drivers/net/ethernet/ibm/ibmvnic.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /*  IBM System i and System p Virtual NIC Device Driver                   */
   4 /*  Copyright (C) 2014 IBM Corp.                                          */
   5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   8 /*                                                                        */
   9 /*  This program is free software; you can redistribute it and/or modify  */
  10 /*  it under the terms of the GNU General Public License as published by  */
  11 /*  the Free Software Foundation; either version 2 of the License, or     */
  12 /*  (at your option) any later version.                                   */
  13 /*                                                                        */
  14 /*  This program is distributed in the hope that it will be useful,       */
  15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  17 /*  GNU General Public License for more details.                          */
  18 /*                                                                        */
  19 /*  You should have received a copy of the GNU General Public License     */
  20 /*  along with this program.                                              */
  21 /*                                                                        */
  22 /* This module contains the implementation of a virtual ethernet device   */
  23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  26 /*                                                                         */
  27 /* Messages are passed between the VNIC driver and the VNIC server using  */
  28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  29 /* issue and receive commands that initiate communication with the server */
  30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  31 /* are used by the driver to notify the server that a packet is           */
  32 /* ready for transmission or that a buffer has been added to receive a    */
  33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  34 /* driver that a packet transmission has been completed or that a packet  */
  35 /* has been received and placed in a waiting buffer.                      */
  36 /*                                                                        */
  37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  39 /* or receive has been completed, the VNIC driver is required to use      */
  40 /* "long term mapping". This entails that large, continuous DMA mapped    */
  41 /* buffers are allocated on driver initialization and these buffers are   */
  42 /* then continuously reused to pass skbs to and from the VNIC server.     */
  43 /*                                                                        */
  44 /**************************************************************************/
  45
  46 #include <linux/module.h>
  47 #include <linux/moduleparam.h>
  48 #include <linux/types.h>
  49 #include <linux/errno.h>
  50 #include <linux/completion.h>
  51 #include <linux/ioport.h>
  52 #include <linux/dma-mapping.h>
  53 #include <linux/kernel.h>
  54 #include <linux/netdevice.h>
  55 #include <linux/etherdevice.h>
  56 #include <linux/skbuff.h>
  57 #include <linux/init.h>
  58 #include <linux/delay.h>
  59 #include <linux/mm.h>
  60 #include <linux/ethtool.h>
  61 #include <linux/proc_fs.h>
  62 #include <linux/in.h>
  63 #include <linux/ip.h>
  64 #include <linux/ipv6.h>
  65 #include <linux/irq.h>
  66 #include <linux/kthread.h>
  67 #include <linux/seq_file.h>
  68 #include <linux/interrupt.h>
  69 #include <net/net_namespace.h>
  70 #include <asm/hvcall.h>
  71 #include <linux/atomic.h>
  72 #include <asm/vio.h>
  73 #include <asm/iommu.h>
  74 #include <linux/uaccess.h>
  75 #include <asm/firmware.h>
  76 #include <linux/workqueue.h>
  77 #include <linux/if_vlan.h>
  78
  79 #include "ibmvnic.h"
  80
  81 static const char ibmvnic_driver_name[] = "ibmvnic";
  82 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  83
  84 MODULE_AUTHOR("Santiago Leon <santi_leon@yahoo.com>");
  85 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  86 MODULE_LICENSE("GPL");
  87 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  88
  89 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  90 static int ibmvnic_remove(struct vio_dev *);
  91 static void release_sub_crqs(struct ibmvnic_adapter *);
  92 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  93 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  94 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  95 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  96 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  97                        union sub_crq *sub_crq);
  98 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
  99 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 100 static int enable_scrq_irq(struct ibmvnic_adapter *,
 101                            struct ibmvnic_sub_crq_queue *);
 102 static int disable_scrq_irq(struct ibmvnic_adapter *,
 103                             struct ibmvnic_sub_crq_queue *);
 104 static int pending_scrq(struct ibmvnic_adapter *,
 105                         struct ibmvnic_sub_crq_queue *);
 106 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 107                                         struct ibmvnic_sub_crq_queue *);
 108 static int ibmvnic_poll(struct napi_struct *napi, int data);
 109 static void send_map_query(struct ibmvnic_adapter *adapter);
 110 static void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 111 static void send_request_unmap(struct ibmvnic_adapter *, u8);
 112 static void send_login(struct ibmvnic_adapter *adapter);
 113 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 114 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 115 static int ibmvnic_init(struct ibmvnic_adapter *);
 116 static void release_crq_queue(struct ibmvnic_adapter *);
 117
 118 struct ibmvnic_stat {
 119         char name[ETH_GSTRING_LEN];
 120         int offset;
 121 };
 122
 123 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 124                              offsetof(struct ibmvnic_statistics, stat))
 125 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 126
 127 static const struct ibmvnic_stat ibmvnic_stats[] = {
 128         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 129         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 130         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 131         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 132         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 133         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 134         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 135         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 136         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 137         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 138         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 139         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 140         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 141         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 142         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 143         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 144         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 145         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 146         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 147         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 148         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 149         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 150 };
 151
 152 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 153                           unsigned long length, unsigned long *number,
 154                           unsigned long *irq)
 155 {
 156         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 157         long rc;
 158
 159         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 160         *number = retbuf[0];
 161         *irq = retbuf[1];
 162
 163         return rc;
 164 }
 165
 166 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 167                                 struct ibmvnic_long_term_buff *ltb, int size)
 168 {
 169         struct device *dev = &adapter->vdev->dev;
 170
 171         ltb->size = size;
 172         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 173                                        GFP_KERNEL);
 174
 175         if (!ltb->buff) {
 176                 dev_err(dev, "Couldn't alloc long term buffer\n");
 177                 return -ENOMEM;
 178         }
 179         ltb->map_id = adapter->map_id;
 180         adapter->map_id++;
 181
 182         init_completion(&adapter->fw_done);
 183         send_request_map(adapter, ltb->addr,
 184                          ltb->size, ltb->map_id);
 185         wait_for_completion(&adapter->fw_done);
 186         return 0;
 187 }
 188
 189 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 190                                 struct ibmvnic_long_term_buff *ltb)
 191 {
 192         struct device *dev = &adapter->vdev->dev;
 193
 194         if (!ltb->buff)
 195                 return;
 196
 197         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 198             adapter->reset_reason != VNIC_RESET_MOBILITY)
 199                 send_request_unmap(adapter, ltb->map_id);
 200         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 201 }
 202
 203 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 204                               struct ibmvnic_rx_pool *pool)
 205 {
 206         int count = pool->size - atomic_read(&pool->available);
 207         struct device *dev = &adapter->vdev->dev;
 208         int buffers_added = 0;
 209         unsigned long lpar_rc;
 210         union sub_crq sub_crq;
 211         struct sk_buff *skb;
 212         unsigned int offset;
 213         dma_addr_t dma_addr;
 214         unsigned char *dst;
 215         u64 *handle_array;
 216         int shift = 0;
 217         int index;
 218         int i;
 219
 220         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 221                                       be32_to_cpu(adapter->login_rsp_buf->
 222                                       off_rxadd_subcrqs));
 223
 224         for (i = 0; i < count; ++i) {
 225                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 226                 if (!skb) {
 227                         dev_err(dev, "Couldn't replenish rx buff\n");
 228                         adapter->replenish_no_mem++;
 229                         break;
 230                 }
 231
 232                 index = pool->free_map[pool->next_free];
 233
 234                 if (pool->rx_buff[index].skb)
 235                         dev_err(dev, "Inconsistent free_map!\n");
 236
 237                 /* Copy the skb to the long term mapped DMA buffer */
 238                 offset = index * pool->buff_size;
 239                 dst = pool->long_term_buff.buff + offset;
 240                 memset(dst, 0, pool->buff_size);
 241                 dma_addr = pool->long_term_buff.addr + offset;
 242                 pool->rx_buff[index].data = dst;
 243
 244                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 245                 pool->rx_buff[index].dma = dma_addr;
 246                 pool->rx_buff[index].skb = skb;
 247                 pool->rx_buff[index].pool_index = pool->index;
 248                 pool->rx_buff[index].size = pool->buff_size;
 249
 250                 memset(&sub_crq, 0, sizeof(sub_crq));
 251                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 252                 sub_crq.rx_add.correlator =
 253                     cpu_to_be64((u64)&pool->rx_buff[index]);
 254                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 255                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 256
 257                 /* The length field of the sCRQ is defined to be 24 bits so the
 258                  * buffer size needs to be left shifted by a byte before it is
 259                  * converted to big endian to prevent the last byte from being
 260                  * truncated.
 261                  */
 262 #ifdef __LITTLE_ENDIAN__
 263                 shift = 8;
 264 #endif
 265                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 266
 267                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 268                                       &sub_crq);
 269                 if (lpar_rc != H_SUCCESS)
 270                         goto failure;
 271
 272                 buffers_added++;
 273                 adapter->replenish_add_buff_success++;
 274                 pool->next_free = (pool->next_free + 1) % pool->size;
 275         }
 276         atomic_add(buffers_added, &pool->available);
 277         return;
 278
 279 failure:
 280         dev_info(dev, "replenish pools failure\n");
 281         pool->free_map[pool->next_free] = index;
 282         pool->rx_buff[index].skb = NULL;
 283         if (!dma_mapping_error(dev, dma_addr))
 284                 dma_unmap_single(dev, dma_addr, pool->buff_size,
 285                                  DMA_FROM_DEVICE);
 286
 287         dev_kfree_skb_any(skb);
 288         adapter->replenish_add_buff_failure++;
 289         atomic_add(buffers_added, &pool->available);
 290 }
 291
 292 static void replenish_pools(struct ibmvnic_adapter *adapter)
 293 {
 294         int i;
 295
 296         adapter->replenish_task_cycles++;
 297         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 298              i++) {
 299                 if (adapter->rx_pool[i].active)
 300                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 301         }
 302 }
 303
 304 static void release_stats_token(struct ibmvnic_adapter *adapter)
 305 {
 306         struct device *dev = &adapter->vdev->dev;
 307
 308         if (!adapter->stats_token)
 309                 return;
 310
 311         dma_unmap_single(dev, adapter->stats_token,
 312                          sizeof(struct ibmvnic_statistics),
 313                          DMA_FROM_DEVICE);
 314         adapter->stats_token = 0;
 315 }
 316
 317 static int init_stats_token(struct ibmvnic_adapter *adapter)
 318 {
 319         struct device *dev = &adapter->vdev->dev;
 320         dma_addr_t stok;
 321
 322         stok = dma_map_single(dev, &adapter->stats,
 323                               sizeof(struct ibmvnic_statistics),
 324                               DMA_FROM_DEVICE);
 325         if (dma_mapping_error(dev, stok)) {
 326                 dev_err(dev, "Couldn't map stats buffer\n");
 327                 return -1;
 328         }
 329
 330         adapter->stats_token = stok;
 331         return 0;
 332 }
 333
 334 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 335 {
 336         struct ibmvnic_rx_pool *rx_pool;
 337         int rx_scrqs;
 338         int i, j;
 339
 340         if (!adapter->rx_pool)
 341                 return;
 342
 343         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 344         for (i = 0; i < rx_scrqs; i++) {
 345                 rx_pool = &adapter->rx_pool[i];
 346
 347                 kfree(rx_pool->free_map);
 348                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 349
 350                 if (!rx_pool->rx_buff)
 351                         continue;
 352
 353                 for (j = 0; j < rx_pool->size; j++) {
 354                         if (rx_pool->rx_buff[j].skb) {
 355                                 dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
 356                                 rx_pool->rx_buff[i].skb = NULL;
 357                         }
 358                 }
 359
 360                 kfree(rx_pool->rx_buff);
 361         }
 362
 363         kfree(adapter->rx_pool);
 364         adapter->rx_pool = NULL;
 365 }
 366
 367 static int init_rx_pools(struct net_device *netdev)
 368 {
 369         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 370         struct device *dev = &adapter->vdev->dev;
 371         struct ibmvnic_rx_pool *rx_pool;
 372         int rxadd_subcrqs;
 373         u64 *size_array;
 374         int i, j;
 375
 376         rxadd_subcrqs =
 377                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 378         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 379                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 380
 381         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 382                                    sizeof(struct ibmvnic_rx_pool),
 383                                    GFP_KERNEL);
 384         if (!adapter->rx_pool) {
 385                 dev_err(dev, "Failed to allocate rx pools\n");
 386                 return -1;
 387         }
 388
 389         for (i = 0; i < rxadd_subcrqs; i++) {
 390                 rx_pool = &adapter->rx_pool[i];
 391
 392                 netdev_dbg(adapter->netdev,
 393                            "Initializing rx_pool %d, %lld buffs, %lld bytes each\n",
 394                            i, adapter->req_rx_add_entries_per_subcrq,
 395                            be64_to_cpu(size_array[i]));
 396
 397                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 398                 rx_pool->index = i;
 399                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 400                 rx_pool->active = 1;
 401
 402                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 403                                             GFP_KERNEL);
 404                 if (!rx_pool->free_map) {
 405                         release_rx_pools(adapter);
 406                         return -1;
 407                 }
 408
 409                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 410                                            sizeof(struct ibmvnic_rx_buff),
 411                                            GFP_KERNEL);
 412                 if (!rx_pool->rx_buff) {
 413                         dev_err(dev, "Couldn't alloc rx buffers\n");
 414                         release_rx_pools(adapter);
 415                         return -1;
 416                 }
 417
 418                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 419                                          rx_pool->size * rx_pool->buff_size)) {
 420                         release_rx_pools(adapter);
 421                         return -1;
 422                 }
 423
 424                 for (j = 0; j < rx_pool->size; ++j)
 425                         rx_pool->free_map[j] = j;
 426
 427                 atomic_set(&rx_pool->available, 0);
 428                 rx_pool->next_alloc = 0;
 429                 rx_pool->next_free = 0;
 430         }
 431
 432         return 0;
 433 }
 434
 435 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 436 {
 437         struct ibmvnic_tx_pool *tx_pool;
 438         int i, tx_scrqs;
 439
 440         if (!adapter->tx_pool)
 441                 return;
 442
 443         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 444         for (i = 0; i < tx_scrqs; i++) {
 445                 tx_pool = &adapter->tx_pool[i];
 446                 kfree(tx_pool->tx_buff);
 447                 free_long_term_buff(adapter, &tx_pool->long_term_buff);
 448                 kfree(tx_pool->free_map);
 449         }
 450
 451         kfree(adapter->tx_pool);
 452         adapter->tx_pool = NULL;
 453 }
 454
 455 static int init_tx_pools(struct net_device *netdev)
 456 {
 457         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 458         struct device *dev = &adapter->vdev->dev;
 459         struct ibmvnic_tx_pool *tx_pool;
 460         int tx_subcrqs;
 461         int i, j;
 462
 463         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 464         adapter->tx_pool = kcalloc(tx_subcrqs,
 465                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 466         if (!adapter->tx_pool)
 467                 return -1;
 468
 469         for (i = 0; i < tx_subcrqs; i++) {
 470                 tx_pool = &adapter->tx_pool[i];
 471                 tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
 472                                            sizeof(struct ibmvnic_tx_buff),
 473                                            GFP_KERNEL);
 474                 if (!tx_pool->tx_buff) {
 475                         dev_err(dev, "tx pool buffer allocation failed\n");
 476                         release_tx_pools(adapter);
 477                         return -1;
 478                 }
 479
 480                 if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 481                                          adapter->req_tx_entries_per_subcrq *
 482                                          adapter->req_mtu)) {
 483                         release_tx_pools(adapter);
 484                         return -1;
 485                 }
 486
 487                 tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
 488                                             sizeof(int), GFP_KERNEL);
 489                 if (!tx_pool->free_map) {
 490                         release_tx_pools(adapter);
 491                         return -1;
 492                 }
 493
 494                 for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
 495                         tx_pool->free_map[j] = j;
 496
 497                 tx_pool->consumer_index = 0;
 498                 tx_pool->producer_index = 0;
 499         }
 500
 501         return 0;
 502 }
 503
 504 static void release_error_buffers(struct ibmvnic_adapter *adapter)
 505 {
 506         struct device *dev = &adapter->vdev->dev;
 507         struct ibmvnic_error_buff *error_buff, *tmp;
 508         unsigned long flags;
 509
 510         spin_lock_irqsave(&adapter->error_list_lock, flags);
 511         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
 512                 list_del(&error_buff->list);
 513                 dma_unmap_single(dev, error_buff->dma, error_buff->len,
 514                                  DMA_FROM_DEVICE);
 515                 kfree(error_buff->buff);
 516                 kfree(error_buff);
 517         }
 518         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
 519 }
 520
 521 static int ibmvnic_login(struct net_device *netdev)
 522 {
 523         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 524         unsigned long timeout = msecs_to_jiffies(30000);
 525         struct device *dev = &adapter->vdev->dev;
 526
 527         do {
 528                 if (adapter->renegotiate) {
 529                         adapter->renegotiate = false;
 530                         release_sub_crqs(adapter);
 531
 532                         reinit_completion(&adapter->init_done);
 533                         send_cap_queries(adapter);
 534                         if (!wait_for_completion_timeout(&adapter->init_done,
 535                                                          timeout)) {
 536                                 dev_err(dev, "Capabilities query timeout\n");
 537                                 return -1;
 538                         }
 539                 }
 540
 541                 reinit_completion(&adapter->init_done);
 542                 send_login(adapter);
 543                 if (!wait_for_completion_timeout(&adapter->init_done,
 544                                                  timeout)) {
 545                         dev_err(dev, "Login timeout\n");
 546                         return -1;
 547                 }
 548         } while (adapter->renegotiate);
 549
 550         return 0;
 551 }
 552
 553 static void release_resources(struct ibmvnic_adapter *adapter)
 554 {
 555         int i;
 556
 557         release_tx_pools(adapter);
 558         release_rx_pools(adapter);
 559
 560         release_stats_token(adapter);
 561         release_error_buffers(adapter);
 562
 563         if (adapter->napi) {
 564                 for (i = 0; i < adapter->req_rx_queues; i++) {
 565                         if (&adapter->napi[i])
 566                                 netif_napi_del(&adapter->napi[i]);
 567                 }
 568         }
 569 }
 570
 571 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 572 {
 573         struct net_device *netdev = adapter->netdev;
 574         unsigned long timeout = msecs_to_jiffies(30000);
 575         union ibmvnic_crq crq;
 576         bool resend;
 577         int rc;
 578
 579         netdev_err(netdev, "setting link state %d\n", link_state);
 580         memset(&crq, 0, sizeof(crq));
 581         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 582         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 583         crq.logical_link_state.link_state = link_state;
 584
 585         do {
 586                 resend = false;
 587
 588                 reinit_completion(&adapter->init_done);
 589                 rc = ibmvnic_send_crq(adapter, &crq);
 590                 if (rc) {
 591                         netdev_err(netdev, "Failed to set link state\n");
 592                         return rc;
 593                 }
 594
 595                 if (!wait_for_completion_timeout(&adapter->init_done,
 596                                                  timeout)) {
 597                         netdev_err(netdev, "timeout setting link state\n");
 598                         return -1;
 599                 }
 600
 601                 if (adapter->init_done_rc == 1) {
 602                         /* Partuial success, delay and re-send */
 603                         mdelay(1000);
 604                         resend = true;
 605                 }
 606         } while (resend);
 607
 608         return 0;
 609 }
 610
 611 static int set_real_num_queues(struct net_device *netdev)
 612 {
 613         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 614         int rc;
 615
 616         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 617         if (rc) {
 618                 netdev_err(netdev, "failed to set the number of tx queues\n");
 619                 return rc;
 620         }
 621
 622         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 623         if (rc)
 624                 netdev_err(netdev, "failed to set the number of rx queues\n");
 625
 626         return rc;
 627 }
 628
 629 static int init_resources(struct ibmvnic_adapter *adapter)
 630 {
 631         struct net_device *netdev = adapter->netdev;
 632         int i, rc;
 633
 634         rc = set_real_num_queues(netdev);
 635         if (rc)
 636                 return rc;
 637
 638         rc = init_sub_crq_irqs(adapter);
 639         if (rc) {
 640                 netdev_err(netdev, "failed to initialize sub crq irqs\n");
 641                 return -1;
 642         }
 643
 644         rc = init_stats_token(adapter);
 645         if (rc)
 646                 return rc;
 647
 648         adapter->map_id = 1;
 649         adapter->napi = kcalloc(adapter->req_rx_queues,
 650                                 sizeof(struct napi_struct), GFP_KERNEL);
 651         if (!adapter->napi)
 652                 return -ENOMEM;
 653
 654         for (i = 0; i < adapter->req_rx_queues; i++) {
 655                 netif_napi_add(netdev, &adapter->napi[i], ibmvnic_poll,
 656                                NAPI_POLL_WEIGHT);
 657         }
 658
 659         send_map_query(adapter);
 660
 661         rc = init_rx_pools(netdev);
 662         if (rc)
 663                 return rc;
 664
 665         rc = init_tx_pools(netdev);
 666         return rc;
 667 }
 668
 669 static int __ibmvnic_open(struct net_device *netdev)
 670 {
 671         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 672         enum vnic_state prev_state = adapter->state;
 673         int i, rc;
 674
 675         adapter->state = VNIC_OPENING;
 676         replenish_pools(adapter);
 677
 678         for (i = 0; i < adapter->req_rx_queues; i++)
 679                 napi_enable(&adapter->napi[i]);
 680
 681         /* We're ready to receive frames, enable the sub-crq interrupts and
 682          * set the logical link state to up
 683          */
 684         for (i = 0; i < adapter->req_rx_queues; i++) {
 685                 if (prev_state == VNIC_CLOSED)
 686                         enable_irq(adapter->rx_scrq[i]->irq);
 687                 else
 688                         enable_scrq_irq(adapter, adapter->rx_scrq[i]);
 689         }
 690
 691         for (i = 0; i < adapter->req_tx_queues; i++) {
 692                 if (prev_state == VNIC_CLOSED)
 693                         enable_irq(adapter->tx_scrq[i]->irq);
 694                 else
 695                         enable_scrq_irq(adapter, adapter->tx_scrq[i]);
 696         }
 697
 698         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
 699         if (rc) {
 700                 for (i = 0; i < adapter->req_rx_queues; i++)
 701                         napi_disable(&adapter->napi[i]);
 702                 release_resources(adapter);
 703                 return rc;
 704         }
 705
 706         netif_tx_start_all_queues(netdev);
 707
 708         if (prev_state == VNIC_CLOSED) {
 709                 for (i = 0; i < adapter->req_rx_queues; i++)
 710                         napi_schedule(&adapter->napi[i]);
 711         }
 712
 713         adapter->state = VNIC_OPEN;
 714         return rc;
 715 }
 716
 717 static int ibmvnic_open(struct net_device *netdev)
 718 {
 719         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 720         int rc;
 721
 722         mutex_lock(&adapter->reset_lock);
 723
 724         if (adapter->state != VNIC_CLOSED) {
 725                 rc = ibmvnic_login(netdev);
 726                 if (rc) {
 727                         mutex_unlock(&adapter->reset_lock);
 728                         return rc;
 729                 }
 730
 731                 rc = init_resources(adapter);
 732                 if (rc) {
 733                         netdev_err(netdev, "failed to initialize resources\n");
 734                         release_resources(adapter);
 735                         mutex_unlock(&adapter->reset_lock);
 736                         return rc;
 737                 }
 738         }
 739
 740         rc = __ibmvnic_open(netdev);
 741         mutex_unlock(&adapter->reset_lock);
 742
 743         return rc;
 744 }
 745
 746 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
 747 {
 748         struct ibmvnic_tx_pool *tx_pool;
 749         u64 tx_entries;
 750         int tx_scrqs;
 751         int i, j;
 752
 753         if (!adapter->tx_pool)
 754                 return;
 755
 756         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 757         tx_entries = adapter->req_tx_entries_per_subcrq;
 758
 759         /* Free any remaining skbs in the tx buffer pools */
 760         for (i = 0; i < tx_scrqs; i++) {
 761                 tx_pool = &adapter->tx_pool[i];
 762                 if (!tx_pool)
 763                         continue;
 764
 765                 for (j = 0; j < tx_entries; j++) {
 766                         if (tx_pool->tx_buff[j].skb) {
 767                                 dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
 768                                 tx_pool->tx_buff[j].skb = NULL;
 769                         }
 770                 }
 771         }
 772 }
 773
 774 static int __ibmvnic_close(struct net_device *netdev)
 775 {
 776         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 777         int rc = 0;
 778         int i;
 779
 780         adapter->state = VNIC_CLOSING;
 781         netif_tx_stop_all_queues(netdev);
 782
 783         if (adapter->napi) {
 784                 for (i = 0; i < adapter->req_rx_queues; i++)
 785                         napi_disable(&adapter->napi[i]);
 786         }
 787
 788         clean_tx_pools(adapter);
 789
 790         if (adapter->tx_scrq) {
 791                 for (i = 0; i < adapter->req_tx_queues; i++)
 792                         if (adapter->tx_scrq[i]->irq)
 793                                 disable_irq(adapter->tx_scrq[i]->irq);
 794         }
 795
 796         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
 797         if (rc)
 798                 return rc;
 799
 800         if (adapter->rx_scrq) {
 801                 for (i = 0; i < adapter->req_rx_queues; i++) {
 802                         int retries = 10;
 803
 804                         while (pending_scrq(adapter, adapter->rx_scrq[i])) {
 805                                 retries--;
 806                                 mdelay(100);
 807
 808                                 if (retries == 0)
 809                                         break;
 810                         }
 811
 812                         if (adapter->rx_scrq[i]->irq)
 813                                 disable_irq(adapter->rx_scrq[i]->irq);
 814                 }
 815         }
 816
 817         adapter->state = VNIC_CLOSED;
 818         return rc;
 819 }
 820
 821 static int ibmvnic_close(struct net_device *netdev)
 822 {
 823         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 824         int rc;
 825
 826         mutex_lock(&adapter->reset_lock);
 827         rc = __ibmvnic_close(netdev);
 828         mutex_unlock(&adapter->reset_lock);
 829
 830         return rc;
 831 }
 832
 833 /**
 834  * build_hdr_data - creates L2/L3/L4 header data buffer
 835  * @hdr_field - bitfield determining needed headers
 836  * @skb - socket buffer
 837  * @hdr_len - array of header lengths
 838  * @tot_len - total length of data
 839  *
 840  * Reads hdr_field to determine which headers are needed by firmware.
 841  * Builds a buffer containing these headers.  Saves individual header
 842  * lengths and total buffer length to be used to build descriptors.
 843  */
 844 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
 845                           int *hdr_len, u8 *hdr_data)
 846 {
 847         int len = 0;
 848         u8 *hdr;
 849
 850         hdr_len[0] = sizeof(struct ethhdr);
 851
 852         if (skb->protocol == htons(ETH_P_IP)) {
 853                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
 854                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
 855                         hdr_len[2] = tcp_hdrlen(skb);
 856                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
 857                         hdr_len[2] = sizeof(struct udphdr);
 858         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 859                 hdr_len[1] = sizeof(struct ipv6hdr);
 860                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
 861                         hdr_len[2] = tcp_hdrlen(skb);
 862                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
 863                         hdr_len[2] = sizeof(struct udphdr);
 864         }
 865
 866         memset(hdr_data, 0, 120);
 867         if ((hdr_field >> 6) & 1) {
 868                 hdr = skb_mac_header(skb);
 869                 memcpy(hdr_data, hdr, hdr_len[0]);
 870                 len += hdr_len[0];
 871         }
 872
 873         if ((hdr_field >> 5) & 1) {
 874                 hdr = skb_network_header(skb);
 875                 memcpy(hdr_data + len, hdr, hdr_len[1]);
 876                 len += hdr_len[1];
 877         }
 878
 879         if ((hdr_field >> 4) & 1) {
 880                 hdr = skb_transport_header(skb);
 881                 memcpy(hdr_data + len, hdr, hdr_len[2]);
 882                 len += hdr_len[2];
 883         }
 884         return len;
 885 }
 886
 887 /**
 888  * create_hdr_descs - create header and header extension descriptors
 889  * @hdr_field - bitfield determining needed headers
 890  * @data - buffer containing header data
 891  * @len - length of data buffer
 892  * @hdr_len - array of individual header lengths
 893  * @scrq_arr - descriptor array
 894  *
 895  * Creates header and, if needed, header extension descriptors and
 896  * places them in a descriptor array, scrq_arr
 897  */
 898
 899 static void create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
 900                              union sub_crq *scrq_arr)
 901 {
 902         union sub_crq hdr_desc;
 903         int tmp_len = len;
 904         u8 *data, *cur;
 905         int tmp;
 906
 907         while (tmp_len > 0) {
 908                 cur = hdr_data + len - tmp_len;
 909
 910                 memset(&hdr_desc, 0, sizeof(hdr_desc));
 911                 if (cur != hdr_data) {
 912                         data = hdr_desc.hdr_ext.data;
 913                         tmp = tmp_len > 29 ? 29 : tmp_len;
 914                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
 915                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
 916                         hdr_desc.hdr_ext.len = tmp;
 917                 } else {
 918                         data = hdr_desc.hdr.data;
 919                         tmp = tmp_len > 24 ? 24 : tmp_len;
 920                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
 921                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
 922                         hdr_desc.hdr.len = tmp;
 923                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
 924                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
 925                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
 926                         hdr_desc.hdr.flag = hdr_field << 1;
 927                 }
 928                 memcpy(data, cur, tmp);
 929                 tmp_len -= tmp;
 930                 *scrq_arr = hdr_desc;
 931                 scrq_arr++;
 932         }
 933 }
 934
 935 /**
 936  * build_hdr_descs_arr - build a header descriptor array
 937  * @skb - socket buffer
 938  * @num_entries - number of descriptors to be sent
 939  * @subcrq - first TX descriptor
 940  * @hdr_field - bit field determining which headers will be sent
 941  *
 942  * This function will build a TX descriptor array with applicable
 943  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
 944  */
 945
 946 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
 947                                 int *num_entries, u8 hdr_field)
 948 {
 949         int hdr_len[3] = {0, 0, 0};
 950         int tot_len, len;
 951         u8 *hdr_data = txbuff->hdr_data;
 952
 953         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
 954                                  txbuff->hdr_data);
 955         len = tot_len;
 956         len -= 24;
 957         if (len > 0)
 958                 num_entries += len % 29 ? len / 29 + 1 : len / 29;
 959         create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
 960                          txbuff->indir_arr + 1);
 961 }
 962
 963 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
 964 {
 965         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 966         int queue_num = skb_get_queue_mapping(skb);
 967         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
 968         struct device *dev = &adapter->vdev->dev;
 969         struct ibmvnic_tx_buff *tx_buff = NULL;
 970         struct ibmvnic_sub_crq_queue *tx_scrq;
 971         struct ibmvnic_tx_pool *tx_pool;
 972         unsigned int tx_send_failed = 0;
 973         unsigned int tx_map_failed = 0;
 974         unsigned int tx_dropped = 0;
 975         unsigned int tx_packets = 0;
 976         unsigned int tx_bytes = 0;
 977         dma_addr_t data_dma_addr;
 978         struct netdev_queue *txq;
 979         unsigned long lpar_rc;
 980         union sub_crq tx_crq;
 981         unsigned int offset;
 982         int num_entries = 1;
 983         unsigned char *dst;
 984         u64 *handle_array;
 985         int index = 0;
 986         int ret = 0;
 987
 988         if (adapter->resetting) {
 989                 if (!netif_subqueue_stopped(netdev, skb))
 990                         netif_stop_subqueue(netdev, queue_num);
 991                 dev_kfree_skb_any(skb);
 992
 993                 tx_send_failed++;
 994                 tx_dropped++;
 995                 ret = NETDEV_TX_OK;
 996                 goto out;
 997         }
 998
 999         tx_pool = &adapter->tx_pool[queue_num];
1000         tx_scrq = adapter->tx_scrq[queue_num];
1001         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1002         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1003                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1004
1005         index = tx_pool->free_map[tx_pool->consumer_index];
1006         offset = index * adapter->req_mtu;
1007         dst = tx_pool->long_term_buff.buff + offset;
1008         memset(dst, 0, adapter->req_mtu);
1009         skb_copy_from_linear_data(skb, dst, skb->len);
1010         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1011
1012         tx_pool->consumer_index =
1013             (tx_pool->consumer_index + 1) %
1014                 adapter->req_tx_entries_per_subcrq;
1015
1016         tx_buff = &tx_pool->tx_buff[index];
1017         tx_buff->skb = skb;
1018         tx_buff->data_dma[0] = data_dma_addr;
1019         tx_buff->data_len[0] = skb->len;
1020         tx_buff->index = index;
1021         tx_buff->pool_index = queue_num;
1022         tx_buff->last_frag = true;
1023
1024         memset(&tx_crq, 0, sizeof(tx_crq));
1025         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1026         tx_crq.v1.type = IBMVNIC_TX_DESC;
1027         tx_crq.v1.n_crq_elem = 1;
1028         tx_crq.v1.n_sge = 1;
1029         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1030         tx_crq.v1.correlator = cpu_to_be32(index);
1031         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1032         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1033         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1034
1035         if (adapter->vlan_header_insertion) {
1036                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1037                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1038         }
1039
1040         if (skb->protocol == htons(ETH_P_IP)) {
1041                 if (ip_hdr(skb)->version == 4)
1042                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1043                 else if (ip_hdr(skb)->version == 6)
1044                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1045
1046                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1047                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1048                 else if (ip_hdr(skb)->protocol != IPPROTO_TCP)
1049                         tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1050         }
1051
1052         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1053                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1054                 hdrs += 2;
1055         }
1056         /* determine if l2/3/4 headers are sent to firmware */
1057         if ((*hdrs >> 7) & 1 &&
1058             (skb->protocol == htons(ETH_P_IP) ||
1059              skb->protocol == htons(ETH_P_IPV6))) {
1060                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1061                 tx_crq.v1.n_crq_elem = num_entries;
1062                 tx_buff->indir_arr[0] = tx_crq;
1063                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1064                                                     sizeof(tx_buff->indir_arr),
1065                                                     DMA_TO_DEVICE);
1066                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1067                         dev_kfree_skb_any(skb);
1068                         tx_buff->skb = NULL;
1069                         if (!firmware_has_feature(FW_FEATURE_CMO))
1070                                 dev_err(dev, "tx: unable to map descriptor array\n");
1071                         tx_map_failed++;
1072                         tx_dropped++;
1073                         ret = NETDEV_TX_OK;
1074                         goto out;
1075                 }
1076                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1077                                                (u64)tx_buff->indir_dma,
1078                                                (u64)num_entries);
1079         } else {
1080                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1081                                       &tx_crq);
1082         }
1083         if (lpar_rc != H_SUCCESS) {
1084                 dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1085
1086                 if (tx_pool->consumer_index == 0)
1087                         tx_pool->consumer_index =
1088                                 adapter->req_tx_entries_per_subcrq - 1;
1089                 else
1090                         tx_pool->consumer_index--;
1091
1092                 dev_kfree_skb_any(skb);
1093                 tx_buff->skb = NULL;
1094
1095                 if (lpar_rc == H_CLOSED)
1096                         netif_stop_subqueue(netdev, queue_num);
1097
1098                 tx_send_failed++;
1099                 tx_dropped++;
1100                 ret = NETDEV_TX_OK;
1101                 goto out;
1102         }
1103
1104         if (atomic_inc_return(&tx_scrq->used)
1105                                         >= adapter->req_tx_entries_per_subcrq) {
1106                 netdev_info(netdev, "Stopping queue %d\n", queue_num);
1107                 netif_stop_subqueue(netdev, queue_num);
1108         }
1109
1110         tx_packets++;
1111         tx_bytes += skb->len;
1112         txq->trans_start = jiffies;
1113         ret = NETDEV_TX_OK;
1114
1115 out:
1116         netdev->stats.tx_dropped += tx_dropped;
1117         netdev->stats.tx_bytes += tx_bytes;
1118         netdev->stats.tx_packets += tx_packets;
1119         adapter->tx_send_failed += tx_send_failed;
1120         adapter->tx_map_failed += tx_map_failed;
1121
1122         return ret;
1123 }
1124
1125 static void ibmvnic_set_multi(struct net_device *netdev)
1126 {
1127         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1128         struct netdev_hw_addr *ha;
1129         union ibmvnic_crq crq;
1130
1131         memset(&crq, 0, sizeof(crq));
1132         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1133         crq.request_capability.cmd = REQUEST_CAPABILITY;
1134
1135         if (netdev->flags & IFF_PROMISC) {
1136                 if (!adapter->promisc_supported)
1137                         return;
1138         } else {
1139                 if (netdev->flags & IFF_ALLMULTI) {
1140                         /* Accept all multicast */
1141                         memset(&crq, 0, sizeof(crq));
1142                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1143                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1144                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1145                         ibmvnic_send_crq(adapter, &crq);
1146                 } else if (netdev_mc_empty(netdev)) {
1147                         /* Reject all multicast */
1148                         memset(&crq, 0, sizeof(crq));
1149                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1150                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1151                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1152                         ibmvnic_send_crq(adapter, &crq);
1153                 } else {
1154                         /* Accept one or more multicast(s) */
1155                         netdev_for_each_mc_addr(ha, netdev) {
1156                                 memset(&crq, 0, sizeof(crq));
1157                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1158                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1159                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1160                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1161                                                 ha->addr);
1162                                 ibmvnic_send_crq(adapter, &crq);
1163                         }
1164                 }
1165         }
1166 }
1167
1168 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1169 {
1170         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1171         struct sockaddr *addr = p;
1172         union ibmvnic_crq crq;
1173
1174         if (!is_valid_ether_addr(addr->sa_data))
1175                 return -EADDRNOTAVAIL;
1176
1177         memset(&crq, 0, sizeof(crq));
1178         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1179         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1180         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1181         ibmvnic_send_crq(adapter, &crq);
1182         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1183         return 0;
1184 }
1185
1186 /**
1187  * do_reset returns zero if we are able to keep processing reset events, or
1188  * non-zero if we hit a fatal error and must halt.
1189  */
1190 static int do_reset(struct ibmvnic_adapter *adapter,
1191                     struct ibmvnic_rwi *rwi, u32 reset_state)
1192 {
1193         struct net_device *netdev = adapter->netdev;
1194         int i, rc;
1195
1196         netif_carrier_off(netdev);
1197         adapter->reset_reason = rwi->reset_reason;
1198
1199         if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1200                 rc = ibmvnic_reenable_crq_queue(adapter);
1201                 if (rc)
1202                         return 0;
1203         }
1204
1205         rc = __ibmvnic_close(netdev);
1206         if (rc)
1207                 return rc;
1208
1209         /* remove the closed state so when we call open it appears
1210          * we are coming from the probed state.
1211          */
1212         adapter->state = VNIC_PROBED;
1213
1214         release_resources(adapter);
1215         release_sub_crqs(adapter);
1216         release_crq_queue(adapter);
1217
1218         rc = ibmvnic_init(adapter);
1219         if (rc)
1220                 return 0;
1221
1222         /* If the adapter was in PROBE state prior to the reset, exit here. */
1223         if (reset_state == VNIC_PROBED)
1224                 return 0;
1225
1226         rc = ibmvnic_login(netdev);
1227         if (rc) {
1228                 adapter->state = VNIC_PROBED;
1229                 return 0;
1230         }
1231
1232         rtnl_lock();
1233         rc = init_resources(adapter);
1234         rtnl_unlock();
1235         if (rc)
1236                 return rc;
1237
1238         if (reset_state == VNIC_CLOSED)
1239                 return 0;
1240
1241         rc = __ibmvnic_open(netdev);
1242         if (rc) {
1243                 if (list_empty(&adapter->rwi_list))
1244                         adapter->state = VNIC_CLOSED;
1245                 else
1246                         adapter->state = reset_state;
1247
1248                 return 0;
1249         }
1250
1251         netif_carrier_on(netdev);
1252
1253         /* kick napi */
1254         for (i = 0; i < adapter->req_rx_queues; i++)
1255                 napi_schedule(&adapter->napi[i]);
1256
1257         return 0;
1258 }
1259
1260 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1261 {
1262         struct ibmvnic_rwi *rwi;
1263
1264         mutex_lock(&adapter->rwi_lock);
1265
1266         if (!list_empty(&adapter->rwi_list)) {
1267                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1268                                        list);
1269                 list_del(&rwi->list);
1270         } else {
1271                 rwi = NULL;
1272         }
1273
1274         mutex_unlock(&adapter->rwi_lock);
1275         return rwi;
1276 }
1277
1278 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1279 {
1280         struct ibmvnic_rwi *rwi;
1281
1282         rwi = get_next_rwi(adapter);
1283         while (rwi) {
1284                 kfree(rwi);
1285                 rwi = get_next_rwi(adapter);
1286         }
1287 }
1288
1289 static void __ibmvnic_reset(struct work_struct *work)
1290 {
1291         struct ibmvnic_rwi *rwi;
1292         struct ibmvnic_adapter *adapter;
1293         struct net_device *netdev;
1294         u32 reset_state;
1295         int rc;
1296
1297         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1298         netdev = adapter->netdev;
1299
1300         mutex_lock(&adapter->reset_lock);
1301         adapter->resetting = true;
1302         reset_state = adapter->state;
1303
1304         rwi = get_next_rwi(adapter);
1305         while (rwi) {
1306                 rc = do_reset(adapter, rwi, reset_state);
1307                 kfree(rwi);
1308                 if (rc)
1309                         break;
1310
1311                 rwi = get_next_rwi(adapter);
1312         }
1313
1314         if (rc) {
1315                 free_all_rwi(adapter);
1316                 return;
1317         }
1318
1319         adapter->resetting = false;
1320         mutex_unlock(&adapter->reset_lock);
1321 }
1322
1323 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1324                           enum ibmvnic_reset_reason reason)
1325 {
1326         struct ibmvnic_rwi *rwi, *tmp;
1327         struct net_device *netdev = adapter->netdev;
1328         struct list_head *entry;
1329
1330         if (adapter->state == VNIC_REMOVING ||
1331             adapter->state == VNIC_REMOVED) {
1332                 netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1333                 return;
1334         }
1335
1336         mutex_lock(&adapter->rwi_lock);
1337
1338         list_for_each(entry, &adapter->rwi_list) {
1339                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
1340                 if (tmp->reset_reason == reason) {
1341                         netdev_err(netdev, "Matching reset found, skipping\n");
1342                         mutex_unlock(&adapter->rwi_lock);
1343                         return;
1344                 }
1345         }
1346
1347         rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1348         if (!rwi) {
1349                 mutex_unlock(&adapter->rwi_lock);
1350                 ibmvnic_close(netdev);
1351                 return;
1352         }
1353
1354         rwi->reset_reason = reason;
1355         list_add_tail(&rwi->list, &adapter->rwi_list);
1356         mutex_unlock(&adapter->rwi_lock);
1357         schedule_work(&adapter->ibmvnic_reset);
1358 }
1359
1360 static void ibmvnic_tx_timeout(struct net_device *dev)
1361 {
1362         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1363
1364         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1365 }
1366
1367 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1368                                   struct ibmvnic_rx_buff *rx_buff)
1369 {
1370         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1371
1372         rx_buff->skb = NULL;
1373
1374         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1375         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1376
1377         atomic_dec(&pool->available);
1378 }
1379
1380 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1381 {
1382         struct net_device *netdev = napi->dev;
1383         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1384         int scrq_num = (int)(napi - adapter->napi);
1385         int frames_processed = 0;
1386 restart_poll:
1387         while (frames_processed < budget) {
1388                 struct sk_buff *skb;
1389                 struct ibmvnic_rx_buff *rx_buff;
1390                 union sub_crq *next;
1391                 u32 length;
1392                 u16 offset;
1393                 u8 flags = 0;
1394
1395                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1396                         break;
1397                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1398                 rx_buff =
1399                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1400                                                           rx_comp.correlator);
1401                 /* do error checking */
1402                 if (next->rx_comp.rc) {
1403                         netdev_err(netdev, "rx error %x\n", next->rx_comp.rc);
1404                         /* free the entry */
1405                         next->rx_comp.first = 0;
1406                         remove_buff_from_pool(adapter, rx_buff);
1407                         break;
1408                 }
1409
1410                 length = be32_to_cpu(next->rx_comp.len);
1411                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
1412                 flags = next->rx_comp.flags;
1413                 skb = rx_buff->skb;
1414                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
1415                                         length);
1416
1417                 /* VLAN Header has been stripped by the system firmware and
1418                  * needs to be inserted by the driver
1419                  */
1420                 if (adapter->rx_vlan_header_insertion &&
1421                     (flags & IBMVNIC_VLAN_STRIPPED))
1422                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1423                                                ntohs(next->rx_comp.vlan_tci));
1424
1425                 /* free the entry */
1426                 next->rx_comp.first = 0;
1427                 remove_buff_from_pool(adapter, rx_buff);
1428
1429                 skb_put(skb, length);
1430                 skb->protocol = eth_type_trans(skb, netdev);
1431
1432                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1433                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1434                         skb->ip_summed = CHECKSUM_UNNECESSARY;
1435                 }
1436
1437                 length = skb->len;
1438                 napi_gro_receive(napi, skb); /* send it up */
1439                 netdev->stats.rx_packets++;
1440                 netdev->stats.rx_bytes += length;
1441                 frames_processed++;
1442         }
1443         replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1444
1445         if (frames_processed < budget) {
1446                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1447                 napi_complete_done(napi, frames_processed);
1448                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1449                     napi_reschedule(napi)) {
1450                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1451                         goto restart_poll;
1452                 }
1453         }
1454         return frames_processed;
1455 }
1456
1457 #ifdef CONFIG_NET_POLL_CONTROLLER
1458 static void ibmvnic_netpoll_controller(struct net_device *dev)
1459 {
1460         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1461         int i;
1462
1463         replenish_pools(netdev_priv(dev));
1464         for (i = 0; i < adapter->req_rx_queues; i++)
1465                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1466                                      adapter->rx_scrq[i]);
1467 }
1468 #endif
1469
1470 static const struct net_device_ops ibmvnic_netdev_ops = {
1471         .ndo_open               = ibmvnic_open,
1472         .ndo_stop               = ibmvnic_close,
1473         .ndo_start_xmit         = ibmvnic_xmit,
1474         .ndo_set_rx_mode        = ibmvnic_set_multi,
1475         .ndo_set_mac_address    = ibmvnic_set_mac,
1476         .ndo_validate_addr      = eth_validate_addr,
1477         .ndo_tx_timeout         = ibmvnic_tx_timeout,
1478 #ifdef CONFIG_NET_POLL_CONTROLLER
1479         .ndo_poll_controller    = ibmvnic_netpoll_controller,
1480 #endif
1481 };
1482
1483 /* ethtool functions */
1484
1485 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
1486                                       struct ethtool_link_ksettings *cmd)
1487 {
1488         u32 supported, advertising;
1489
1490         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
1491                           SUPPORTED_FIBRE);
1492         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
1493                             ADVERTISED_FIBRE);
1494         cmd->base.speed = SPEED_1000;
1495         cmd->base.duplex = DUPLEX_FULL;
1496         cmd->base.port = PORT_FIBRE;
1497         cmd->base.phy_address = 0;
1498         cmd->base.autoneg = AUTONEG_ENABLE;
1499
1500         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1501                                                 supported);
1502         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
1503                                                 advertising);
1504
1505         return 0;
1506 }
1507
1508 static void ibmvnic_get_drvinfo(struct net_device *dev,
1509                                 struct ethtool_drvinfo *info)
1510 {
1511         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
1512         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
1513 }
1514
1515 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
1516 {
1517         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1518
1519         return adapter->msg_enable;
1520 }
1521
1522 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
1523 {
1524         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1525
1526         adapter->msg_enable = data;
1527 }
1528
1529 static u32 ibmvnic_get_link(struct net_device *netdev)
1530 {
1531         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1532
1533         /* Don't need to send a query because we request a logical link up at
1534          * init and then we wait for link state indications
1535          */
1536         return adapter->logical_link_state;
1537 }
1538
1539 static void ibmvnic_get_ringparam(struct net_device *netdev,
1540                                   struct ethtool_ringparam *ring)
1541 {
1542         ring->rx_max_pending = 0;
1543         ring->tx_max_pending = 0;
1544         ring->rx_mini_max_pending = 0;
1545         ring->rx_jumbo_max_pending = 0;
1546         ring->rx_pending = 0;
1547         ring->tx_pending = 0;
1548         ring->rx_mini_pending = 0;
1549         ring->rx_jumbo_pending = 0;
1550 }
1551
1552 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
1553 {
1554         int i;
1555
1556         if (stringset != ETH_SS_STATS)
1557                 return;
1558
1559         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
1560                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
1561 }
1562
1563 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
1564 {
1565         switch (sset) {
1566         case ETH_SS_STATS:
1567                 return ARRAY_SIZE(ibmvnic_stats);
1568         default:
1569                 return -EOPNOTSUPP;
1570         }
1571 }
1572
1573 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
1574                                       struct ethtool_stats *stats, u64 *data)
1575 {
1576         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1577         union ibmvnic_crq crq;
1578         int i;
1579
1580         memset(&crq, 0, sizeof(crq));
1581         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
1582         crq.request_statistics.cmd = REQUEST_STATISTICS;
1583         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
1584         crq.request_statistics.len =
1585             cpu_to_be32(sizeof(struct ibmvnic_statistics));
1586
1587         /* Wait for data to be written */
1588         init_completion(&adapter->stats_done);
1589         ibmvnic_send_crq(adapter, &crq);
1590         wait_for_completion(&adapter->stats_done);
1591
1592         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
1593                 data[i] = IBMVNIC_GET_STAT(adapter, ibmvnic_stats[i].offset);
1594 }
1595
1596 static const struct ethtool_ops ibmvnic_ethtool_ops = {
1597         .get_drvinfo            = ibmvnic_get_drvinfo,
1598         .get_msglevel           = ibmvnic_get_msglevel,
1599         .set_msglevel           = ibmvnic_set_msglevel,
1600         .get_link               = ibmvnic_get_link,
1601         .get_ringparam          = ibmvnic_get_ringparam,
1602         .get_strings            = ibmvnic_get_strings,
1603         .get_sset_count         = ibmvnic_get_sset_count,
1604         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
1605         .get_link_ksettings     = ibmvnic_get_link_ksettings,
1606 };
1607
1608 /* Routines for managing CRQs/sCRQs  */
1609
1610 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
1611                                   struct ibmvnic_sub_crq_queue *scrq)
1612 {
1613         struct device *dev = &adapter->vdev->dev;
1614         long rc;
1615
1616         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
1617
1618         /* Close the sub-crqs */
1619         do {
1620                 rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
1621                                         adapter->vdev->unit_address,
1622                                         scrq->crq_num);
1623         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
1624
1625         if (rc) {
1626                 netdev_err(adapter->netdev,
1627                            "Failed to release sub-CRQ %16lx, rc = %ld\n",
1628                            scrq->crq_num, rc);
1629         }
1630
1631         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1632                          DMA_BIDIRECTIONAL);
1633         free_pages((unsigned long)scrq->msgs, 2);
1634         kfree(scrq);
1635 }
1636
1637 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
1638                                                         *adapter)
1639 {
1640         struct device *dev = &adapter->vdev->dev;
1641         struct ibmvnic_sub_crq_queue *scrq;
1642         int rc;
1643
1644         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
1645         if (!scrq)
1646                 return NULL;
1647
1648         scrq->msgs =
1649                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
1650         if (!scrq->msgs) {
1651                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
1652                 goto zero_page_failed;
1653         }
1654
1655         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
1656                                          DMA_BIDIRECTIONAL);
1657         if (dma_mapping_error(dev, scrq->msg_token)) {
1658                 dev_warn(dev, "Couldn't map crq queue messages page\n");
1659                 goto map_failed;
1660         }
1661
1662         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
1663                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
1664
1665         if (rc == H_RESOURCE)
1666                 rc = ibmvnic_reset_crq(adapter);
1667
1668         if (rc == H_CLOSED) {
1669                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
1670         } else if (rc) {
1671                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
1672                 goto reg_failed;
1673         }
1674
1675         scrq->adapter = adapter;
1676         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
1677         spin_lock_init(&scrq->lock);
1678
1679         netdev_dbg(adapter->netdev,
1680                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
1681                    scrq->crq_num, scrq->hw_irq, scrq->irq);
1682
1683         return scrq;
1684
1685 reg_failed:
1686         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1687                          DMA_BIDIRECTIONAL);
1688 map_failed:
1689         free_pages((unsigned long)scrq->msgs, 2);
1690 zero_page_failed:
1691         kfree(scrq);
1692
1693         return NULL;
1694 }
1695
1696 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
1697 {
1698         int i;
1699
1700         if (adapter->tx_scrq) {
1701                 for (i = 0; i < adapter->req_tx_queues; i++) {
1702                         if (!adapter->tx_scrq[i])
1703                                 continue;
1704
1705                         if (adapter->tx_scrq[i]->irq) {
1706                                 free_irq(adapter->tx_scrq[i]->irq,
1707                                          adapter->tx_scrq[i]);
1708                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
1709                                 adapter->tx_scrq[i]->irq = 0;
1710                         }
1711
1712                         release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
1713                 }
1714
1715                 kfree(adapter->tx_scrq);
1716                 adapter->tx_scrq = NULL;
1717         }
1718
1719         if (adapter->rx_scrq) {
1720                 for (i = 0; i < adapter->req_rx_queues; i++) {
1721                         if (!adapter->rx_scrq[i])
1722                                 continue;
1723
1724                         if (adapter->rx_scrq[i]->irq) {
1725                                 free_irq(adapter->rx_scrq[i]->irq,
1726                                          adapter->rx_scrq[i]);
1727                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
1728                                 adapter->rx_scrq[i]->irq = 0;
1729                         }
1730
1731                         release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
1732                 }
1733
1734                 kfree(adapter->rx_scrq);
1735                 adapter->rx_scrq = NULL;
1736         }
1737 }
1738
1739 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
1740                             struct ibmvnic_sub_crq_queue *scrq)
1741 {
1742         struct device *dev = &adapter->vdev->dev;
1743         unsigned long rc;
1744
1745         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
1746                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
1747         if (rc)
1748                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
1749                         scrq->hw_irq, rc);
1750         return rc;
1751 }
1752
1753 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
1754                            struct ibmvnic_sub_crq_queue *scrq)
1755 {
1756         struct device *dev = &adapter->vdev->dev;
1757         unsigned long rc;
1758
1759         if (scrq->hw_irq > 0x100000000ULL) {
1760                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
1761                 return 1;
1762         }
1763
1764         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
1765                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
1766         if (rc)
1767                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
1768                         scrq->hw_irq, rc);
1769         return rc;
1770 }
1771
1772 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
1773                                struct ibmvnic_sub_crq_queue *scrq)
1774 {
1775         struct device *dev = &adapter->vdev->dev;
1776         struct ibmvnic_tx_buff *txbuff;
1777         union sub_crq *next;
1778         int index;
1779         int i, j;
1780         u8 first;
1781
1782 restart_loop:
1783         while (pending_scrq(adapter, scrq)) {
1784                 unsigned int pool = scrq->pool_index;
1785
1786                 next = ibmvnic_next_scrq(adapter, scrq);
1787                 for (i = 0; i < next->tx_comp.num_comps; i++) {
1788                         if (next->tx_comp.rcs[i]) {
1789                                 dev_err(dev, "tx error %x\n",
1790                                         next->tx_comp.rcs[i]);
1791                                 continue;
1792                         }
1793                         index = be32_to_cpu(next->tx_comp.correlators[i]);
1794                         txbuff = &adapter->tx_pool[pool].tx_buff[index];
1795
1796                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
1797                                 if (!txbuff->data_dma[j])
1798                                         continue;
1799
1800                                 txbuff->data_dma[j] = 0;
1801                         }
1802                         /* if sub_crq was sent indirectly */
1803                         first = txbuff->indir_arr[0].generic.first;
1804                         if (first == IBMVNIC_CRQ_CMD) {
1805                                 dma_unmap_single(dev, txbuff->indir_dma,
1806                                                  sizeof(txbuff->indir_arr),
1807                                                  DMA_TO_DEVICE);
1808                         }
1809
1810                         if (txbuff->last_frag) {
1811                                 if (atomic_sub_return(next->tx_comp.num_comps,
1812                                                       &scrq->used) <=
1813                                     (adapter->req_tx_entries_per_subcrq / 2) &&
1814                                     netif_subqueue_stopped(adapter->netdev,
1815                                                            txbuff->skb)) {
1816                                         netif_wake_subqueue(adapter->netdev,
1817                                                             scrq->pool_index);
1818                                         netdev_dbg(adapter->netdev,
1819                                                    "Started queue %d\n",
1820                                                    scrq->pool_index);
1821                                 }
1822
1823                                 dev_kfree_skb_any(txbuff->skb);
1824                         }
1825
1826                         adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
1827                                                      producer_index] = index;
1828                         adapter->tx_pool[pool].producer_index =
1829                             (adapter->tx_pool[pool].producer_index + 1) %
1830                             adapter->req_tx_entries_per_subcrq;
1831                 }
1832                 /* remove tx_comp scrq*/
1833                 next->tx_comp.first = 0;
1834         }
1835
1836         enable_scrq_irq(adapter, scrq);
1837
1838         if (pending_scrq(adapter, scrq)) {
1839                 disable_scrq_irq(adapter, scrq);
1840                 goto restart_loop;
1841         }
1842
1843         return 0;
1844 }
1845
1846 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
1847 {
1848         struct ibmvnic_sub_crq_queue *scrq = instance;
1849         struct ibmvnic_adapter *adapter = scrq->adapter;
1850
1851         disable_scrq_irq(adapter, scrq);
1852         ibmvnic_complete_tx(adapter, scrq);
1853
1854         return IRQ_HANDLED;
1855 }
1856
1857 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
1858 {
1859         struct ibmvnic_sub_crq_queue *scrq = instance;
1860         struct ibmvnic_adapter *adapter = scrq->adapter;
1861
1862         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
1863                 disable_scrq_irq(adapter, scrq);
1864                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
1865         }
1866
1867         return IRQ_HANDLED;
1868 }
1869
1870 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
1871 {
1872         struct device *dev = &adapter->vdev->dev;
1873         struct ibmvnic_sub_crq_queue *scrq;
1874         int i = 0, j = 0;
1875         int rc = 0;
1876
1877         for (i = 0; i < adapter->req_tx_queues; i++) {
1878                 scrq = adapter->tx_scrq[i];
1879                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
1880
1881                 if (!scrq->irq) {
1882                         rc = -EINVAL;
1883                         dev_err(dev, "Error mapping irq\n");
1884                         goto req_tx_irq_failed;
1885                 }
1886
1887                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
1888                                  0, "ibmvnic_tx", scrq);
1889
1890                 if (rc) {
1891                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
1892                                 scrq->irq, rc);
1893                         irq_dispose_mapping(scrq->irq);
1894                         goto req_rx_irq_failed;
1895                 }
1896         }
1897
1898         for (i = 0; i < adapter->req_rx_queues; i++) {
1899                 scrq = adapter->rx_scrq[i];
1900                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
1901                 if (!scrq->irq) {
1902                         rc = -EINVAL;
1903                         dev_err(dev, "Error mapping irq\n");
1904                         goto req_rx_irq_failed;
1905                 }
1906                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
1907                                  0, "ibmvnic_rx", scrq);
1908                 if (rc) {
1909                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
1910                                 scrq->irq, rc);
1911                         irq_dispose_mapping(scrq->irq);
1912                         goto req_rx_irq_failed;
1913                 }
1914         }
1915         return rc;
1916
1917 req_rx_irq_failed:
1918         for (j = 0; j < i; j++) {
1919                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
1920                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
1921         }
1922         i = adapter->req_tx_queues;
1923 req_tx_irq_failed:
1924         for (j = 0; j < i; j++) {
1925                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
1926                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
1927         }
1928         release_sub_crqs(adapter);
1929         return rc;
1930 }
1931
1932 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
1933 {
1934         struct device *dev = &adapter->vdev->dev;
1935         struct ibmvnic_sub_crq_queue **allqueues;
1936         int registered_queues = 0;
1937         int total_queues;
1938         int more = 0;
1939         int i;
1940
1941         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
1942
1943         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
1944         if (!allqueues)
1945                 return -1;
1946
1947         for (i = 0; i < total_queues; i++) {
1948                 allqueues[i] = init_sub_crq_queue(adapter);
1949                 if (!allqueues[i]) {
1950                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
1951                         break;
1952                 }
1953                 registered_queues++;
1954         }
1955
1956         /* Make sure we were able to register the minimum number of queues */
1957         if (registered_queues <
1958             adapter->min_tx_queues + adapter->min_rx_queues) {
1959                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
1960                 goto tx_failed;
1961         }
1962
1963         /* Distribute the failed allocated queues*/
1964         for (i = 0; i < total_queues - registered_queues + more ; i++) {
1965                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
1966                 switch (i % 3) {
1967                 case 0:
1968                         if (adapter->req_rx_queues > adapter->min_rx_queues)
1969                                 adapter->req_rx_queues--;
1970                         else
1971                                 more++;
1972                         break;
1973                 case 1:
1974                         if (adapter->req_tx_queues > adapter->min_tx_queues)
1975                                 adapter->req_tx_queues--;
1976                         else
1977                                 more++;
1978                         break;
1979                 }
1980         }
1981
1982         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
1983                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
1984         if (!adapter->tx_scrq)
1985                 goto tx_failed;
1986
1987         for (i = 0; i < adapter->req_tx_queues; i++) {
1988                 adapter->tx_scrq[i] = allqueues[i];
1989                 adapter->tx_scrq[i]->pool_index = i;
1990         }
1991
1992         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
1993                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
1994         if (!adapter->rx_scrq)
1995                 goto rx_failed;
1996
1997         for (i = 0; i < adapter->req_rx_queues; i++) {
1998                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
1999                 adapter->rx_scrq[i]->scrq_num = i;
2000         }
2001
2002         kfree(allqueues);
2003         return 0;
2004
2005 rx_failed:
2006         kfree(adapter->tx_scrq);
2007         adapter->tx_scrq = NULL;
2008 tx_failed:
2009         for (i = 0; i < registered_queues; i++)
2010                 release_sub_crq_queue(adapter, allqueues[i]);
2011         kfree(allqueues);
2012         return -1;
2013 }
2014
2015 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2016 {
2017         struct device *dev = &adapter->vdev->dev;
2018         union ibmvnic_crq crq;
2019
2020         if (!retry) {
2021                 /* Sub-CRQ entries are 32 byte long */
2022                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2023
2024                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
2025                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
2026                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2027                         return;
2028                 }
2029
2030                 /* Get the minimum between the queried max and the entries
2031                  * that fit in our PAGE_SIZE
2032                  */
2033                 adapter->req_tx_entries_per_subcrq =
2034                     adapter->max_tx_entries_per_subcrq > entries_page ?
2035                     entries_page : adapter->max_tx_entries_per_subcrq;
2036                 adapter->req_rx_add_entries_per_subcrq =
2037                     adapter->max_rx_add_entries_per_subcrq > entries_page ?
2038                     entries_page : adapter->max_rx_add_entries_per_subcrq;
2039
2040                 adapter->req_tx_queues = adapter->opt_tx_comp_sub_queues;
2041                 adapter->req_rx_queues = adapter->opt_rx_comp_queues;
2042                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2043
2044                 adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2045         }
2046
2047         memset(&crq, 0, sizeof(crq));
2048         crq.request_capability.first = IBMVNIC_CRQ_CMD;
2049         crq.request_capability.cmd = REQUEST_CAPABILITY;
2050
2051         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2052         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2053         atomic_inc(&adapter->running_cap_crqs);
2054         ibmvnic_send_crq(adapter, &crq);
2055
2056         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2057         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2058         atomic_inc(&adapter->running_cap_crqs);
2059         ibmvnic_send_crq(adapter, &crq);
2060
2061         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2062         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2063         atomic_inc(&adapter->running_cap_crqs);
2064         ibmvnic_send_crq(adapter, &crq);
2065
2066         crq.request_capability.capability =
2067             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2068         crq.request_capability.number =
2069             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2070         atomic_inc(&adapter->running_cap_crqs);
2071         ibmvnic_send_crq(adapter, &crq);
2072
2073         crq.request_capability.capability =
2074             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2075         crq.request_capability.number =
2076             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2077         atomic_inc(&adapter->running_cap_crqs);
2078         ibmvnic_send_crq(adapter, &crq);
2079
2080         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2081         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2082         atomic_inc(&adapter->running_cap_crqs);
2083         ibmvnic_send_crq(adapter, &crq);
2084
2085         if (adapter->netdev->flags & IFF_PROMISC) {
2086                 if (adapter->promisc_supported) {
2087                         crq.request_capability.capability =
2088                             cpu_to_be16(PROMISC_REQUESTED);
2089                         crq.request_capability.number = cpu_to_be64(1);
2090                         atomic_inc(&adapter->running_cap_crqs);
2091                         ibmvnic_send_crq(adapter, &crq);
2092                 }
2093         } else {
2094                 crq.request_capability.capability =
2095                     cpu_to_be16(PROMISC_REQUESTED);
2096                 crq.request_capability.number = cpu_to_be64(0);
2097                 atomic_inc(&adapter->running_cap_crqs);
2098                 ibmvnic_send_crq(adapter, &crq);
2099         }
2100 }
2101
2102 static int pending_scrq(struct ibmvnic_adapter *adapter,
2103                         struct ibmvnic_sub_crq_queue *scrq)
2104 {
2105         union sub_crq *entry = &scrq->msgs[scrq->cur];
2106
2107         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP ||
2108             adapter->state == VNIC_CLOSING)
2109                 return 1;
2110         else
2111                 return 0;
2112 }
2113
2114 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2115                                         struct ibmvnic_sub_crq_queue *scrq)
2116 {
2117         union sub_crq *entry;
2118         unsigned long flags;
2119
2120         spin_lock_irqsave(&scrq->lock, flags);
2121         entry = &scrq->msgs[scrq->cur];
2122         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2123                 if (++scrq->cur == scrq->size)
2124                         scrq->cur = 0;
2125         } else {
2126                 entry = NULL;
2127         }
2128         spin_unlock_irqrestore(&scrq->lock, flags);
2129
2130         return entry;
2131 }
2132
2133 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2134 {
2135         struct ibmvnic_crq_queue *queue = &adapter->crq;
2136         union ibmvnic_crq *crq;
2137
2138         crq = &queue->msgs[queue->cur];
2139         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2140                 if (++queue->cur == queue->size)
2141                         queue->cur = 0;
2142         } else {
2143                 crq = NULL;
2144         }
2145
2146         return crq;
2147 }
2148
2149 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2150                        union sub_crq *sub_crq)
2151 {
2152         unsigned int ua = adapter->vdev->unit_address;
2153         struct device *dev = &adapter->vdev->dev;
2154         u64 *u64_crq = (u64 *)sub_crq;
2155         int rc;
2156
2157         netdev_dbg(adapter->netdev,
2158                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2159                    (unsigned long int)cpu_to_be64(remote_handle),
2160                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2161                    (unsigned long int)cpu_to_be64(u64_crq[1]),
2162                    (unsigned long int)cpu_to_be64(u64_crq[2]),
2163                    (unsigned long int)cpu_to_be64(u64_crq[3]));
2164
2165         /* Make sure the hypervisor sees the complete request */
2166         mb();
2167
2168         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2169                                 cpu_to_be64(remote_handle),
2170                                 cpu_to_be64(u64_crq[0]),
2171                                 cpu_to_be64(u64_crq[1]),
2172                                 cpu_to_be64(u64_crq[2]),
2173                                 cpu_to_be64(u64_crq[3]));
2174
2175         if (rc) {
2176                 if (rc == H_CLOSED)
2177                         dev_warn(dev, "CRQ Queue closed\n");
2178                 dev_err(dev, "Send error (rc=%d)\n", rc);
2179         }
2180
2181         return rc;
2182 }
2183
2184 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2185                                 u64 remote_handle, u64 ioba, u64 num_entries)
2186 {
2187         unsigned int ua = adapter->vdev->unit_address;
2188         struct device *dev = &adapter->vdev->dev;
2189         int rc;
2190
2191         /* Make sure the hypervisor sees the complete request */
2192         mb();
2193         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2194                                 cpu_to_be64(remote_handle),
2195                                 ioba, num_entries);
2196
2197         if (rc) {
2198                 if (rc == H_CLOSED)
2199                         dev_warn(dev, "CRQ Queue closed\n");
2200                 dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2201         }
2202
2203         return rc;
2204 }
2205
2206 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2207                             union ibmvnic_crq *crq)
2208 {
2209         unsigned int ua = adapter->vdev->unit_address;
2210         struct device *dev = &adapter->vdev->dev;
2211         u64 *u64_crq = (u64 *)crq;
2212         int rc;
2213
2214         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2215                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2216                    (unsigned long int)cpu_to_be64(u64_crq[1]));
2217
2218         /* Make sure the hypervisor sees the complete request */
2219         mb();
2220
2221         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2222                                 cpu_to_be64(u64_crq[0]),
2223                                 cpu_to_be64(u64_crq[1]));
2224
2225         if (rc) {
2226                 if (rc == H_CLOSED)
2227                         dev_warn(dev, "CRQ Queue closed\n");
2228                 dev_warn(dev, "Send error (rc=%d)\n", rc);
2229         }
2230
2231         return rc;
2232 }
2233
2234 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2235 {
2236         union ibmvnic_crq crq;
2237
2238         memset(&crq, 0, sizeof(crq));
2239         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2240         crq.generic.cmd = IBMVNIC_CRQ_INIT;
2241         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2242
2243         return ibmvnic_send_crq(adapter, &crq);
2244 }
2245
2246 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2247 {
2248         union ibmvnic_crq crq;
2249
2250         memset(&crq, 0, sizeof(crq));
2251         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2252         crq.version_exchange.cmd = VERSION_EXCHANGE;
2253         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2254
2255         return ibmvnic_send_crq(adapter, &crq);
2256 }
2257
2258 static void send_login(struct ibmvnic_adapter *adapter)
2259 {
2260         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
2261         struct ibmvnic_login_buffer *login_buffer;
2262         struct device *dev = &adapter->vdev->dev;
2263         dma_addr_t rsp_buffer_token;
2264         dma_addr_t buffer_token;
2265         size_t rsp_buffer_size;
2266         union ibmvnic_crq crq;
2267         size_t buffer_size;
2268         __be64 *tx_list_p;
2269         __be64 *rx_list_p;
2270         int i;
2271
2272         buffer_size =
2273             sizeof(struct ibmvnic_login_buffer) +
2274             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues);
2275
2276         login_buffer = kmalloc(buffer_size, GFP_ATOMIC);
2277         if (!login_buffer)
2278                 goto buf_alloc_failed;
2279
2280         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
2281                                       DMA_TO_DEVICE);
2282         if (dma_mapping_error(dev, buffer_token)) {
2283                 dev_err(dev, "Couldn't map login buffer\n");
2284                 goto buf_map_failed;
2285         }
2286
2287         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
2288                           sizeof(u64) * adapter->req_tx_queues +
2289                           sizeof(u64) * adapter->req_rx_queues +
2290                           sizeof(u64) * adapter->req_rx_queues +
2291                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
2292
2293         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
2294         if (!login_rsp_buffer)
2295                 goto buf_rsp_alloc_failed;
2296
2297         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
2298                                           rsp_buffer_size, DMA_FROM_DEVICE);
2299         if (dma_mapping_error(dev, rsp_buffer_token)) {
2300                 dev_err(dev, "Couldn't map login rsp buffer\n");
2301                 goto buf_rsp_map_failed;
2302         }
2303
2304         adapter->login_buf = login_buffer;
2305         adapter->login_buf_token = buffer_token;
2306         adapter->login_buf_sz = buffer_size;
2307         adapter->login_rsp_buf = login_rsp_buffer;
2308         adapter->login_rsp_buf_token = rsp_buffer_token;
2309         adapter->login_rsp_buf_sz = rsp_buffer_size;
2310
2311         login_buffer->len = cpu_to_be32(buffer_size);
2312         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
2313         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
2314         login_buffer->off_txcomp_subcrqs =
2315             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
2316         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
2317         login_buffer->off_rxcomp_subcrqs =
2318             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
2319                         sizeof(u64) * adapter->req_tx_queues);
2320         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
2321         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
2322
2323         tx_list_p = (__be64 *)((char *)login_buffer +
2324                                       sizeof(struct ibmvnic_login_buffer));
2325         rx_list_p = (__be64 *)((char *)login_buffer +
2326                                       sizeof(struct ibmvnic_login_buffer) +
2327                                       sizeof(u64) * adapter->req_tx_queues);
2328
2329         for (i = 0; i < adapter->req_tx_queues; i++) {
2330                 if (adapter->tx_scrq[i]) {
2331                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
2332                                                    crq_num);
2333                 }
2334         }
2335
2336         for (i = 0; i < adapter->req_rx_queues; i++) {
2337                 if (adapter->rx_scrq[i]) {
2338                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
2339                                                    crq_num);
2340                 }
2341         }
2342
2343         netdev_dbg(adapter->netdev, "Login Buffer:\n");
2344         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
2345                 netdev_dbg(adapter->netdev, "%016lx\n",
2346                            ((unsigned long int *)(adapter->login_buf))[i]);
2347         }
2348
2349         memset(&crq, 0, sizeof(crq));
2350         crq.login.first = IBMVNIC_CRQ_CMD;
2351         crq.login.cmd = LOGIN;
2352         crq.login.ioba = cpu_to_be32(buffer_token);
2353         crq.login.len = cpu_to_be32(buffer_size);
2354         ibmvnic_send_crq(adapter, &crq);
2355
2356         return;
2357
2358 buf_rsp_map_failed:
2359         kfree(login_rsp_buffer);
2360 buf_rsp_alloc_failed:
2361         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
2362 buf_map_failed:
2363         kfree(login_buffer);
2364 buf_alloc_failed:
2365         return;
2366 }
2367
2368 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
2369                              u32 len, u8 map_id)
2370 {
2371         union ibmvnic_crq crq;
2372
2373         memset(&crq, 0, sizeof(crq));
2374         crq.request_map.first = IBMVNIC_CRQ_CMD;
2375         crq.request_map.cmd = REQUEST_MAP;
2376         crq.request_map.map_id = map_id;
2377         crq.request_map.ioba = cpu_to_be32(addr);
2378         crq.request_map.len = cpu_to_be32(len);
2379         ibmvnic_send_crq(adapter, &crq);
2380 }
2381
2382 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
2383 {
2384         union ibmvnic_crq crq;
2385
2386         memset(&crq, 0, sizeof(crq));
2387         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
2388         crq.request_unmap.cmd = REQUEST_UNMAP;
2389         crq.request_unmap.map_id = map_id;
2390         ibmvnic_send_crq(adapter, &crq);
2391 }
2392
2393 static void send_map_query(struct ibmvnic_adapter *adapter)
2394 {
2395         union ibmvnic_crq crq;
2396
2397         memset(&crq, 0, sizeof(crq));
2398         crq.query_map.first = IBMVNIC_CRQ_CMD;
2399         crq.query_map.cmd = QUERY_MAP;
2400         ibmvnic_send_crq(adapter, &crq);
2401 }
2402
2403 /* Send a series of CRQs requesting various capabilities of the VNIC server */
2404 static void send_cap_queries(struct ibmvnic_adapter *adapter)
2405 {
2406         union ibmvnic_crq crq;
2407
2408         atomic_set(&adapter->running_cap_crqs, 0);
2409         memset(&crq, 0, sizeof(crq));
2410         crq.query_capability.first = IBMVNIC_CRQ_CMD;
2411         crq.query_capability.cmd = QUERY_CAPABILITY;
2412
2413         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
2414         atomic_inc(&adapter->running_cap_crqs);
2415         ibmvnic_send_crq(adapter, &crq);
2416
2417         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
2418         atomic_inc(&adapter->running_cap_crqs);
2419         ibmvnic_send_crq(adapter, &crq);
2420
2421         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
2422         atomic_inc(&adapter->running_cap_crqs);
2423         ibmvnic_send_crq(adapter, &crq);
2424
2425         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
2426         atomic_inc(&adapter->running_cap_crqs);
2427         ibmvnic_send_crq(adapter, &crq);
2428
2429         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
2430         atomic_inc(&adapter->running_cap_crqs);
2431         ibmvnic_send_crq(adapter, &crq);
2432
2433         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
2434         atomic_inc(&adapter->running_cap_crqs);
2435         ibmvnic_send_crq(adapter, &crq);
2436
2437         crq.query_capability.capability =
2438             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
2439         atomic_inc(&adapter->running_cap_crqs);
2440         ibmvnic_send_crq(adapter, &crq);
2441
2442         crq.query_capability.capability =
2443             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
2444         atomic_inc(&adapter->running_cap_crqs);
2445         ibmvnic_send_crq(adapter, &crq);
2446
2447         crq.query_capability.capability =
2448             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
2449         atomic_inc(&adapter->running_cap_crqs);
2450         ibmvnic_send_crq(adapter, &crq);
2451
2452         crq.query_capability.capability =
2453             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
2454         atomic_inc(&adapter->running_cap_crqs);
2455         ibmvnic_send_crq(adapter, &crq);
2456
2457         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
2458         atomic_inc(&adapter->running_cap_crqs);
2459         ibmvnic_send_crq(adapter, &crq);
2460
2461         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
2462         atomic_inc(&adapter->running_cap_crqs);
2463         ibmvnic_send_crq(adapter, &crq);
2464
2465         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
2466         atomic_inc(&adapter->running_cap_crqs);
2467         ibmvnic_send_crq(adapter, &crq);
2468
2469         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
2470         atomic_inc(&adapter->running_cap_crqs);
2471         ibmvnic_send_crq(adapter, &crq);
2472
2473         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
2474         atomic_inc(&adapter->running_cap_crqs);
2475         ibmvnic_send_crq(adapter, &crq);
2476
2477         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
2478         atomic_inc(&adapter->running_cap_crqs);
2479         ibmvnic_send_crq(adapter, &crq);
2480
2481         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
2482         atomic_inc(&adapter->running_cap_crqs);
2483         ibmvnic_send_crq(adapter, &crq);
2484
2485         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
2486         atomic_inc(&adapter->running_cap_crqs);
2487         ibmvnic_send_crq(adapter, &crq);
2488
2489         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
2490         atomic_inc(&adapter->running_cap_crqs);
2491         ibmvnic_send_crq(adapter, &crq);
2492
2493         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
2494         atomic_inc(&adapter->running_cap_crqs);
2495         ibmvnic_send_crq(adapter, &crq);
2496
2497         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
2498         atomic_inc(&adapter->running_cap_crqs);
2499         ibmvnic_send_crq(adapter, &crq);
2500
2501         crq.query_capability.capability =
2502                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
2503         atomic_inc(&adapter->running_cap_crqs);
2504         ibmvnic_send_crq(adapter, &crq);
2505
2506         crq.query_capability.capability =
2507                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
2508         atomic_inc(&adapter->running_cap_crqs);
2509         ibmvnic_send_crq(adapter, &crq);
2510
2511         crq.query_capability.capability =
2512                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
2513         atomic_inc(&adapter->running_cap_crqs);
2514         ibmvnic_send_crq(adapter, &crq);
2515
2516         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
2517         atomic_inc(&adapter->running_cap_crqs);
2518         ibmvnic_send_crq(adapter, &crq);
2519 }
2520
2521 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
2522 {
2523         struct device *dev = &adapter->vdev->dev;
2524         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
2525         union ibmvnic_crq crq;
2526         int i;
2527
2528         dma_unmap_single(dev, adapter->ip_offload_tok,
2529                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
2530
2531         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
2532         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
2533                 netdev_dbg(adapter->netdev, "%016lx\n",
2534                            ((unsigned long int *)(buf))[i]);
2535
2536         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
2537         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
2538         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
2539                    buf->tcp_ipv4_chksum);
2540         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
2541                    buf->tcp_ipv6_chksum);
2542         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
2543                    buf->udp_ipv4_chksum);
2544         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
2545                    buf->udp_ipv6_chksum);
2546         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
2547                    buf->large_tx_ipv4);
2548         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
2549                    buf->large_tx_ipv6);
2550         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
2551                    buf->large_rx_ipv4);
2552         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
2553                    buf->large_rx_ipv6);
2554         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
2555                    buf->max_ipv4_header_size);
2556         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
2557                    buf->max_ipv6_header_size);
2558         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
2559                    buf->max_tcp_header_size);
2560         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
2561                    buf->max_udp_header_size);
2562         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
2563                    buf->max_large_tx_size);
2564         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
2565                    buf->max_large_rx_size);
2566         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
2567                    buf->ipv6_extension_header);
2568         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
2569                    buf->tcp_pseudosum_req);
2570         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
2571                    buf->num_ipv6_ext_headers);
2572         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
2573                    buf->off_ipv6_ext_headers);
2574
2575         adapter->ip_offload_ctrl_tok =
2576             dma_map_single(dev, &adapter->ip_offload_ctrl,
2577                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
2578
2579         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
2580                 dev_err(dev, "Couldn't map ip offload control buffer\n");
2581                 return;
2582         }
2583
2584         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
2585         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
2586         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
2587         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
2588         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
2589
2590         /* large_tx/rx disabled for now, additional features needed */
2591         adapter->ip_offload_ctrl.large_tx_ipv4 = 0;
2592         adapter->ip_offload_ctrl.large_tx_ipv6 = 0;
2593         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
2594         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
2595
2596         adapter->netdev->features = NETIF_F_GSO;
2597
2598         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
2599                 adapter->netdev->features |= NETIF_F_IP_CSUM;
2600
2601         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
2602                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
2603
2604         if ((adapter->netdev->features &
2605             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
2606                 adapter->netdev->features |= NETIF_F_RXCSUM;
2607
2608         memset(&crq, 0, sizeof(crq));
2609         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
2610         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
2611         crq.control_ip_offload.len =
2612             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
2613         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
2614         ibmvnic_send_crq(adapter, &crq);
2615 }
2616
2617 static void handle_error_info_rsp(union ibmvnic_crq *crq,
2618                                   struct ibmvnic_adapter *adapter)
2619 {
2620         struct device *dev = &adapter->vdev->dev;
2621         struct ibmvnic_error_buff *error_buff, *tmp;
2622         unsigned long flags;
2623         bool found = false;
2624         int i;
2625
2626         if (!crq->request_error_rsp.rc.code) {
2627                 dev_info(dev, "Request Error Rsp returned with rc=%x\n",
2628                          crq->request_error_rsp.rc.code);
2629                 return;
2630         }
2631
2632         spin_lock_irqsave(&adapter->error_list_lock, flags);
2633         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
2634                 if (error_buff->error_id == crq->request_error_rsp.error_id) {
2635                         found = true;
2636                         list_del(&error_buff->list);
2637                         break;
2638                 }
2639         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2640
2641         if (!found) {
2642                 dev_err(dev, "Couldn't find error id %x\n",
2643                         be32_to_cpu(crq->request_error_rsp.error_id));
2644                 return;
2645         }
2646
2647         dev_err(dev, "Detailed info for error id %x:",
2648                 be32_to_cpu(crq->request_error_rsp.error_id));
2649
2650         for (i = 0; i < error_buff->len; i++) {
2651                 pr_cont("%02x", (int)error_buff->buff[i]);
2652                 if (i % 8 == 7)
2653                         pr_cont(" ");
2654         }
2655         pr_cont("\n");
2656
2657         dma_unmap_single(dev, error_buff->dma, error_buff->len,
2658                          DMA_FROM_DEVICE);
2659         kfree(error_buff->buff);
2660         kfree(error_buff);
2661 }
2662
2663 static void request_error_information(struct ibmvnic_adapter *adapter,
2664                                       union ibmvnic_crq *err_crq)
2665 {
2666         struct device *dev = &adapter->vdev->dev;
2667         struct net_device *netdev = adapter->netdev;
2668         struct ibmvnic_error_buff *error_buff;
2669         unsigned long timeout = msecs_to_jiffies(30000);
2670         union ibmvnic_crq crq;
2671         unsigned long flags;
2672         int rc, detail_len;
2673
2674         error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
2675         if (!error_buff)
2676                 return;
2677
2678         detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
2679         error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
2680         if (!error_buff->buff) {
2681                 kfree(error_buff);
2682                 return;
2683         }
2684
2685         error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
2686                                          DMA_FROM_DEVICE);
2687         if (dma_mapping_error(dev, error_buff->dma)) {
2688                 netdev_err(netdev, "Couldn't map error buffer\n");
2689                 kfree(error_buff->buff);
2690                 kfree(error_buff);
2691                 return;
2692         }
2693
2694         error_buff->len = detail_len;
2695         error_buff->error_id = err_crq->error_indication.error_id;
2696
2697         spin_lock_irqsave(&adapter->error_list_lock, flags);
2698         list_add_tail(&error_buff->list, &adapter->errors);
2699         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2700
2701         memset(&crq, 0, sizeof(crq));
2702         crq.request_error_info.first = IBMVNIC_CRQ_CMD;
2703         crq.request_error_info.cmd = REQUEST_ERROR_INFO;
2704         crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
2705         crq.request_error_info.len = cpu_to_be32(detail_len);
2706         crq.request_error_info.error_id = err_crq->error_indication.error_id;
2707
2708         rc = ibmvnic_send_crq(adapter, &crq);
2709         if (rc) {
2710                 netdev_err(netdev, "failed to request error information\n");
2711                 goto err_info_fail;
2712         }
2713
2714         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
2715                 netdev_err(netdev, "timeout waiting for error information\n");
2716                 goto err_info_fail;
2717         }
2718
2719         return;
2720
2721 err_info_fail:
2722         spin_lock_irqsave(&adapter->error_list_lock, flags);
2723         list_del(&error_buff->list);
2724         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2725
2726         kfree(error_buff->buff);
2727         kfree(error_buff);
2728 }
2729
2730 static void handle_error_indication(union ibmvnic_crq *crq,
2731                                     struct ibmvnic_adapter *adapter)
2732 {
2733         struct device *dev = &adapter->vdev->dev;
2734
2735         dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
2736                 crq->error_indication.flags
2737                         & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
2738                 be32_to_cpu(crq->error_indication.error_id),
2739                 be16_to_cpu(crq->error_indication.error_cause));
2740
2741         if (be32_to_cpu(crq->error_indication.error_id))
2742                 request_error_information(adapter, crq);
2743
2744         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
2745                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
2746 }
2747
2748 static void handle_change_mac_rsp(union ibmvnic_crq *crq,
2749                                   struct ibmvnic_adapter *adapter)
2750 {
2751         struct net_device *netdev = adapter->netdev;
2752         struct device *dev = &adapter->vdev->dev;
2753         long rc;
2754
2755         rc = crq->change_mac_addr_rsp.rc.code;
2756         if (rc) {
2757                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
2758                 return;
2759         }
2760         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
2761                ETH_ALEN);
2762 }
2763
2764 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
2765                                    struct ibmvnic_adapter *adapter)
2766 {
2767         struct device *dev = &adapter->vdev->dev;
2768         u64 *req_value;
2769         char *name;
2770
2771         atomic_dec(&adapter->running_cap_crqs);
2772         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
2773         case REQ_TX_QUEUES:
2774                 req_value = &adapter->req_tx_queues;
2775                 name = "tx";
2776                 break;
2777         case REQ_RX_QUEUES:
2778                 req_value = &adapter->req_rx_queues;
2779                 name = "rx";
2780                 break;
2781         case REQ_RX_ADD_QUEUES:
2782                 req_value = &adapter->req_rx_add_queues;
2783                 name = "rx_add";
2784                 break;
2785         case REQ_TX_ENTRIES_PER_SUBCRQ:
2786                 req_value = &adapter->req_tx_entries_per_subcrq;
2787                 name = "tx_entries_per_subcrq";
2788                 break;
2789         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
2790                 req_value = &adapter->req_rx_add_entries_per_subcrq;
2791                 name = "rx_add_entries_per_subcrq";
2792                 break;
2793         case REQ_MTU:
2794                 req_value = &adapter->req_mtu;
2795                 name = "mtu";
2796                 break;
2797         case PROMISC_REQUESTED:
2798                 req_value = &adapter->promisc;
2799                 name = "promisc";
2800                 break;
2801         default:
2802                 dev_err(dev, "Got invalid cap request rsp %d\n",
2803                         crq->request_capability.capability);
2804                 return;
2805         }
2806
2807         switch (crq->request_capability_rsp.rc.code) {
2808         case SUCCESS:
2809                 break;
2810         case PARTIALSUCCESS:
2811                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
2812                          *req_value,
2813                          (long int)be64_to_cpu(crq->request_capability_rsp.
2814                                                number), name);
2815                 release_sub_crqs(adapter);
2816                 *req_value = be64_to_cpu(crq->request_capability_rsp.number);
2817                 ibmvnic_send_req_caps(adapter, 1);
2818                 return;
2819         default:
2820                 dev_err(dev, "Error %d in request cap rsp\n",
2821                         crq->request_capability_rsp.rc.code);
2822                 return;
2823         }
2824
2825         /* Done receiving requested capabilities, query IP offload support */
2826         if (atomic_read(&adapter->running_cap_crqs) == 0) {
2827                 union ibmvnic_crq newcrq;
2828                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
2829                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
2830                     &adapter->ip_offload_buf;
2831
2832                 adapter->wait_capability = false;
2833                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
2834                                                          buf_sz,
2835                                                          DMA_FROM_DEVICE);
2836
2837                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
2838                         if (!firmware_has_feature(FW_FEATURE_CMO))
2839                                 dev_err(dev, "Couldn't map offload buffer\n");
2840                         return;
2841                 }
2842
2843                 memset(&newcrq, 0, sizeof(newcrq));
2844                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
2845                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
2846                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
2847                 newcrq.query_ip_offload.ioba =
2848                     cpu_to_be32(adapter->ip_offload_tok);
2849
2850                 ibmvnic_send_crq(adapter, &newcrq);
2851         }
2852 }
2853
2854 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
2855                             struct ibmvnic_adapter *adapter)
2856 {
2857         struct device *dev = &adapter->vdev->dev;
2858         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
2859         struct ibmvnic_login_buffer *login = adapter->login_buf;
2860         int i;
2861
2862         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
2863                          DMA_BIDIRECTIONAL);
2864         dma_unmap_single(dev, adapter->login_rsp_buf_token,
2865                          adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
2866
2867         /* If the number of queues requested can't be allocated by the
2868          * server, the login response will return with code 1. We will need
2869          * to resend the login buffer with fewer queues requested.
2870          */
2871         if (login_rsp_crq->generic.rc.code) {
2872                 adapter->renegotiate = true;
2873                 complete(&adapter->init_done);
2874                 return 0;
2875         }
2876
2877         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
2878         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
2879                 netdev_dbg(adapter->netdev, "%016lx\n",
2880                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
2881         }
2882
2883         /* Sanity checks */
2884         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
2885             (be32_to_cpu(login->num_rxcomp_subcrqs) *
2886              adapter->req_rx_add_queues !=
2887              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
2888                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
2889                 ibmvnic_remove(adapter->vdev);
2890                 return -EIO;
2891         }
2892         complete(&adapter->init_done);
2893
2894         return 0;
2895 }
2896
2897 static void handle_request_map_rsp(union ibmvnic_crq *crq,
2898                                    struct ibmvnic_adapter *adapter)
2899 {
2900         struct device *dev = &adapter->vdev->dev;
2901         u8 map_id = crq->request_map_rsp.map_id;
2902         int tx_subcrqs;
2903         int rx_subcrqs;
2904         long rc;
2905         int i;
2906
2907         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
2908         rx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
2909
2910         rc = crq->request_map_rsp.rc.code;
2911         if (rc) {
2912                 dev_err(dev, "Error %ld in REQUEST_MAP_RSP\n", rc);
2913                 adapter->map_id--;
2914                 /* need to find and zero tx/rx_pool map_id */
2915                 for (i = 0; i < tx_subcrqs; i++) {
2916                         if (adapter->tx_pool[i].long_term_buff.map_id == map_id)
2917                                 adapter->tx_pool[i].long_term_buff.map_id = 0;
2918                 }
2919                 for (i = 0; i < rx_subcrqs; i++) {
2920                         if (adapter->rx_pool[i].long_term_buff.map_id == map_id)
2921                                 adapter->rx_pool[i].long_term_buff.map_id = 0;
2922                 }
2923         }
2924         complete(&adapter->fw_done);
2925 }
2926
2927 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
2928                                      struct ibmvnic_adapter *adapter)
2929 {
2930         struct device *dev = &adapter->vdev->dev;
2931         long rc;
2932
2933         rc = crq->request_unmap_rsp.rc.code;
2934         if (rc)
2935                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
2936 }
2937
2938 static void handle_query_map_rsp(union ibmvnic_crq *crq,
2939                                  struct ibmvnic_adapter *adapter)
2940 {
2941         struct net_device *netdev = adapter->netdev;
2942         struct device *dev = &adapter->vdev->dev;
2943         long rc;
2944
2945         rc = crq->query_map_rsp.rc.code;
2946         if (rc) {
2947                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
2948                 return;
2949         }
2950         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
2951                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
2952                    crq->query_map_rsp.free_pages);
2953 }
2954
2955 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
2956                                  struct ibmvnic_adapter *adapter)
2957 {
2958         struct net_device *netdev = adapter->netdev;
2959         struct device *dev = &adapter->vdev->dev;
2960         long rc;
2961
2962         atomic_dec(&adapter->running_cap_crqs);
2963         netdev_dbg(netdev, "Outstanding queries: %d\n",
2964                    atomic_read(&adapter->running_cap_crqs));
2965         rc = crq->query_capability.rc.code;
2966         if (rc) {
2967                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
2968                 goto out;
2969         }
2970
2971         switch (be16_to_cpu(crq->query_capability.capability)) {
2972         case MIN_TX_QUEUES:
2973                 adapter->min_tx_queues =
2974                     be64_to_cpu(crq->query_capability.number);
2975                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
2976                            adapter->min_tx_queues);
2977                 break;
2978         case MIN_RX_QUEUES:
2979                 adapter->min_rx_queues =
2980                     be64_to_cpu(crq->query_capability.number);
2981                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
2982                            adapter->min_rx_queues);
2983                 break;
2984         case MIN_RX_ADD_QUEUES:
2985                 adapter->min_rx_add_queues =
2986                     be64_to_cpu(crq->query_capability.number);
2987                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
2988                            adapter->min_rx_add_queues);
2989                 break;
2990         case MAX_TX_QUEUES:
2991                 adapter->max_tx_queues =
2992                     be64_to_cpu(crq->query_capability.number);
2993                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
2994                            adapter->max_tx_queues);
2995                 break;
2996         case MAX_RX_QUEUES:
2997                 adapter->max_rx_queues =
2998                     be64_to_cpu(crq->query_capability.number);
2999                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
3000                            adapter->max_rx_queues);
3001                 break;
3002         case MAX_RX_ADD_QUEUES:
3003                 adapter->max_rx_add_queues =
3004                     be64_to_cpu(crq->query_capability.number);
3005                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3006                            adapter->max_rx_add_queues);
3007                 break;
3008         case MIN_TX_ENTRIES_PER_SUBCRQ:
3009                 adapter->min_tx_entries_per_subcrq =
3010                     be64_to_cpu(crq->query_capability.number);
3011                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3012                            adapter->min_tx_entries_per_subcrq);
3013                 break;
3014         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3015                 adapter->min_rx_add_entries_per_subcrq =
3016                     be64_to_cpu(crq->query_capability.number);
3017                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3018                            adapter->min_rx_add_entries_per_subcrq);
3019                 break;
3020         case MAX_TX_ENTRIES_PER_SUBCRQ:
3021                 adapter->max_tx_entries_per_subcrq =
3022                     be64_to_cpu(crq->query_capability.number);
3023                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3024                            adapter->max_tx_entries_per_subcrq);
3025                 break;
3026         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3027                 adapter->max_rx_add_entries_per_subcrq =
3028                     be64_to_cpu(crq->query_capability.number);
3029                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3030                            adapter->max_rx_add_entries_per_subcrq);
3031                 break;
3032         case TCP_IP_OFFLOAD:
3033                 adapter->tcp_ip_offload =
3034                     be64_to_cpu(crq->query_capability.number);
3035                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3036                            adapter->tcp_ip_offload);
3037                 break;
3038         case PROMISC_SUPPORTED:
3039                 adapter->promisc_supported =
3040                     be64_to_cpu(crq->query_capability.number);
3041                 netdev_dbg(netdev, "promisc_supported = %lld\n",
3042                            adapter->promisc_supported);
3043                 break;
3044         case MIN_MTU:
3045                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3046                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3047                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3048                 break;
3049         case MAX_MTU:
3050                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3051                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3052                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3053                 break;
3054         case MAX_MULTICAST_FILTERS:
3055                 adapter->max_multicast_filters =
3056                     be64_to_cpu(crq->query_capability.number);
3057                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3058                            adapter->max_multicast_filters);
3059                 break;
3060         case VLAN_HEADER_INSERTION:
3061                 adapter->vlan_header_insertion =
3062                     be64_to_cpu(crq->query_capability.number);
3063                 if (adapter->vlan_header_insertion)
3064                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3065                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3066                            adapter->vlan_header_insertion);
3067                 break;
3068         case RX_VLAN_HEADER_INSERTION:
3069                 adapter->rx_vlan_header_insertion =
3070                     be64_to_cpu(crq->query_capability.number);
3071                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3072                            adapter->rx_vlan_header_insertion);
3073                 break;
3074         case MAX_TX_SG_ENTRIES:
3075                 adapter->max_tx_sg_entries =
3076                     be64_to_cpu(crq->query_capability.number);
3077                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3078                            adapter->max_tx_sg_entries);
3079                 break;
3080         case RX_SG_SUPPORTED:
3081                 adapter->rx_sg_supported =
3082                     be64_to_cpu(crq->query_capability.number);
3083                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3084                            adapter->rx_sg_supported);
3085                 break;
3086         case OPT_TX_COMP_SUB_QUEUES:
3087                 adapter->opt_tx_comp_sub_queues =
3088                     be64_to_cpu(crq->query_capability.number);
3089                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3090                            adapter->opt_tx_comp_sub_queues);
3091                 break;
3092         case OPT_RX_COMP_QUEUES:
3093                 adapter->opt_rx_comp_queues =
3094                     be64_to_cpu(crq->query_capability.number);
3095                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3096                            adapter->opt_rx_comp_queues);
3097                 break;
3098         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3099                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
3100                     be64_to_cpu(crq->query_capability.number);
3101                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3102                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
3103                 break;
3104         case OPT_TX_ENTRIES_PER_SUBCRQ:
3105                 adapter->opt_tx_entries_per_subcrq =
3106                     be64_to_cpu(crq->query_capability.number);
3107                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3108                            adapter->opt_tx_entries_per_subcrq);
3109                 break;
3110         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3111                 adapter->opt_rxba_entries_per_subcrq =
3112                     be64_to_cpu(crq->query_capability.number);
3113                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3114                            adapter->opt_rxba_entries_per_subcrq);
3115                 break;
3116         case TX_RX_DESC_REQ:
3117                 adapter->tx_rx_desc_req = crq->query_capability.number;
3118                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3119                            adapter->tx_rx_desc_req);
3120                 break;
3121
3122         default:
3123                 netdev_err(netdev, "Got invalid cap rsp %d\n",
3124                            crq->query_capability.capability);
3125         }
3126
3127 out:
3128         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3129                 adapter->wait_capability = false;
3130                 ibmvnic_send_req_caps(adapter, 0);
3131         }
3132 }
3133
3134 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
3135                                struct ibmvnic_adapter *adapter)
3136 {
3137         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
3138         struct net_device *netdev = adapter->netdev;
3139         struct device *dev = &adapter->vdev->dev;
3140         u64 *u64_crq = (u64 *)crq;
3141         long rc;
3142
3143         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
3144                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3145                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3146         switch (gen_crq->first) {
3147         case IBMVNIC_CRQ_INIT_RSP:
3148                 switch (gen_crq->cmd) {
3149                 case IBMVNIC_CRQ_INIT:
3150                         dev_info(dev, "Partner initialized\n");
3151                         break;
3152                 case IBMVNIC_CRQ_INIT_COMPLETE:
3153                         dev_info(dev, "Partner initialization complete\n");
3154                         send_version_xchg(adapter);
3155                         break;
3156                 default:
3157                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
3158                 }
3159                 return;
3160         case IBMVNIC_CRQ_XPORT_EVENT:
3161                 netif_carrier_off(netdev);
3162                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
3163                         dev_info(dev, "Migrated, re-enabling adapter\n");
3164                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
3165                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
3166                         dev_info(dev, "Backing device failover detected\n");
3167                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
3168                 } else {
3169                         /* The adapter lost the connection */
3170                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
3171                                 gen_crq->cmd);
3172                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3173                 }
3174                 return;
3175         case IBMVNIC_CRQ_CMD_RSP:
3176                 break;
3177         default:
3178                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
3179                         gen_crq->first);
3180                 return;
3181         }
3182
3183         switch (gen_crq->cmd) {
3184         case VERSION_EXCHANGE_RSP:
3185                 rc = crq->version_exchange_rsp.rc.code;
3186                 if (rc) {
3187                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
3188                         break;
3189                 }
3190                 dev_info(dev, "Partner protocol version is %d\n",
3191                          crq->version_exchange_rsp.version);
3192                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
3193                     ibmvnic_version)
3194                         ibmvnic_version =
3195                             be16_to_cpu(crq->version_exchange_rsp.version);
3196                 send_cap_queries(adapter);
3197                 break;
3198         case QUERY_CAPABILITY_RSP:
3199                 handle_query_cap_rsp(crq, adapter);
3200                 break;
3201         case QUERY_MAP_RSP:
3202                 handle_query_map_rsp(crq, adapter);
3203                 break;
3204         case REQUEST_MAP_RSP:
3205                 handle_request_map_rsp(crq, adapter);
3206                 break;
3207         case REQUEST_UNMAP_RSP:
3208                 handle_request_unmap_rsp(crq, adapter);
3209                 break;
3210         case REQUEST_CAPABILITY_RSP:
3211                 handle_request_cap_rsp(crq, adapter);
3212                 break;
3213         case LOGIN_RSP:
3214                 netdev_dbg(netdev, "Got Login Response\n");
3215                 handle_login_rsp(crq, adapter);
3216                 break;
3217         case LOGICAL_LINK_STATE_RSP:
3218                 netdev_dbg(netdev,
3219                            "Got Logical Link State Response, state: %d rc: %d\n",
3220                            crq->logical_link_state_rsp.link_state,
3221                            crq->logical_link_state_rsp.rc.code);
3222                 adapter->logical_link_state =
3223                     crq->logical_link_state_rsp.link_state;
3224                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
3225                 complete(&adapter->init_done);
3226                 break;
3227         case LINK_STATE_INDICATION:
3228                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
3229                 adapter->phys_link_state =
3230                     crq->link_state_indication.phys_link_state;
3231                 adapter->logical_link_state =
3232                     crq->link_state_indication.logical_link_state;
3233                 break;
3234         case CHANGE_MAC_ADDR_RSP:
3235                 netdev_dbg(netdev, "Got MAC address change Response\n");
3236                 handle_change_mac_rsp(crq, adapter);
3237                 break;
3238         case ERROR_INDICATION:
3239                 netdev_dbg(netdev, "Got Error Indication\n");
3240                 handle_error_indication(crq, adapter);
3241                 break;
3242         case REQUEST_ERROR_RSP:
3243                 netdev_dbg(netdev, "Got Error Detail Response\n");
3244                 handle_error_info_rsp(crq, adapter);
3245                 break;
3246         case REQUEST_STATISTICS_RSP:
3247                 netdev_dbg(netdev, "Got Statistics Response\n");
3248                 complete(&adapter->stats_done);
3249                 break;
3250         case QUERY_IP_OFFLOAD_RSP:
3251                 netdev_dbg(netdev, "Got Query IP offload Response\n");
3252                 handle_query_ip_offload_rsp(adapter);
3253                 break;
3254         case MULTICAST_CTRL_RSP:
3255                 netdev_dbg(netdev, "Got multicast control Response\n");
3256                 break;
3257         case CONTROL_IP_OFFLOAD_RSP:
3258                 netdev_dbg(netdev, "Got Control IP offload Response\n");
3259                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
3260                                  sizeof(adapter->ip_offload_ctrl),
3261                                  DMA_TO_DEVICE);
3262                 complete(&adapter->init_done);
3263                 break;
3264         case COLLECT_FW_TRACE_RSP:
3265                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
3266                 complete(&adapter->fw_done);
3267                 break;
3268         default:
3269                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
3270                            gen_crq->cmd);
3271         }
3272 }
3273
3274 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
3275 {
3276         struct ibmvnic_adapter *adapter = instance;
3277
3278         tasklet_schedule(&adapter->tasklet);
3279         return IRQ_HANDLED;
3280 }
3281
3282 static void ibmvnic_tasklet(void *data)
3283 {
3284         struct ibmvnic_adapter *adapter = data;
3285         struct ibmvnic_crq_queue *queue = &adapter->crq;
3286         union ibmvnic_crq *crq;
3287         unsigned long flags;
3288         bool done = false;
3289
3290         spin_lock_irqsave(&queue->lock, flags);
3291         while (!done) {
3292                 /* Pull all the valid messages off the CRQ */
3293                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
3294                         ibmvnic_handle_crq(crq, adapter);
3295                         crq->generic.first = 0;
3296                 }
3297
3298                 /* remain in tasklet until all
3299                  * capabilities responses are received
3300                  */
3301                 if (!adapter->wait_capability)
3302                         done = true;
3303         }
3304         /* if capabilities CRQ's were sent in this tasklet, the following
3305          * tasklet must wait until all responses are received
3306          */
3307         if (atomic_read(&adapter->running_cap_crqs) != 0)
3308                 adapter->wait_capability = true;
3309         spin_unlock_irqrestore(&queue->lock, flags);
3310 }
3311
3312 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
3313 {
3314         struct vio_dev *vdev = adapter->vdev;
3315         int rc;
3316
3317         do {
3318                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
3319         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
3320
3321         if (rc)
3322                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
3323
3324         return rc;
3325 }
3326
3327 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
3328 {
3329         struct ibmvnic_crq_queue *crq = &adapter->crq;
3330         struct device *dev = &adapter->vdev->dev;
3331         struct vio_dev *vdev = adapter->vdev;
3332         int rc;
3333
3334         /* Close the CRQ */
3335         do {
3336                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3337         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3338
3339         /* Clean out the queue */
3340         memset(crq->msgs, 0, PAGE_SIZE);
3341         crq->cur = 0;
3342
3343         /* And re-open it again */
3344         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3345                                 crq->msg_token, PAGE_SIZE);
3346
3347         if (rc == H_CLOSED)
3348                 /* Adapter is good, but other end is not ready */
3349                 dev_warn(dev, "Partner adapter not ready\n");
3350         else if (rc != 0)
3351                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
3352
3353         return rc;
3354 }
3355
3356 static void release_crq_queue(struct ibmvnic_adapter *adapter)
3357 {
3358         struct ibmvnic_crq_queue *crq = &adapter->crq;
3359         struct vio_dev *vdev = adapter->vdev;
3360         long rc;
3361
3362         if (!crq->msgs)
3363                 return;
3364
3365         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
3366         free_irq(vdev->irq, adapter);
3367         tasklet_kill(&adapter->tasklet);
3368         do {
3369                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3370         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3371
3372         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
3373                          DMA_BIDIRECTIONAL);
3374         free_page((unsigned long)crq->msgs);
3375         crq->msgs = NULL;
3376 }
3377
3378 static int init_crq_queue(struct ibmvnic_adapter *adapter)
3379 {
3380         struct ibmvnic_crq_queue *crq = &adapter->crq;
3381         struct device *dev = &adapter->vdev->dev;
3382         struct vio_dev *vdev = adapter->vdev;
3383         int rc, retrc = -ENOMEM;
3384
3385         if (crq->msgs)
3386                 return 0;
3387
3388         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
3389         /* Should we allocate more than one page? */
3390
3391         if (!crq->msgs)
3392                 return -ENOMEM;
3393
3394         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
3395         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
3396                                         DMA_BIDIRECTIONAL);
3397         if (dma_mapping_error(dev, crq->msg_token))
3398                 goto map_failed;
3399
3400         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3401                                 crq->msg_token, PAGE_SIZE);
3402
3403         if (rc == H_RESOURCE)
3404                 /* maybe kexecing and resource is busy. try a reset */
3405                 rc = ibmvnic_reset_crq(adapter);
3406         retrc = rc;
3407
3408         if (rc == H_CLOSED) {
3409                 dev_warn(dev, "Partner adapter not ready\n");
3410         } else if (rc) {
3411                 dev_warn(dev, "Error %d opening adapter\n", rc);
3412                 goto reg_crq_failed;
3413         }
3414
3415         retrc = 0;
3416
3417         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
3418                      (unsigned long)adapter);
3419
3420         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
3421         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
3422                          adapter);
3423         if (rc) {
3424                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
3425                         vdev->irq, rc);
3426                 goto req_irq_failed;
3427         }
3428
3429         rc = vio_enable_interrupts(vdev);
3430         if (rc) {
3431                 dev_err(dev, "Error %d enabling interrupts\n", rc);
3432                 goto req_irq_failed;
3433         }
3434
3435         crq->cur = 0;
3436         spin_lock_init(&crq->lock);
3437
3438         return retrc;
3439
3440 req_irq_failed:
3441         tasklet_kill(&adapter->tasklet);
3442         do {
3443                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3444         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3445 reg_crq_failed:
3446         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
3447 map_failed:
3448         free_page((unsigned long)crq->msgs);
3449         crq->msgs = NULL;
3450         return retrc;
3451 }
3452
3453 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
3454 {
3455         struct device *dev = &adapter->vdev->dev;
3456         unsigned long timeout = msecs_to_jiffies(30000);
3457         int rc;
3458
3459         rc = init_crq_queue(adapter);
3460         if (rc) {
3461                 dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
3462                 return rc;
3463         }
3464
3465         init_completion(&adapter->init_done);
3466         ibmvnic_send_crq_init(adapter);
3467         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3468                 dev_err(dev, "Initialization sequence timed out\n");
3469                 release_crq_queue(adapter);
3470                 return -1;
3471         }
3472
3473         rc = init_sub_crqs(adapter);
3474         if (rc) {
3475                 dev_err(dev, "Initialization of sub crqs failed\n");
3476                 release_crq_queue(adapter);
3477         }
3478
3479         return rc;
3480 }
3481
3482 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
3483 {
3484         struct ibmvnic_adapter *adapter;
3485         struct net_device *netdev;
3486         unsigned char *mac_addr_p;
3487         int rc;
3488
3489         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
3490                 dev->unit_address);
3491
3492         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
3493                                                         VETH_MAC_ADDR, NULL);
3494         if (!mac_addr_p) {
3495                 dev_err(&dev->dev,
3496                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
3497                         __FILE__, __LINE__);
3498                 return 0;
3499         }
3500
3501         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
3502                                    IBMVNIC_MAX_TX_QUEUES);
3503         if (!netdev)
3504                 return -ENOMEM;
3505
3506         adapter = netdev_priv(netdev);
3507         adapter->state = VNIC_PROBING;
3508         dev_set_drvdata(&dev->dev, netdev);
3509         adapter->vdev = dev;
3510         adapter->netdev = netdev;
3511
3512         ether_addr_copy(adapter->mac_addr, mac_addr_p);
3513         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
3514         netdev->irq = dev->irq;
3515         netdev->netdev_ops = &ibmvnic_netdev_ops;
3516         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
3517         SET_NETDEV_DEV(netdev, &dev->dev);
3518
3519         spin_lock_init(&adapter->stats_lock);
3520
3521         INIT_LIST_HEAD(&adapter->errors);
3522         spin_lock_init(&adapter->error_list_lock);
3523
3524         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
3525         INIT_LIST_HEAD(&adapter->rwi_list);
3526         mutex_init(&adapter->reset_lock);
3527         mutex_init(&adapter->rwi_lock);
3528         adapter->resetting = false;
3529
3530         rc = ibmvnic_init(adapter);
3531         if (rc) {
3532                 free_netdev(netdev);
3533                 return rc;
3534         }
3535
3536         netdev->mtu = adapter->req_mtu - ETH_HLEN;
3537
3538         rc = register_netdev(netdev);
3539         if (rc) {
3540                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
3541                 free_netdev(netdev);
3542                 return rc;
3543         }
3544         dev_info(&dev->dev, "ibmvnic registered\n");
3545
3546         adapter->state = VNIC_PROBED;
3547         return 0;
3548 }
3549
3550 static int ibmvnic_remove(struct vio_dev *dev)
3551 {
3552         struct net_device *netdev = dev_get_drvdata(&dev->dev);
3553         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3554
3555         adapter->state = VNIC_REMOVING;
3556         unregister_netdev(netdev);
3557         mutex_lock(&adapter->reset_lock);
3558
3559         release_resources(adapter);
3560         release_sub_crqs(adapter);
3561         release_crq_queue(adapter);
3562
3563         adapter->state = VNIC_REMOVED;
3564
3565         mutex_unlock(&adapter->reset_lock);
3566         free_netdev(netdev);
3567         dev_set_drvdata(&dev->dev, NULL);
3568
3569         return 0;
3570 }
3571
3572 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
3573 {
3574         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
3575         struct ibmvnic_adapter *adapter;
3576         struct iommu_table *tbl;
3577         unsigned long ret = 0;
3578         int i;
3579
3580         tbl = get_iommu_table_base(&vdev->dev);
3581
3582         /* netdev inits at probe time along with the structures we need below*/
3583         if (!netdev)
3584                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
3585
3586         adapter = netdev_priv(netdev);
3587
3588         ret += PAGE_SIZE; /* the crq message queue */
3589         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
3590
3591         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
3592                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
3593
3594         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
3595              i++)
3596                 ret += adapter->rx_pool[i].size *
3597                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
3598
3599         return ret;
3600 }
3601
3602 static int ibmvnic_resume(struct device *dev)
3603 {
3604         struct net_device *netdev = dev_get_drvdata(dev);
3605         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3606         int i;
3607
3608         /* kick the interrupt handlers just in case we lost an interrupt */
3609         for (i = 0; i < adapter->req_rx_queues; i++)
3610                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
3611                                      adapter->rx_scrq[i]);
3612
3613         return 0;
3614 }
3615
3616 static struct vio_device_id ibmvnic_device_table[] = {
3617         {"network", "IBM,vnic"},
3618         {"", "" }
3619 };
3620 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
3621
3622 static const struct dev_pm_ops ibmvnic_pm_ops = {
3623         .resume = ibmvnic_resume
3624 };
3625
3626 static struct vio_driver ibmvnic_driver = {
3627         .id_table       = ibmvnic_device_table,
3628         .probe          = ibmvnic_probe,
3629         .remove         = ibmvnic_remove,
3630         .get_desired_dma = ibmvnic_get_desired_dma,
3631         .name           = ibmvnic_driver_name,
3632         .pm             = &ibmvnic_pm_ops,
3633 };
3634
3635 /* module functions */
3636 static int __init ibmvnic_module_init(void)
3637 {
3638         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
3639                 IBMVNIC_DRIVER_VERSION);
3640
3641         return vio_register_driver(&ibmvnic_driver);
3642 }
3643
3644 static void __exit ibmvnic_module_exit(void)
3645 {
3646         vio_unregister_driver(&ibmvnic_driver);
3647 }
3648
3649 module_init(ibmvnic_module_init);
3650 module_exit(ibmvnic_module_exit);