inc(&adapter->running_cap_crqs);
3445         ibmvnic_send_crq(adapter, &crq);
3446
3447         crq.request_capability.capability =
3448             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3449         crq.request_capability.number =
3450             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3451         atomic_inc(&adapter->running_cap_crqs);
3452         ibmvnic_send_crq(adapter, &crq);
3453
3454         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3455         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3456         atomic_inc(&adapter->running_cap_crqs);
3457         ibmvnic_send_crq(adapter, &crq);
3458
3459         if (adapter->netdev->flags & IFF_PROMISC) {
3460                 if (adapter->promisc_supported) {
3461                         crq.request_capability.capability =
3462                             cpu_to_be16(PROMISC_REQUESTED);
3463                         crq.request_capability.number = cpu_to_be64(1);
3464                         atomic_inc(&adapter->running_cap_crqs);
3465                         ibmvnic_send_crq(adapter, &crq);
3466                 }
3467         } else {
3468                 crq.request_capability.capability =
3469                     cpu_to_be16(PROMISC_REQUESTED);
3470                 crq.request_capability.number = cpu_to_be64(0);
3471                 atomic_inc(&adapter->running_cap_crqs);
3472                 ibmvnic_send_crq(adapter, &crq);
3473         }
3474 }
3475
3476 static int pending_scrq(struct ibmvnic_adapter *adapter,
3477                         struct ibmvnic_sub_crq_queue *scrq)
3478 {
3479         union sub_crq *entry = &scrq->msgs[scrq->cur];
3480
3481         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3482                 return 1;
3483         else
3484                 return 0;
3485 }
3486
3487 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3488                                         struct ibmvnic_sub_crq_queue *scrq)
3489 {
3490         union sub_crq *entry;
3491         unsigned long flags;
3492
3493         spin_lock_irqsave(&scrq->lock, flags);
3494         entry = &scrq->msgs[scrq->cur];
3495         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3496                 if (++scrq->cur == scrq->size)
3497                         scrq->cur = 0;
3498         } else {
3499                 entry = NULL;
3500         }
3501         spin_unlock_irqrestore(&scrq->lock, flags);
3502
3503         return entry;
3504 }
3505
3506 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3507 {
3508         struct ibmvnic_crq_queue *queue = &adapter->crq;
3509         union ibmvnic_crq *crq;
3510
3511         crq = &queue->msgs[queue->cur];
3512         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3513                 if (++queue->cur == queue->size)
3514                         queue->cur = 0;
3515         } else {
3516                 crq = NULL;
3517         }
3518
3519         return crq;
3520 }
3521
3522 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3523 {
3524         switch (rc) {
3525         case H_PARAMETER:
3526                 dev_warn_ratelimited(dev,
3527                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3528                                      func, rc);
3529                 break;
3530         case H_CLOSED:
3531                 dev_warn_ratelimited(dev,
3532                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3533                                      func, rc);
3534                 break;
3535         default:
3536                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3537                 break;
3538         }
3539 }
3540
3541 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3542                        union sub_crq *sub_crq)
3543 {
3544         unsigned int ua = adapter->vdev->unit_address;
3545         struct device *dev = &adapter->vdev->dev;
3546         u64 *u64_crq = (u64 *)sub_crq;
3547         int rc;
3548
3549         netdev_dbg(adapter->netdev,
3550                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3551                    (unsigned long int)cpu_to_be64(remote_handle),
3552                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3553                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3554                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3555                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3556
3557         /* Make sure the hypervisor sees the complete request */
3558         mb();
3559
3560         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3561                                 cpu_to_be64(remote_handle),
3562                                 cpu_to_be64(u64_crq[0]),
3563                                 cpu_to_be64(u64_crq[1]),
3564                                 cpu_to_be64(u64_crq[2]),
3565                                 cpu_to_be64(u64_crq[3]));
3566
3567         if (rc)
3568                 print_subcrq_error(dev, rc, __func__);
3569
3570         return rc;
3571 }
3572
3573 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3574                                 u64 remote_handle, u64 ioba, u64 num_entries)
3575 {
3576         unsigned int ua = adapter->vdev->unit_address;
3577         struct device *dev = &adapter->vdev->dev;
3578         int rc;
3579
3580         /* Make sure the hypervisor sees the complete request */
3581         mb();
3582         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3583                                 cpu_to_be64(remote_handle),
3584                                 ioba, num_entries);
3585
3586         if (rc)
3587                 print_subcrq_error(dev, rc, __func__);
3588
3589         return rc;
3590 }
3591
3592 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3593                             union ibmvnic_crq *crq)
3594 {
3595         unsigned int ua = adapter->vdev->unit_address;
3596         struct device *dev = &adapter->vdev->dev;
3597         u64 *u64_crq = (u64 *)crq;
3598         int rc;
3599
3600         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3601                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3602                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3603
3604         if (!adapter->crq.active &&
3605             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3606                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3607                 return -EINVAL;
3608         }
3609
3610         /* Make sure the hypervisor sees the complete request */
3611         mb();
3612
3613         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3614                                 cpu_to_be64(u64_crq[0]),
3615                                 cpu_to_be64(u64_crq[1]));
3616
3617         if (rc) {
3618                 if (rc == H_CLOSED) {
3619                         dev_warn(dev, "CRQ Queue closed\n");
3620                         /* do not reset, report the fail, wait for passive init from server */
3621                 }
3622
3623                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3624         }
3625
3626         return rc;
3627 }
3628
3629 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3630 {
3631         struct device *dev = &adapter->vdev->dev;
3632         union ibmvnic_crq crq;
3633         int retries = 100;
3634         int rc;
3635
3636         memset(&crq, 0, sizeof(crq));
3637         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3638         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3639         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3640
3641         do {
3642                 rc = ibmvnic_send_crq(adapter, &crq);
3643                 if (rc != H_CLOSED)
3644                         break;
3645                 retries--;
3646                 msleep(50);
3647
3648         } while (retries > 0);
3649
3650         if (rc) {
3651                 dev_err(dev, "Failed to send init request, rc = %d\n", rc);
3652                 return rc;
3653         }
3654
3655         return 0;
3656 }
3657
3658 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3659 {
3660         union ibmvnic_crq crq;
3661
3662         memset(&crq, 0, sizeof(crq));
3663         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3664         crq.version_exchange.cmd = VERSION_EXCHANGE;
3665         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3666
3667         return ibmvnic_send_crq(adapter, &crq);
3668 }
3669
3670 struct vnic_login_client_data {
3671         u8      type;
3672         __be16  len;
3673         char    name[];
3674 } __packed;
3675
3676 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3677 {
3678         int len;
3679
3680         /* Calculate the amount of buffer space needed for the
3681          * vnic client data in the login buffer. There are four entries,
3682          * OS name, LPAR name, device name, and a null last entry.
3683          */
3684         len = 4 * sizeof(struct vnic_login_client_data);
3685         len += 6; /* "Linux" plus NULL */
3686         len += strlen(utsname()->nodename) + 1;
3687         len += strlen(adapter->netdev->name) + 1;
3688
3689         return len;
3690 }
3691
3692 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3693                                  struct vnic_login_client_data *vlcd)
3694 {
3695         const char *os_name = "Linux";
3696         int len;
3697
3698         /* Type 1 - LPAR OS */
3699         vlcd->type = 1;
3700         len = strlen(os_name) + 1;
3701         vlcd->len = cpu_to_be16(len);
3702         strncpy(vlcd->name, os_name, len);
3703         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3704
3705         /* Type 2 - LPAR name */
3706         vlcd->type = 2;
3707         len = strlen(utsname()->nodename) + 1;
3708         vlcd->len = cpu_to_be16(len);
3709         strncpy(vlcd->name, utsname()->nodename, len);
3710         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3711
3712         /* Type 3 - device name */
3713         vlcd->type = 3;
3714         len = strlen(adapter->netdev->name) + 1;
3715         vlcd->len = cpu_to_be16(len);
3716         strncpy(vlcd->name, adapter->netdev->name, len);
3717 }
3718
3719 static int send_login(struct ibmvnic_adapter *adapter)
3720 {
3721         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3722         struct ibmvnic_login_buffer *login_buffer;
3723         struct device *dev = &adapter->vdev->dev;
3724         dma_addr_t rsp_buffer_token;
3725         dma_addr_t buffer_token;
3726         size_t rsp_buffer_size;
3727         union ibmvnic_crq crq;
3728         size_t buffer_size;
3729         __be64 *tx_list_p;
3730         __be64 *rx_list_p;
3731         int client_data_len;
3732         struct vnic_login_client_data *vlcd;
3733         int i;
3734
3735         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3736                 netdev_err(adapter->netdev,
3737                            "RX or TX queues are not allocated, device login failed\n");
3738                 return -1;
3739         }
3740
3741         release_login_rsp_buffer(adapter);
3742         client_data_len = vnic_client_data_len(adapter);
3743
3744         buffer_size =
3745             sizeof(struct ibmvnic_login_buffer) +
3746             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3747             client_data_len;
3748
3749         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3750         if (!login_buffer)
3751                 goto buf_alloc_failed;
3752
3753         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3754                                       DMA_TO_DEVICE);
3755         if (dma_mapping_error(dev, buffer_token)) {
3756                 dev_err(dev, "Couldn't map login buffer\n");
3757                 goto buf_map_failed;
3758         }
3759
3760         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3761                           sizeof(u64) * adapter->req_tx_queues +
3762                           sizeof(u64) * adapter->req_rx_queues +
3763                           sizeof(u64) * adapter->req_rx_queues +
3764                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3765
3766         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3767         if (!login_rsp_buffer)
3768                 goto buf_rsp_alloc_failed;
3769
3770         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3771                                           rsp_buffer_size, DMA_FROM_DEVICE);
3772         if (dma_mapping_error(dev, rsp_buffer_token)) {
3773                 dev_err(dev, "Couldn't map login rsp buffer\n");
3774                 goto buf_rsp_map_failed;
3775         }
3776
3777         adapter->login_buf = login_buffer;
3778         adapter->login_buf_token = buffer_token;
3779         adapter->login_buf_sz = buffer_size;
3780         adapter->login_rsp_buf = login_rsp_buffer;
3781         adapter->login_rsp_buf_token = rsp_buffer_token;
3782         adapter->login_rsp_buf_sz = rsp_buffer_size;
3783
3784         login_buffer->len = cpu_to_be32(buffer_size);
3785         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3786         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3787         login_buffer->off_txcomp_subcrqs =
3788             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3789         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3790         login_buffer->off_rxcomp_subcrqs =
3791             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3792                         sizeof(u64) * adapter->req_tx_queues);
3793         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3794         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3795
3796         tx_list_p = (__be64 *)((char *)login_buffer +
3797                                       sizeof(struct ibmvnic_login_buffer));
3798         rx_list_p = (__be64 *)((char *)login_buffer +
3799                                       sizeof(struct ibmvnic_login_buffer) +
3800                                       sizeof(u64) * adapter->req_tx_queues);
3801
3802         for (i = 0; i < adapter->req_tx_queues; i++) {
3803                 if (adapter->tx_scrq[i]) {
3804                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3805                                                    crq_num);
3806                 }
3807         }
3808
3809         for (i = 0; i < adapter->req_rx_queues; i++) {
3810                 if (adapter->rx_scrq[i]) {
3811                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3812                                                    crq_num);
3813                 }
3814         }
3815
3816         /* Insert vNIC login client data */
3817         vlcd = (struct vnic_login_client_data *)
3818                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3819         login_buffer->client_data_offset =
3820                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3821         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3822
3823         vnic_add_client_data(adapter, vlcd);
3824
3825         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3826         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3827                 netdev_dbg(adapter->netdev, "%016lx\n",
3828                            ((unsigned long int *)(adapter->login_buf))[i]);
3829         }
3830
3831         memset(&crq, 0, sizeof(crq));
3832         crq.login.first = IBMVNIC_CRQ_CMD;
3833         crq.login.cmd = LOGIN;
3834         crq.login.ioba = cpu_to_be32(buffer_token);
3835         crq.login.len = cpu_to_be32(buffer_size);
3836         ibmvnic_send_crq(adapter, &crq);
3837
3838         return 0;
3839
3840 buf_rsp_map_failed:
3841         kfree(login_rsp_buffer);
3842 buf_rsp_alloc_failed:
3843         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3844 buf_map_failed:
3845         kfree(login_buffer);
3846 buf_alloc_failed:
3847         return -1;
3848 }
3849
3850 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3851                             u32 len, u8 map_id)
3852 {
3853         union ibmvnic_crq crq;
3854
3855         memset(&crq, 0, sizeof(crq));
3856         crq.request_map.first = IBMVNIC_CRQ_CMD;
3857         crq.request_map.cmd = REQUEST_MAP;
3858         crq.request_map.map_id = map_id;
3859         crq.request_map.ioba = cpu_to_be32(addr);
3860         crq.request_map.len = cpu_to_be32(len);
3861         return ibmvnic_send_crq(adapter, &crq);
3862 }
3863
3864 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3865 {
3866         union ibmvnic_crq crq;
3867
3868         memset(&crq, 0, sizeof(crq));
3869         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3870         crq.request_unmap.cmd = REQUEST_UNMAP;
3871         crq.request_unmap.map_id = map_id;
3872         return ibmvnic_send_crq(adapter, &crq);
3873 }
3874
3875 static void send_query_map(struct ibmvnic_adapter *adapter)
3876 {
3877         union ibmvnic_crq crq;
3878
3879         memset(&crq, 0, sizeof(crq));
3880         crq.query_map.first = IBMVNIC_CRQ_CMD;
3881         crq.query_map.cmd = QUERY_MAP;
3882         ibmvnic_send_crq(adapter, &crq);
3883 }
3884
3885 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3886 static void send_query_cap(struct ibmvnic_adapter *adapter)
3887 {
3888         union ibmvnic_crq crq;
3889
3890         atomic_set(&adapter->running_cap_crqs, 0);
3891         memset(&crq, 0, sizeof(crq));
3892         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3893         crq.query_capability.cmd = QUERY_CAPABILITY;
3894
3895         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3896         atomic_inc(&adapter->running_cap_crqs);
3897         ibmvnic_send_crq(adapter, &crq);
3898
3899         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3900         atomic_inc(&adapter->running_cap_crqs);
3901         ibmvnic_send_crq(adapter, &crq);
3902
3903         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3904         atomic_inc(&adapter->running_cap_crqs);
3905         ibmvnic_send_crq(adapter, &crq);
3906
3907         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3908         atomic_inc(&adapter->running_cap_crqs);
3909         ibmvnic_send_crq(adapter, &crq);
3910
3911         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3912         atomic_inc(&adapter->running_cap_crqs);
3913         ibmvnic_send_crq(adapter, &crq);
3914
3915         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3916         atomic_inc(&adapter->running_cap_crqs);
3917         ibmvnic_send_crq(adapter, &crq);
3918
3919         crq.query_capability.capability =
3920             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3921         atomic_inc(&adapter->running_cap_crqs);
3922         ibmvnic_send_crq(adapter, &crq);
3923
3924         crq.query_capability.capability =
3925             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3926         atomic_inc(&adapter->running_cap_crqs);
3927         ibmvnic_send_crq(adapter, &crq);
3928
3929         crq.query_capability.capability =
3930             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3931         atomic_inc(&adapter->running_cap_crqs);
3932         ibmvnic_send_crq(adapter, &crq);
3933
3934         crq.query_capability.capability =
3935             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3936         atomic_inc(&adapter->running_cap_crqs);
3937         ibmvnic_send_crq(adapter, &crq);
3938
3939         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3940         atomic_inc(&adapter->running_cap_crqs);
3941         ibmvnic_send_crq(adapter, &crq);
3942
3943         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3944         atomic_inc(&adapter->running_cap_crqs);
3945         ibmvnic_send_crq(adapter, &crq);
3946
3947         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3948         atomic_inc(&adapter->running_cap_crqs);
3949         ibmvnic_send_crq(adapter, &crq);
3950
3951         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3952         atomic_inc(&adapter->running_cap_crqs);
3953         ibmvnic_send_crq(adapter, &crq);
3954
3955         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3956         atomic_inc(&adapter->running_cap_crqs);
3957         ibmvnic_send_crq(adapter, &crq);
3958
3959         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3960         atomic_inc(&adapter->running_cap_crqs);
3961         ibmvnic_send_crq(adapter, &crq);
3962
3963         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3964         atomic_inc(&adapter->running_cap_crqs);
3965         ibmvnic_send_crq(adapter, &crq);
3966
3967         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3968         atomic_inc(&adapter->running_cap_crqs);
3969         ibmvnic_send_crq(adapter, &crq);
3970
3971         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3972         atomic_inc(&adapter->running_cap_crqs);
3973         ibmvnic_send_crq(adapter, &crq);
3974
3975         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3976         atomic_inc(&adapter->running_cap_crqs);
3977         ibmvnic_send_crq(adapter, &crq);
3978
3979         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3980         atomic_inc(&adapter->running_cap_crqs);
3981         ibmvnic_send_crq(adapter, &crq);
3982
3983         crq.query_capability.capability =
3984                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3985         atomic_inc(&adapter->running_cap_crqs);
3986         ibmvnic_send_crq(adapter, &crq);
3987
3988         crq.query_capability.capability =
3989                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3990         atomic_inc(&adapter->running_cap_crqs);
3991         ibmvnic_send_crq(adapter, &crq);
3992
3993         crq.query_capability.capability =
3994                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3995         atomic_inc(&adapter->running_cap_crqs);
3996         ibmvnic_send_crq(adapter, &crq);
3997
3998         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3999         atomic_inc(&adapter->running_cap_crqs);
4000         ibmvnic_send_crq(adapter, &crq);
4001 }
4002
4003 static void send_query_ip_offload(struct ibmvnic_adapter *adapter)
4004 {
4005         int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4006         struct device *dev = &adapter->vdev->dev;
4007         union ibmvnic_crq crq;
4008
4009         adapter->ip_offload_tok =
4010                 dma_map_single(dev,
4011                                &adapter->ip_offload_buf,
4012                                buf_sz,
4013                                DMA_FROM_DEVICE);
4014
4015         if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4016                 if (!firmware_has_feature(FW_FEATURE_CMO))
4017                         dev_err(dev, "Couldn't map offload buffer\n");
4018                 return;
4019         }
4020
4021         memset(&crq, 0, sizeof(crq));
4022         crq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4023         crq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4024         crq.query_ip_offload.len = cpu_to_be32(buf_sz);
4025         crq.query_ip_offload.ioba =
4026             cpu_to_be32(adapter->ip_offload_tok);
4027
4028         ibmvnic_send_crq(adapter, &crq);
4029 }
4030
4031 static void send_control_ip_offload(struct ibmvnic_adapter *adapter)
4032 {
4033         struct ibmvnic_control_ip_offload_buffer *ctrl_buf = &adapter->ip_offload_ctrl;
4034         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4035         struct device *dev = &adapter->vdev->dev;
4036         netdev_features_t old_hw_features = 0;
4037         union ibmvnic_crq crq;
4038
4039         adapter->ip_offload_ctrl_tok =
4040                 dma_map_single(dev,
4041                                ctrl_buf,
4042                                sizeof(adapter->ip_offload_ctrl),
4043                                DMA_TO_DEVICE);
4044
4045         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4046                 dev_err(dev, "Couldn't map ip offload control buffer\n");
4047                 return;
4048         }
4049
4050         ctrl_buf->len = cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4051         ctrl_buf->version = cpu_to_be32(INITIAL_VERSION_IOB);
4052         ctrl_buf->ipv4_chksum = buf->ipv4_chksum;
4053         ctrl_buf->ipv6_chksum = buf->ipv6_chksum;
4054         ctrl_buf->tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4055         ctrl_buf->udp_ipv4_chksum = buf->udp_ipv4_chksum;
4056         ctrl_buf->tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4057         ctrl_buf->udp_ipv6_chksum = buf->udp_ipv6_chksum;
4058         ctrl_buf->large_tx_ipv4 = buf->large_tx_ipv4;
4059         ctrl_buf->large_tx_ipv6 = buf->large_tx_ipv6;
4060
4061         /* large_rx disabled for now, additional features needed */
4062         ctrl_buf->large_rx_ipv4 = 0;
4063         ctrl_buf->large_rx_ipv6 = 0;
4064
4065         if (adapter->state != VNIC_PROBING) {
4066                 old_hw_features = adapter->netdev->hw_features;
4067                 adapter->netdev->hw_features = 0;
4068         }
4069
4070         adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4071
4072         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4073                 adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4074
4075         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4076                 adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4077
4078         if ((adapter->netdev->features &
4079             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4080                 adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4081
4082         if (buf->large_tx_ipv4)
4083                 adapter->netdev->hw_features |= NETIF_F_TSO;
4084         if (buf->large_tx_ipv6)
4085                 adapter->netdev->hw_features |= NETIF_F_TSO6;
4086
4087         if (adapter->state == VNIC_PROBING) {
4088                 adapter->netdev->features |= adapter->netdev->hw_features;
4089         } else if (old_hw_features != adapter->netdev->hw_features) {
4090                 netdev_features_t tmp = 0;
4091
4092                 /* disable features no longer supported */
4093                 adapter->netdev->features &= adapter->netdev->hw_features;
4094                 /* turn on features now supported if previously enabled */
4095                 tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4096                         adapter->netdev->hw_features;
4097                 adapter->netdev->features |=
4098                                 tmp & adapter->netdev->wanted_features;
4099         }
4100
4101         memset(&crq, 0, sizeof(crq));
4102         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4103         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4104         crq.control_ip_offload.len =
4105             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4106         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4107         ibmvnic_send_crq(adapter, &crq);
4108 }
4109
4110 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
4111                                 struct ibmvnic_adapter *adapter)
4112 {
4113         struct device *dev = &adapter->vdev->dev;
4114
4115         if (crq->get_vpd_size_rsp.rc.code) {
4116                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
4117                         crq->get_vpd_size_rsp.rc.code);
4118                 complete(&adapter->fw_done);
4119                 return;
4120         }
4121
4122         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
4123         complete(&adapter->fw_done);
4124 }
4125
4126 static void handle_vpd_rsp(union ibmvnic_crq *crq,
4127                            struct ibmvnic_adapter *adapter)
4128 {
4129         struct device *dev = &adapter->vdev->dev;
4130         unsigned char *substr = NULL;
4131         u8 fw_level_len = 0;
4132
4133         memset(adapter->fw_version, 0, 32);
4134
4135         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
4136                          DMA_FROM_DEVICE);
4137
4138         if (crq->get_vpd_rsp.rc.code) {
4139                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
4140                         crq->get_vpd_rsp.rc.code);
4141                 goto complete;
4142         }
4143
4144         /* get the position of the firmware version info
4145          * located after the ASCII 'RM' substring in the buffer
4146          */
4147         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
4148         if (!substr) {
4149                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
4150                 goto complete;
4151         }
4152
4153         /* get length of firmware level ASCII substring */
4154         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
4155                 fw_level_len = *(substr + 2);
4156         } else {
4157                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
4158                 goto complete;
4159         }
4160
4161         /* copy firmware version string from vpd into adapter */
4162         if ((substr + 3 + fw_level_len) <
4163             (adapter->vpd->buff + adapter->vpd->len)) {
4164                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
4165         } else {
4166                 dev_info(dev, "FW substr extrapolated VPD buff\n");
4167         }
4168
4169 complete:
4170         if (adapter->fw_version[0] == '\0')
4171                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
4172         complete(&adapter->fw_done);
4173 }
4174
4175 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4176 {
4177         struct device *dev = &adapter->vdev->dev;
4178         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4179         int i;
4180
4181         dma_unmap_single(dev, adapter->ip_offload_tok,
4182                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4183
4184         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4185         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4186                 netdev_dbg(adapter->netdev, "%016lx\n",
4187                            ((unsigned long int *)(buf))[i]);
4188
4189         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4190         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4191         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4192                    buf->tcp_ipv4_chksum);
4193         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4194                    buf->tcp_ipv6_chksum);
4195         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4196                    buf->udp_ipv4_chksum);
4197         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4198                    buf->udp_ipv6_chksum);
4199         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4200                    buf->large_tx_ipv4);
4201         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4202                    buf->large_tx_ipv6);
4203         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4204                    buf->large_rx_ipv4);
4205         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4206                    buf->large_rx_ipv6);
4207         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4208                    buf->max_ipv4_header_size);
4209         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4210                    buf->max_ipv6_header_size);
4211         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4212                    buf->max_tcp_header_size);
4213         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4214                    buf->max_udp_header_size);
4215         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4216                    buf->max_large_tx_size);
4217         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4218                    buf->max_large_rx_size);
4219         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4220                    buf->ipv6_extension_header);
4221         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4222                    buf->tcp_pseudosum_req);
4223         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4224                    buf->num_ipv6_ext_headers);
4225         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4226                    buf->off_ipv6_ext_headers);
4227
4228         send_control_ip_offload(adapter);
4229 }
4230
4231 static const char *ibmvnic_fw_err_cause(u16 cause)
4232 {
4233         switch (cause) {
4234         case ADAPTER_PROBLEM:
4235                 return "adapter problem";
4236         case BUS_PROBLEM:
4237                 return "bus problem";
4238         case FW_PROBLEM:
4239                 return "firmware problem";
4240         case DD_PROBLEM:
4241                 return "device driver problem";
4242         case EEH_RECOVERY:
4243                 return "EEH recovery";
4244         case FW_UPDATED:
4245                 return "firmware updated";
4246         case LOW_MEMORY:
4247                 return "low Memory";
4248         default:
4249                 return "unknown";
4250         }
4251 }
4252
4253 static void handle_error_indication(union ibmvnic_crq *crq,
4254                                     struct ibmvnic_adapter *adapter)
4255 {
4256         struct device *dev = &adapter->vdev->dev;
4257         u16 cause;
4258
4259         cause = be16_to_cpu(crq->error_indication.error_cause);
4260
4261         dev_warn_ratelimited(dev,
4262                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
4263                              crq->error_indication.flags
4264                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4265                              ibmvnic_fw_err_cause(cause));
4266
4267         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4268                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4269         else
4270                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4271 }
4272
4273 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4274                                  struct ibmvnic_adapter *adapter)
4275 {
4276         struct net_device *netdev = adapter->netdev;
4277         struct device *dev = &adapter->vdev->dev;
4278         long rc;
4279
4280         rc = crq->change_mac_addr_rsp.rc.code;
4281         if (rc) {
4282                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4283                 goto out;
4284         }
4285         /* crq->change_mac_addr.mac_addr is the requested one
4286          * crq->change_mac_addr_rsp.mac_addr is the returned valid one.
4287          */
4288         ether_addr_copy(netdev->dev_addr,
4289                         &crq->change_mac_addr_rsp.mac_addr[0]);
4290         ether_addr_copy(adapter->mac_addr,
4291                         &crq->change_mac_addr_rsp.mac_addr[0]);
4292 out:
4293         complete(&adapter->fw_done);
4294         return rc;
4295 }
4296
4297 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4298                                    struct ibmvnic_adapter *adapter)
4299 {
4300         struct device *dev = &adapter->vdev->dev;
4301         u64 *req_value;
4302         char *name;
4303
4304         atomic_dec(&adapter->running_cap_crqs);
4305         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4306         case REQ_TX_QUEUES:
4307                 req_value = &adapter->req_tx_queues;
4308                 name = "tx";
4309                 break;
4310         case REQ_RX_QUEUES:
4311                 req_value = &adapter->req_rx_queues;
4312                 name = "rx";
4313                 break;
4314         case REQ_RX_ADD_QUEUES:
4315                 req_value = &adapter->req_rx_add_queues;
4316                 name = "rx_add";
4317                 break;
4318         case REQ_TX_ENTRIES_PER_SUBCRQ:
4319                 req_value = &adapter->req_tx_entries_per_subcrq;
4320                 name = "tx_entries_per_subcrq";
4321                 break;
4322         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4323                 req_value = &adapter->req_rx_add_entries_per_subcrq;
4324                 name = "rx_add_entries_per_subcrq";
4325                 break;
4326         case REQ_MTU:
4327                 req_value = &adapter->req_mtu;
4328                 name = "mtu";
4329                 break;
4330         case PROMISC_REQUESTED:
4331                 req_value = &adapter->promisc;
4332                 name = "promisc";
4333                 break;
4334         default:
4335                 dev_err(dev, "Got invalid cap request rsp %d\n",
4336                         crq->request_capability.capability);
4337                 return;
4338         }
4339
4340         switch (crq->request_capability_rsp.rc.code) {
4341         case SUCCESS:
4342                 break;
4343         case PARTIALSUCCESS:
4344                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4345                          *req_value,
4346                          (long int)be64_to_cpu(crq->request_capability_rsp.
4347                                                number), name);
4348
4349                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4350                     REQ_MTU) {
4351                         pr_err("mtu of %llu is not supported. Reverting.\n",
4352                                *req_value);
4353                         *req_value = adapter->fallback.mtu;
4354                 } else {
4355                         *req_value =
4356                                 be64_to_cpu(crq->request_capability_rsp.number);
4357                 }
4358
4359                 send_request_cap(adapter, 1);
4360                 return;
4361         default:
4362                 dev_err(dev, "Error %d in request cap rsp\n",
4363                         crq->request_capability_rsp.rc.code);
4364                 return;
4365         }
4366
4367         /* Done receiving requested capabilities, query IP offload support */
4368         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4369                 adapter->wait_capability = false;
4370                 send_query_ip_offload(adapter);
4371         }
4372 }
4373
4374 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4375                             struct ibmvnic_adapter *adapter)
4376 {
4377         struct device *dev = &adapter->vdev->dev;
4378         struct net_device *netdev = adapter->netdev;
4379         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4380         struct ibmvnic_login_buffer *login = adapter->login_buf;
4381         u64 *tx_handle_array;
4382         u64 *rx_handle_array;
4383         int num_tx_pools;
4384         int num_rx_pools;
4385         u64 *size_array;
4386         int i;
4387
4388         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4389                          DMA_TO_DEVICE);
4390         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4391                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4392
4393         /* If the number of queues requested can't be allocated by the
4394          * server, the login response will return with code 1. We will need
4395          * to resend the login buffer with fewer queues requested.
4396          */
4397         if (login_rsp_crq->generic.rc.code) {
4398                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4399                 complete(&adapter->init_done);
4400                 return 0;
4401         }
4402
4403         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4404
4405         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4406         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4407                 netdev_dbg(adapter->netdev, "%016lx\n",
4408                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4409         }
4410
4411         /* Sanity checks */
4412         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4413             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4414              adapter->req_rx_add_queues !=
4415              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4416                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4417                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4418                 return -EIO;
4419         }
4420         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4421                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
4422         /* variable buffer sizes are not supported, so just read the
4423          * first entry.
4424          */
4425         adapter->cur_rx_buf_sz = be64_to_cpu(size_array[0]);
4426
4427         num_tx_pools = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
4428         num_rx_pools = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4429
4430         tx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4431                                   be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
4432         rx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4433                                   be32_to_cpu(adapter->login_rsp_buf->off_rxadd_subcrqs));
4434
4435         for (i = 0; i < num_tx_pools; i++)
4436                 adapter->tx_scrq[i]->handle = tx_handle_array[i];
4437
4438         for (i = 0; i < num_rx_pools; i++)
4439                 adapter->rx_scrq[i]->handle = rx_handle_array[i];
4440
4441         adapter->num_active_tx_scrqs = num_tx_pools;
4442         adapter->num_active_rx_scrqs = num_rx_pools;
4443         release_login_rsp_buffer(adapter);
4444         release_login_buffer(adapter);
4445         complete(&adapter->init_done);
4446
4447         return 0;
4448 }
4449
4450 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4451                                      struct ibmvnic_adapter *adapter)
4452 {
4453         struct device *dev = &adapter->vdev->dev;
4454         long rc;
4455
4456         rc = crq->request_unmap_rsp.rc.code;
4457         if (rc)
4458                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4459 }
4460
4461 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4462                                  struct ibmvnic_adapter *adapter)
4463 {
4464         struct net_device *netdev = adapter->netdev;
4465         struct device *dev = &adapter->vdev->dev;
4466         long rc;
4467
4468         rc = crq->query_map_rsp.rc.code;
4469         if (rc) {
4470                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4471                 return;
4472         }
4473         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4474                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4475                    crq->query_map_rsp.free_pages);
4476 }
4477
4478 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4479                                  struct ibmvnic_adapter *adapter)
4480 {
4481         struct net_device *netdev = adapter->netdev;
4482         struct device *dev = &adapter->vdev->dev;
4483         long rc;
4484
4485         atomic_dec(&adapter->running_cap_crqs);
4486         netdev_dbg(netdev, "Outstanding queries: %d\n",
4487                    atomic_read(&adapter->running_cap_crqs));
4488         rc = crq->query_capability.rc.code;
4489         if (rc) {
4490                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4491                 goto out;
4492         }
4493
4494         switch (be16_to_cpu(crq->query_capability.capability)) {
4495         case MIN_TX_QUEUES:
4496                 adapter->min_tx_queues =
4497                     be64_to_cpu(crq->query_capability.number);
4498                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4499                            adapter->min_tx_queues);
4500                 break;
4501         case MIN_RX_QUEUES:
4502                 adapter->min_rx_queues =
4503                     be64_to_cpu(crq->query_capability.number);
4504                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4505                            adapter->min_rx_queues);
4506                 break;
4507         case MIN_RX_ADD_QUEUES:
4508                 adapter->min_rx_add_queues =
4509                     be64_to_cpu(crq->query_capability.number);
4510                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4511                            adapter->min_rx_add_queues);
4512                 break;
4513         case MAX_TX_QUEUES:
4514                 adapter->max_tx_queues =
4515                     be64_to_cpu(crq->query_capability.number);
4516                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4517                            adapter->max_tx_queues);
4518                 break;
4519         case MAX_RX_QUEUES:
4520                 adapter->max_rx_queues =
4521                     be64_to_cpu(crq->query_capability.number);
4522                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4523                            adapter->max_rx_queues);
4524                 break;
4525         case MAX_RX_ADD_QUEUES:
4526                 adapter->max_rx_add_queues =
4527                     be64_to_cpu(crq->query_capability.number);
4528                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4529                            adapter->max_rx_add_queues);
4530                 break;
4531         case MIN_TX_ENTRIES_PER_SUBCRQ:
4532                 adapter->min_tx_entries_per_subcrq =
4533                     be64_to_cpu(crq->query_capability.number);
4534                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4535                            adapter->min_tx_entries_per_subcrq);
4536                 break;
4537         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4538                 adapter->min_rx_add_entries_per_subcrq =
4539                     be64_to_cpu(crq->query_capability.number);
4540                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4541                            adapter->min_rx_add_entries_per_subcrq);
4542                 break;
4543         case MAX_TX_ENTRIES_PER_SUBCRQ:
4544                 adapter->max_tx_entries_per_subcrq =
4545                     be64_to_cpu(crq->query_capability.number);
4546                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4547                            adapter->max_tx_entries_per_subcrq);
4548                 break;
4549         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4550                 adapter->max_rx_add_entries_per_subcrq =
4551                     be64_to_cpu(crq->query_capability.number);
4552                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4553                            adapter->max_rx_add_entries_per_subcrq);
4554                 break;
4555         case TCP_IP_OFFLOAD:
4556                 adapter->tcp_ip_offload =
4557                     be64_to_cpu(crq->query_capability.number);
4558                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4559                            adapter->tcp_ip_offload);
4560                 break;
4561         case PROMISC_SUPPORTED:
4562                 adapter->promisc_supported =
4563                     be64_to_cpu(crq->query_capability.number);
4564                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4565                            adapter->promisc_supported);
4566                 break;
4567         case MIN_MTU:
4568                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4569                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4570                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4571                 break;
4572         case MAX_MTU:
4573                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4574                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4575                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4576                 break;
4577         case MAX_MULTICAST_FILTERS:
4578                 adapter->max_multicast_filters =
4579                     be64_to_cpu(crq->query_capability.number);
4580                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4581                            adapter->max_multicast_filters);
4582                 break;
4583         case VLAN_HEADER_INSERTION:
4584                 adapter->vlan_header_insertion =
4585                     be64_to_cpu(crq->query_capability.number);
4586                 if (adapter->vlan_header_insertion)
4587                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4588                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4589                            adapter->vlan_header_insertion);
4590                 break;
4591         case RX_VLAN_HEADER_INSERTION:
4592                 adapter->rx_vlan_header_insertion =
4593                     be64_to_cpu(crq->query_capability.number);
4594                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4595                            adapter->rx_vlan_header_insertion);
4596                 break;
4597         case MAX_TX_SG_ENTRIES:
4598                 adapter->max_tx_sg_entries =
4599                     be64_to_cpu(crq->query_capability.number);
4600                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4601                            adapter->max_tx_sg_entries);
4602                 break;
4603         case RX_SG_SUPPORTED:
4604                 adapter->rx_sg_supported =
4605                     be64_to_cpu(crq->query_capability.number);
4606                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4607                            adapter->rx_sg_supported);
4608                 break;
4609         case OPT_TX_COMP_SUB_QUEUES:
4610                 adapter->opt_tx_comp_sub_queues =
4611                     be64_to_cpu(crq->query_capability.number);
4612                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4613                            adapter->opt_tx_comp_sub_queues);
4614                 break;
4615         case OPT_RX_COMP_QUEUES:
4616                 adapter->opt_rx_comp_queues =
4617                     be64_to_cpu(crq->query_capability.number);
4618                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4619                            adapter->opt_rx_comp_queues);
4620                 break;
4621         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4622                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4623                     be64_to_cpu(crq->query_capability.number);
4624                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4625                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4626                 break;
4627         case OPT_TX_ENTRIES_PER_SUBCRQ:
4628                 adapter->opt_tx_entries_per_subcrq =
4629                     be64_to_cpu(crq->query_capability.number);
4630                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4631                            adapter->opt_tx_entries_per_subcrq);
4632                 break;
4633         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4634                 adapter->opt_rxba_entries_per_subcrq =
4635                     be64_to_cpu(crq->query_capability.number);
4636                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4637                            adapter->opt_rxba_entries_per_subcrq);
4638                 break;
4639         case TX_RX_DESC_REQ:
4640                 adapter->tx_rx_desc_req = crq->query_capability.number;
4641                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4642                            adapter->tx_rx_desc_req);
4643                 break;
4644
4645         default:
4646                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4647                            crq->query_capability.capability);
4648         }
4649
4650 out:
4651         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4652                 adapter->wait_capability = false;
4653                 send_request_cap(adapter, 0);
4654         }
4655 }
4656
4657 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4658 {
4659         union ibmvnic_crq crq;
4660         int rc;
4661
4662         memset(&crq, 0, sizeof(crq));
4663         crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4664         crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4665
4666         mutex_lock(&adapter->fw_lock);
4667         adapter->fw_done_rc = 0;
4668         reinit_completion(&adapter->fw_done);
4669
4670         rc = ibmvnic_send_crq(adapter, &crq);
4671         if (rc) {
4672                 mutex_unlock(&adapter->fw_lock);
4673                 return rc;
4674         }
4675
4676         rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4677         if (rc) {
4678                 mutex_unlock(&adapter->fw_lock);
4679                 return rc;
4680         }
4681
4682         mutex_unlock(&adapter->fw_lock);
4683         return adapter->fw_done_rc ? -EIO : 0;
4684 }
4685
4686 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4687                                        struct ibmvnic_adapter *adapter)
4688 {
4689         struct net_device *netdev = adapter->netdev;
4690         int rc;
4691         __be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4692
4693         rc = crq->query_phys_parms_rsp.rc.code;
4694         if (rc) {
4695                 netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4696                 return rc;
4697         }
4698         switch (rspeed) {
4699         case IBMVNIC_10MBPS:
4700                 adapter->speed = SPEED_10;
4701                 break;
4702         case IBMVNIC_100MBPS:
4703                 adapter->speed = SPEED_100;
4704                 break;
4705         case IBMVNIC_1GBPS:
4706                 adapter->speed = SPEED_1000;
4707                 break;
4708         case IBMVNIC_10GBPS:
4709                 adapter->speed = SPEED_10000;
4710                 break;
4711         case IBMVNIC_25GBPS:
4712                 adapter->speed = SPEED_25000;
4713                 break;
4714         case IBMVNIC_40GBPS:
4715                 adapter->speed = SPEED_40000;
4716                 break;
4717         case IBMVNIC_50GBPS:
4718                 adapter->speed = SPEED_50000;
4719                 break;
4720         case IBMVNIC_100GBPS:
4721                 adapter->speed = SPEED_100000;
4722                 break;
4723         case IBMVNIC_200GBPS:
4724                 adapter->speed = SPEED_200000;
4725                 break;
4726         default:
4727                 if (netif_carrier_ok(netdev))
4728                         netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4729                 adapter->speed = SPEED_UNKNOWN;
4730         }
4731         if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4732                 adapter->duplex = DUPLEX_FULL;
4733         else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4734                 adapter->duplex = DUPLEX_HALF;
4735         else
4736                 adapter->duplex = DUPLEX_UNKNOWN;
4737
4738         return rc;
4739 }
4740
4741 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4742                                struct ibmvnic_adapter *adapter)
4743 {
4744         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4745         struct net_device *netdev = adapter->netdev;
4746         struct device *dev = &adapter->vdev->dev;
4747         u64 *u64_crq = (u64 *)crq;
4748         long rc;
4749
4750         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4751                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4752                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4753         switch (gen_crq->first) {
4754         case IBMVNIC_CRQ_INIT_RSP:
4755                 switch (gen_crq->cmd) {
4756                 case IBMVNIC_CRQ_INIT:
4757                         dev_info(dev, "Partner initialized\n");
4758                         adapter->from_passive_init = true;
4759                         if (!completion_done(&adapter->init_done)) {
4760                                 complete(&adapter->init_done);
4761                                 adapter->init_done_rc = -EIO;
4762                         }
4763                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4764                         break;
4765                 case IBMVNIC_CRQ_INIT_COMPLETE:
4766                         dev_info(dev, "Partner initialization complete\n");
4767                         adapter->crq.active = true;
4768                         send_version_xchg(adapter);
4769                         break;
4770                 default:
4771                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4772                 }
4773                 return;
4774         case IBMVNIC_CRQ_XPORT_EVENT:
4775                 netif_carrier_off(netdev);
4776                 adapter->crq.active = false;
4777                 /* terminate any thread waiting for a response
4778                  * from the device
4779                  */
4780                 if (!completion_done(&adapter->fw_done)) {
4781                         adapter->fw_done_rc = -EIO;
4782                         complete(&adapter->fw_done);
4783                 }
4784                 if (!completion_done(&adapter->stats_done))
4785                         complete(&adapter->stats_done);
4786                 if (test_bit(0, &adapter->resetting))
4787                         adapter->force_reset_recovery = true;
4788                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4789                         dev_info(dev, "Migrated, re-enabling adapter\n");
4790                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4791                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4792                         dev_info(dev, "Backing device failover detected\n");
4793                         adapter->failover_pending = true;
4794                 } else {
4795                         /* The adapter lost the connection */
4796                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4797                                 gen_crq->cmd);
4798                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4799                 }
4800                 return;
4801         case IBMVNIC_CRQ_CMD_RSP:
4802                 break;
4803         default:
4804                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4805                         gen_crq->first);
4806                 return;
4807         }
4808
4809         switch (gen_crq->cmd) {
4810         case VERSION_EXCHANGE_RSP:
4811                 rc = crq->version_exchange_rsp.rc.code;
4812                 if (rc) {
4813                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4814                         break;
4815                 }
4816                 ibmvnic_version =
4817                             be16_to_cpu(crq->version_exchange_rsp.version);
4818                 dev_info(dev, "Partner protocol version is %d\n",
4819                          ibmvnic_version);
4820                 send_query_cap(adapter);
4821                 break;
4822         case QUERY_CAPABILITY_RSP:
4823                 handle_query_cap_rsp(crq, adapter);
4824                 break;
4825         case QUERY_MAP_RSP:
4826                 handle_query_map_rsp(crq, adapter);
4827                 break;
4828         case REQUEST_MAP_RSP:
4829                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4830                 complete(&adapter->fw_done);
4831                 break;
4832         case REQUEST_UNMAP_RSP:
4833                 handle_request_unmap_rsp(crq, adapter);
4834                 break;
4835         case REQUEST_CAPABILITY_RSP:
4836                 handle_request_cap_rsp(crq, adapter);
4837                 break;
4838         case LOGIN_RSP:
4839                 netdev_dbg(netdev, "Got Login Response\n");
4840                 handle_login_rsp(crq, adapter);
4841                 break;
4842         case LOGICAL_LINK_STATE_RSP:
4843                 netdev_dbg(netdev,
4844                            "Got Logical Link State Response, state: %d rc: %d\n",
4845                            crq->logical_link_state_rsp.link_state,
4846                            crq->logical_link_state_rsp.rc.code);
4847                 adapter->logical_link_state =
4848                     crq->logical_link_state_rsp.link_state;
4849                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4850                 complete(&adapter->init_done);
4851                 break;
4852         case LINK_STATE_INDICATION:
4853                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4854                 adapter->phys_link_state =
4855                     crq->link_state_indication.phys_link_state;
4856                 adapter->logical_link_state =
4857                     crq->link_state_indication.logical_link_state;
4858                 if (adapter->phys_link_state && adapter->logical_link_state)
4859                         netif_carrier_on(netdev);
4860                 else
4861                         netif_carrier_off(netdev);
4862                 break;
4863         case CHANGE_MAC_ADDR_RSP:
4864                 netdev_dbg(netdev, "Got MAC address change Response\n");
4865                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4866                 break;
4867         case ERROR_INDICATION:
4868                 netdev_dbg(netdev, "Got Error Indication\n");
4869                 handle_error_indication(crq, adapter);
4870                 break;
4871         case REQUEST_STATISTICS_RSP:
4872                 netdev_dbg(netdev, "Got Statistics Response\n");
4873                 complete(&adapter->stats_done);
4874                 break;
4875         case QUERY_IP_OFFLOAD_RSP:
4876                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4877                 handle_query_ip_offload_rsp(adapter);
4878                 break;
4879         case MULTICAST_CTRL_RSP:
4880                 netdev_dbg(netdev, "Got multicast control Response\n");
4881                 break;
4882         case CONTROL_IP_OFFLOAD_RSP:
4883                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4884                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4885                                  sizeof(adapter->ip_offload_ctrl),
4886                                  DMA_TO_DEVICE);
4887                 complete(&adapter->init_done);
4888                 break;
4889         case COLLECT_FW_TRACE_RSP:
4890                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4891                 complete(&adapter->fw_done);
4892                 break;
4893         case GET_VPD_SIZE_RSP:
4894                 handle_vpd_size_rsp(crq, adapter);
4895                 break;
4896         case GET_VPD_RSP:
4897                 handle_vpd_rsp(crq, adapter);
4898                 break;
4899         case QUERY_PHYS_PARMS_RSP:
4900                 adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4901                 complete(&adapter->fw_done);
4902                 break;
4903         default:
4904                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4905                            gen_crq->cmd);
4906         }
4907 }
4908
4909 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4910 {
4911         struct ibmvnic_adapter *adapter = instance;
4912
4913         tasklet_schedule(&adapter->tasklet);
4914         return IRQ_HANDLED;
4915 }
4916
4917 static void ibmvnic_tasklet(struct tasklet_struct *t)
4918 {
4919         struct ibmvnic_adapter *adapter = from_tasklet(adapter, t, tasklet);
4920         struct ibmvnic_crq_queue *queue = &adapter->crq;
4921         union ibmvnic_crq *crq;
4922         unsigned long flags;
4923         bool done = false;
4924
4925         spin_lock_irqsave(&queue->lock, flags);
4926         while (!done) {
4927                 /* Pull all the valid messages off the CRQ */
4928                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4929                         ibmvnic_handle_crq(crq, adapter);
4930                         crq->generic.first = 0;
4931                 }
4932
4933                 /* remain in tasklet until all
4934                  * capabilities responses are received
4935                  */
4936                 if (!adapter->wait_capability)
4937                         done = true;
4938         }
4939         /* if capabilities CRQ's were sent in this tasklet, the following
4940          * tasklet must wait until all responses are received
4941          */
4942         if (atomic_read(&adapter->running_cap_crqs) != 0)
4943                 adapter->wait_capability = true;
4944         spin_unlock_irqrestore(&queue->lock, flags);
4945 }
4946
4947 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4948 {
4949         struct vio_dev *vdev = adapter->vdev;
4950         int rc;
4951
4952         do {
4953                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4954         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4955
4956         if (rc)
4957                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4958
4959         return rc;
4960 }
4961
4962 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4963 {
4964         struct ibmvnic_crq_queue *crq = &adapter->crq;
4965         struct device *dev = &adapter->vdev->dev;
4966         struct vio_dev *vdev = adapter->vdev;
4967         int rc;
4968
4969         /* Close the CRQ */
4970         do {
4971                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4972         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4973
4974         /* Clean out the queue */
4975         if (!crq->msgs)
4976                 return -EINVAL;
4977
4978         memset(crq->msgs, 0, PAGE_SIZE);
4979         crq->cur = 0;
4980         crq->active = false;
4981
4982         /* And re-open it again */
4983         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4984                                 crq->msg_token, PAGE_SIZE);
4985
4986         if (rc == H_CLOSED)
4987                 /* Adapter is good, but other end is not ready */
4988                 dev_warn(dev, "Partner adapter not ready\n");
4989         else if (rc != 0)
4990                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4991
4992         return rc;
4993 }
4994
4995 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4996 {
4997         struct ibmvnic_crq_queue *crq = &adapter->crq;
4998         struct vio_dev *vdev = adapter->vdev;
4999         long rc;
5000
5001         if (!crq->msgs)
5002                 return;
5003
5004         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
5005         free_irq(vdev->irq, adapter);
5006         tasklet_kill(&adapter->tasklet);
5007         do {
5008                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5009         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5010
5011         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
5012                          DMA_BIDIRECTIONAL);
5013         free_page((unsigned long)crq->msgs);
5014         crq->msgs = NULL;
5015         crq->active = false;
5016 }
5017
5018 static int init_crq_queue(struct ibmvnic_adapter *adapter)
5019 {
5020         struct ibmvnic_crq_queue *crq = &adapter->crq;
5021         struct device *dev = &adapter->vdev->dev;
5022         struct vio_dev *vdev = adapter->vdev;
5023         int rc, retrc = -ENOMEM;
5024
5025         if (crq->msgs)
5026                 return 0;
5027
5028         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
5029         /* Should we allocate more than one page? */
5030
5031         if (!crq->msgs)
5032                 return -ENOMEM;
5033
5034         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
5035         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
5036                                         DMA_BIDIRECTIONAL);
5037         if (dma_mapping_error(dev, crq->msg_token))
5038                 goto map_failed;
5039
5040         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
5041                                 crq->msg_token, PAGE_SIZE);
5042
5043         if (rc == H_RESOURCE)
5044                 /* maybe kexecing and resource is busy. try a reset */
5045                 rc = ibmvnic_reset_crq(adapter);
5046         retrc = rc;
5047
5048         if (rc == H_CLOSED) {
5049                 dev_warn(dev, "Partner adapter not ready\n");
5050         } else if (rc) {
5051                 dev_warn(dev, "Error %d opening adapter\n", rc);
5052                 goto reg_crq_failed;
5053         }
5054
5055         retrc = 0;
5056
5057         tasklet_setup(&adapter->tasklet, (void *)ibmvnic_tasklet);
5058
5059         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
5060         snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
5061                  adapter->vdev->unit_address);
5062         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
5063         if (rc) {
5064                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
5065                         vdev->irq, rc);
5066                 goto req_irq_failed;
5067         }
5068
5069         rc = vio_enable_interrupts(vdev);
5070         if (rc) {
5071                 dev_err(dev, "Error %d enabling interrupts\n", rc);
5072                 goto req_irq_failed;
5073         }
5074
5075         crq->cur = 0;
5076         spin_lock_init(&crq->lock);
5077
5078         return retrc;
5079
5080 req_irq_failed:
5081         tasklet_kill(&adapter->tasklet);
5082         do {
5083                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5084         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5085 reg_crq_failed:
5086         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
5087 map_failed:
5088         free_page((unsigned long)crq->msgs);
5089         crq->msgs = NULL;
5090         return retrc;
5091 }
5092
5093 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter, bool reset)
5094 {
5095         struct device *dev = &adapter->vdev->dev;
5096         unsigned long timeout = msecs_to_jiffies(30000);
5097         u64 old_num_rx_queues, old_num_tx_queues;
5098         int rc;
5099
5100         adapter->from_passive_init = false;
5101
5102         if (reset) {
5103                 old_num_rx_queues = adapter->req_rx_queues;
5104                 old_num_tx_queues = adapter->req_tx_queues;
5105                 reinit_completion(&adapter->init_done);
5106         }
5107
5108         adapter->init_done_rc = 0;
5109         rc = ibmvnic_send_crq_init(adapter);
5110         if (rc) {
5111                 dev_err(dev, "Send crq init failed with error %d\n", rc);
5112                 return rc;
5113         }
5114
5115         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5116                 dev_err(dev, "Initialization sequence timed out\n");
5117                 return -1;
5118         }
5119
5120         if (adapter->init_done_rc) {
5121                 release_crq_queue(adapter);
5122                 return adapter->init_done_rc;
5123         }
5124
5125         if (adapter->from_passive_init) {
5126                 adapter->state = VNIC_OPEN;
5127                 adapter->from_passive_init = false;
5128                 return -1;
5129         }
5130
5131         if (reset &&
5132             test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5133             adapter->reset_reason != VNIC_RESET_MOBILITY) {
5134                 if (adapter->req_rx_queues != old_num_rx_queues ||
5135                     adapter->req_tx_queues != old_num_tx_queues) {
5136                         release_sub_crqs(adapter, 0);
5137                         rc = init_sub_crqs(adapter);
5138                 } else {
5139                         rc = reset_sub_crq_queues(adapter);
5140                 }
5141         } else {
5142                 rc = init_sub_crqs(adapter);
5143         }
5144
5145         if (rc) {
5146                 dev_err(dev, "Initialization of sub crqs failed\n");
5147                 release_crq_queue(adapter);
5148                 return rc;
5149         }
5150
5151         rc = init_sub_crq_irqs(adapter);
5152         if (rc) {
5153                 dev_err(dev, "Failed to initialize sub crq irqs\n");
5154                 release_crq_queue(adapter);
5155         }
5156
5157         return rc;
5158 }
5159
5160 static struct device_attribute dev_attr_failover;
5161
5162 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5163 {
5164         struct ibmvnic_adapter *adapter;
5165         struct net_device *netdev;
5166         unsigned char *mac_addr_p;
5167         int rc;
5168
5169         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5170                 dev->unit_address);
5171
5172         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5173                                                         VETH_MAC_ADDR, NULL);
5174         if (!mac_addr_p) {
5175                 dev_err(&dev->dev,
5176                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5177                         __FILE__, __LINE__);
5178                 return 0;
5179         }
5180
5181         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5182                                    IBMVNIC_MAX_QUEUES);
5183         if (!netdev)
5184                 return -ENOMEM;
5185
5186         adapter = netdev_priv(netdev);
5187         adapter->state = VNIC_PROBING;
5188         dev_set_drvdata(&dev->dev, netdev);
5189         adapter->vdev = dev;
5190         adapter->netdev = netdev;
5191
5192         ether_addr_copy(adapter->mac_addr, mac_addr_p);
5193         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5194         netdev->irq = dev->irq;
5195         netdev->netdev_ops = &ibmvnic_netdev_ops;
5196         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5197         SET_NETDEV_DEV(netdev, &dev->dev);
5198
5199         spin_lock_init(&adapter->stats_lock);
5200
5201         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5202         INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5203                           __ibmvnic_delayed_reset);
5204         INIT_LIST_HEAD(&adapter->rwi_list);
5205         spin_lock_init(&adapter->rwi_lock);
5206         spin_lock_init(&adapter->state_lock);
5207         mutex_init(&adapter->fw_lock);
5208         init_completion(&adapter->init_done);
5209         init_completion(&adapter->fw_done);
5210         init_completion(&adapter->reset_done);
5211         init_completion(&adapter->stats_done);
5212         clear_bit(0, &adapter->resetting);
5213
5214         do {
5215                 rc = init_crq_queue(adapter);
5216                 if (rc) {
5217                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5218                                 rc);
5219                         goto ibmvnic_init_fail;
5220                 }
5221
5222                 rc = ibmvnic_reset_init(adapter, false);
5223                 if (rc && rc != EAGAIN)
5224                         goto ibmvnic_init_fail;
5225         } while (rc == EAGAIN);
5226
5227         rc = init_stats_buffers(adapter);
5228         if (rc)
5229                 goto ibmvnic_init_fail;
5230
5231         rc = init_stats_token(adapter);
5232         if (rc)
5233                 goto ibmvnic_stats_fail;
5234
5235         netdev->mtu = adapter->req_mtu - ETH_HLEN;
5236         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5237         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5238
5239         rc = device_create_file(&dev->dev, &dev_attr_failover);
5240         if (rc)
5241                 goto ibmvnic_dev_file_err;
5242
5243         netif_carrier_off(netdev);
5244         rc = register_netdev(netdev);
5245         if (rc) {
5246                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5247                 goto ibmvnic_register_fail;
5248         }
5249         dev_info(&dev->dev, "ibmvnic registered\n");
5250
5251         adapter->state = VNIC_PROBED;
5252
5253         adapter->wait_for_reset = false;
5254
5255         return 0;
5256
5257 ibmvnic_register_fail:
5258         device_remove_file(&dev->dev, &dev_attr_failover);
5259
5260 ibmvnic_dev_file_err:
5261         release_stats_token(adapter);
5262
5263 ibmvnic_stats_fail:
5264         release_stats_buffers(adapter);
5265
5266 ibmvnic_init_fail:
5267         release_sub_crqs(adapter, 1);
5268         release_crq_queue(adapter);
5269         mutex_destroy(&adapter->fw_lock);
5270         free_netdev(netdev);
5271
5272         return rc;
5273 }
5274
5275 static int ibmvnic_remove(struct vio_dev *dev)
5276 {
5277         struct net_device *netdev = dev_get_drvdata(&dev->dev);
5278         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5279         unsigned long flags;
5280
5281         spin_lock_irqsave(&adapter->state_lock, flags);
5282         if (test_bit(0, &adapter->resetting)) {
5283                 spin_unlock_irqrestore(&adapter->state_lock, flags);
5284                 return -EBUSY;
5285         }
5286
5287         adapter->state = VNIC_REMOVING;
5288         spin_unlock_irqrestore(&adapter->state_lock, flags);
5289
5290         flush_work(&adapter->ibmvnic_reset);
5291         flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5292
5293         rtnl_lock();
5294         unregister_netdevice(netdev);
5295
5296         release_resources(adapter);
5297         release_sub_crqs(adapter, 1);
5298         release_crq_queue(adapter);
5299
5300         release_stats_token(adapter);
5301         release_stats_buffers(adapter);
5302
5303         adapter->state = VNIC_REMOVED;
5304
5305         rtnl_unlock();
5306         mutex_destroy(&adapter->fw_lock);
5307         device_remove_file(&dev->dev, &dev_attr_failover);
5308         free_netdev(netdev);
5309         dev_set_drvdata(&dev->dev, NULL);
5310
5311         return 0;
5312 }
5313
5314 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5315                               const char *buf, size_t count)
5316 {
5317         struct net_device *netdev = dev_get_drvdata(dev);
5318         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5319         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5320         __be64 session_token;
5321         long rc;
5322
5323         if (!sysfs_streq(buf, "1"))
5324                 return -EINVAL;
5325
5326         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5327                          H_GET_SESSION_TOKEN, 0, 0, 0);
5328         if (rc) {
5329                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5330                            rc);
5331                 return -EINVAL;
5332         }
5333
5334         session_token = (__be64)retbuf[0];
5335         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5336                    be64_to_cpu(session_token));
5337         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5338                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
5339         if (rc) {
5340                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5341                            rc);
5342                 return -EINVAL;
5343         }
5344
5345         return count;
5346 }
5347
5348 static DEVICE_ATTR_WO(failover);
5349
5350 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5351 {
5352         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5353         struct ibmvnic_adapter *adapter;
5354         struct iommu_table *tbl;
5355         unsigned long ret = 0;
5356         int i;
5357
5358         tbl = get_iommu_table_base(&vdev->dev);
5359
5360         /* netdev inits at probe time along with the structures we need below*/
5361         if (!netdev)
5362                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5363
5364         adapter = netdev_priv(netdev);
5365
5366         ret += PAGE_SIZE; /* the crq message queue */
5367         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5368
5369         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5370                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
5371
5372         for (i = 0; i < adapter->num_active_rx_pools; i++)
5373                 ret += adapter->rx_pool[i].size *
5374                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5375
5376         return ret;
5377 }
5378
5379 static int ibmvnic_resume(struct device *dev)
5380 {
5381         struct net_device *netdev = dev_get_drvdata(dev);
5382         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5383
5384         if (adapter->state != VNIC_OPEN)
5385                 return 0;
5386
5387         tasklet_schedule(&adapter->tasklet);
5388
5389         return 0;
5390 }
5391
5392 static const struct vio_device_id ibmvnic_device_table[] = {
5393         {"network", "IBM,vnic"},
5394         {"", "" }
5395 };
5396 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5397
5398 static const struct dev_pm_ops ibmvnic_pm_ops = {
5399         .resume = ibmvnic_resume
5400 };
5401
5402 static struct vio_driver ibmvnic_driver = {
5403         .id_table       = ibmvnic_device_table,
5404         .probe          = ibmvnic_probe,
5405         .remove         = ibmvnic_remove,
5406         .get_desired_dma = ibmvnic_get_desired_dma,
5407         .name           = ibmvnic_driver_name,
5408         .pm             = &ibmvnic_pm_ops,
5409 };
5410
5411 /* module functions */
5412 static int __init ibmvnic_module_init(void)
5413 {
5414         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5415                 IBMVNIC_DRIVER_VERSION);
5416
5417         return vio_register_driver(&ibmvnic_driver);
5418 }
5419
5420 static void __exit ibmvnic_module_exit(void)
5421 {
5422         vio_unregister_driver(&ibmvnic_driver);
5423 }
5424
5425 module_init(ibmvnic_module_init);
5426 module_exit(ibmvnic_module_exit);