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                         continue;
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                 skb_record_rx_queue(skb, scrq_num);
1432
1433                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1434                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1435                         skb->ip_summed = CHECKSUM_UNNECESSARY;
1436                 }
1437
1438                 length = skb->len;
1439                 napi_gro_receive(napi, skb); /* send it up */
1440                 netdev->stats.rx_packets++;
1441                 netdev->stats.rx_bytes += length;
1442                 frames_processed++;
1443         }
1444         replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1445
1446         if (frames_processed < budget) {
1447                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1448                 napi_complete_done(napi, frames_processed);
1449                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1450                     napi_reschedule(napi)) {
1451                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1452                         goto restart_poll;
1453                 }
1454         }
1455         return frames_processed;
1456 }
1457
1458 #ifdef CONFIG_NET_POLL_CONTROLLER
1459 static void ibmvnic_netpoll_controller(struct net_device *dev)
1460 {
1461         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1462         int i;
1463
1464         replenish_pools(netdev_priv(dev));
1465         for (i = 0; i < adapter->req_rx_queues; i++)
1466                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1467                                      adapter->rx_scrq[i]);
1468 }
1469 #endif
1470
1471 static const struct net_device_ops ibmvnic_netdev_ops = {
1472         .ndo_open               = ibmvnic_open,
1473         .ndo_stop               = ibmvnic_close,
1474         .ndo_start_xmit         = ibmvnic_xmit,
1475         .ndo_set_rx_mode        = ibmvnic_set_multi,
1476         .ndo_set_mac_address    = ibmvnic_set_mac,
1477         .ndo_validate_addr      = eth_validate_addr,
1478         .ndo_tx_timeout         = ibmvnic_tx_timeout,
1479 #ifdef CONFIG_NET_POLL_CONTROLLER
1480         .ndo_poll_controller    = ibmvnic_netpoll_controller,
1481 #endif
1482 };
1483
1484 /* ethtool functions */
1485
1486 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
1487                                       struct ethtool_link_ksettings *cmd)
1488 {
1489         u32 supported, advertising;
1490
1491         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
1492                           SUPPORTED_FIBRE);
1493         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
1494                             ADVERTISED_FIBRE);
1495         cmd->base.speed = SPEED_1000;
1496         cmd->base.duplex = DUPLEX_FULL;
1497         cmd->base.port = PORT_FIBRE;
1498         cmd->base.phy_address = 0;
1499         cmd->base.autoneg = AUTONEG_ENABLE;
1500
1501         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1502                                                 supported);
1503         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
1504                                                 advertising);
1505
1506         return 0;
1507 }
1508
1509 static void ibmvnic_get_drvinfo(struct net_device *dev,
1510                                 struct ethtool_drvinfo *info)
1511 {
1512         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
1513         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
1514 }
1515
1516 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
1517 {
1518         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1519
1520         return adapter->msg_enable;
1521 }
1522
1523 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
1524 {
1525         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1526
1527         adapter->msg_enable = data;
1528 }
1529
1530 static u32 ibmvnic_get_link(struct net_device *netdev)
1531 {
1532         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1533
1534         /* Don't need to send a query because we request a logical link up at
1535          * init and then we wait for link state indications
1536          */
1537         return adapter->logical_link_state;
1538 }
1539
1540 static void ibmvnic_get_ringparam(struct net_device *netdev,
1541                                   struct ethtool_ringparam *ring)
1542 {
1543         ring->rx_max_pending = 0;
1544         ring->tx_max_pending = 0;
1545         ring->rx_mini_max_pending = 0;
1546         ring->rx_jumbo_max_pending = 0;
1547         ring->rx_pending = 0;
1548         ring->tx_pending = 0;
1549         ring->rx_mini_pending = 0;
1550         ring->rx_jumbo_pending = 0;
1551 }
1552
1553 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
1554 {
1555         int i;
1556
1557         if (stringset != ETH_SS_STATS)
1558                 return;
1559
1560         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
1561                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
1562 }
1563
1564 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
1565 {
1566         switch (sset) {
1567         case ETH_SS_STATS:
1568                 return ARRAY_SIZE(ibmvnic_stats);
1569         default:
1570                 return -EOPNOTSUPP;
1571         }
1572 }
1573
1574 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
1575                                       struct ethtool_stats *stats, u64 *data)
1576 {
1577         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1578         union ibmvnic_crq crq;
1579         int i;
1580
1581         memset(&crq, 0, sizeof(crq));
1582         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
1583         crq.request_statistics.cmd = REQUEST_STATISTICS;
1584         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
1585         crq.request_statistics.len =
1586             cpu_to_be32(sizeof(struct ibmvnic_statistics));
1587
1588         /* Wait for data to be written */
1589         init_completion(&adapter->stats_done);
1590         ibmvnic_send_crq(adapter, &crq);
1591         wait_for_completion(&adapter->stats_done);
1592
1593         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
1594                 data[i] = IBMVNIC_GET_STAT(adapter, ibmvnic_stats[i].offset);
1595 }
1596
1597 static const struct ethtool_ops ibmvnic_ethtool_ops = {
1598         .get_drvinfo            = ibmvnic_get_drvinfo,
1599         .get_msglevel           = ibmvnic_get_msglevel,
1600         .set_msglevel           = ibmvnic_set_msglevel,
1601         .get_link               = ibmvnic_get_link,
1602         .get_ringparam          = ibmvnic_get_ringparam,
1603         .get_strings            = ibmvnic_get_strings,
1604         .get_sset_count         = ibmvnic_get_sset_count,
1605         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
1606         .get_link_ksettings     = ibmvnic_get_link_ksettings,
1607 };
1608
1609 /* Routines for managing CRQs/sCRQs  */
1610
1611 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
1612                                   struct ibmvnic_sub_crq_queue *scrq)
1613 {
1614         struct device *dev = &adapter->vdev->dev;
1615         long rc;
1616
1617         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
1618
1619         /* Close the sub-crqs */
1620         do {
1621                 rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
1622                                         adapter->vdev->unit_address,
1623                                         scrq->crq_num);
1624         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
1625
1626         if (rc) {
1627                 netdev_err(adapter->netdev,
1628                            "Failed to release sub-CRQ %16lx, rc = %ld\n",
1629                            scrq->crq_num, rc);
1630         }
1631
1632         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1633                          DMA_BIDIRECTIONAL);
1634         free_pages((unsigned long)scrq->msgs, 2);
1635         kfree(scrq);
1636 }
1637
1638 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
1639                                                         *adapter)
1640 {
1641         struct device *dev = &adapter->vdev->dev;
1642         struct ibmvnic_sub_crq_queue *scrq;
1643         int rc;
1644
1645         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
1646         if (!scrq)
1647                 return NULL;
1648
1649         scrq->msgs =
1650                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
1651         if (!scrq->msgs) {
1652                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
1653                 goto zero_page_failed;
1654         }
1655
1656         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
1657                                          DMA_BIDIRECTIONAL);
1658         if (dma_mapping_error(dev, scrq->msg_token)) {
1659                 dev_warn(dev, "Couldn't map crq queue messages page\n");
1660                 goto map_failed;
1661         }
1662
1663         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
1664                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
1665
1666         if (rc == H_RESOURCE)
1667                 rc = ibmvnic_reset_crq(adapter);
1668
1669         if (rc == H_CLOSED) {
1670                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
1671         } else if (rc) {
1672                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
1673                 goto reg_failed;
1674         }
1675
1676         scrq->adapter = adapter;
1677         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
1678         spin_lock_init(&scrq->lock);
1679
1680         netdev_dbg(adapter->netdev,
1681                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
1682                    scrq->crq_num, scrq->hw_irq, scrq->irq);
1683
1684         return scrq;
1685
1686 reg_failed:
1687         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1688                          DMA_BIDIRECTIONAL);
1689 map_failed:
1690         free_pages((unsigned long)scrq->msgs, 2);
1691 zero_page_failed:
1692         kfree(scrq);
1693
1694         return NULL;
1695 }
1696
1697 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
1698 {
1699         int i;
1700
1701         if (adapter->tx_scrq) {
1702                 for (i = 0; i < adapter->req_tx_queues; i++) {
1703                         if (!adapter->tx_scrq[i])
1704                                 continue;
1705
1706                         if (adapter->tx_scrq[i]->irq) {
1707                                 free_irq(adapter->tx_scrq[i]->irq,
1708                                          adapter->tx_scrq[i]);
1709                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
1710                                 adapter->tx_scrq[i]->irq = 0;
1711                         }
1712
1713                         release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
1714                 }
1715
1716                 kfree(adapter->tx_scrq);
1717                 adapter->tx_scrq = NULL;
1718         }
1719
1720         if (adapter->rx_scrq) {
1721                 for (i = 0; i < adapter->req_rx_queues; i++) {
1722                         if (!adapter->rx_scrq[i])
1723                                 continue;
1724
1725                         if (adapter->rx_scrq[i]->irq) {
1726                                 free_irq(adapter->rx_scrq[i]->irq,
1727                                          adapter->rx_scrq[i]);
1728                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
1729                                 adapter->rx_scrq[i]->irq = 0;
1730                         }
1731
1732                         release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
1733                 }
1734
1735                 kfree(adapter->rx_scrq);
1736                 adapter->rx_scrq = NULL;
1737         }
1738 }
1739
1740 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
1741                             struct ibmvnic_sub_crq_queue *scrq)
1742 {
1743         struct device *dev = &adapter->vdev->dev;
1744         unsigned long rc;
1745
1746         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
1747                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
1748         if (rc)
1749                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
1750                         scrq->hw_irq, rc);
1751         return rc;
1752 }
1753
1754 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
1755                            struct ibmvnic_sub_crq_queue *scrq)
1756 {
1757         struct device *dev = &adapter->vdev->dev;
1758         unsigned long rc;
1759
1760         if (scrq->hw_irq > 0x100000000ULL) {
1761                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
1762                 return 1;
1763         }
1764
1765         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
1766                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
1767         if (rc)
1768                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
1769                         scrq->hw_irq, rc);
1770         return rc;
1771 }
1772
1773 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
1774                                struct ibmvnic_sub_crq_queue *scrq)
1775 {
1776         struct device *dev = &adapter->vdev->dev;
1777         struct ibmvnic_tx_buff *txbuff;
1778         union sub_crq *next;
1779         int index;
1780         int i, j;
1781         u8 first;
1782
1783 restart_loop:
1784         while (pending_scrq(adapter, scrq)) {
1785                 unsigned int pool = scrq->pool_index;
1786
1787                 next = ibmvnic_next_scrq(adapter, scrq);
1788                 for (i = 0; i < next->tx_comp.num_comps; i++) {
1789                         if (next->tx_comp.rcs[i]) {
1790                                 dev_err(dev, "tx error %x\n",
1791                                         next->tx_comp.rcs[i]);
1792                                 continue;
1793                         }
1794                         index = be32_to_cpu(next->tx_comp.correlators[i]);
1795                         txbuff = &adapter->tx_pool[pool].tx_buff[index];
1796
1797                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
1798                                 if (!txbuff->data_dma[j])
1799                                         continue;
1800
1801                                 txbuff->data_dma[j] = 0;
1802                         }
1803                         /* if sub_crq was sent indirectly */
1804                         first = txbuff->indir_arr[0].generic.first;
1805                         if (first == IBMVNIC_CRQ_CMD) {
1806                                 dma_unmap_single(dev, txbuff->indir_dma,
1807                                                  sizeof(txbuff->indir_arr),
1808                                                  DMA_TO_DEVICE);
1809                         }
1810
1811                         if (txbuff->last_frag) {
1812                                 dev_kfree_skb_any(txbuff->skb);
1813                                 txbuff->skb = NULL;
1814                         }
1815
1816                         adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
1817                                                      producer_index] = index;
1818                         adapter->tx_pool[pool].producer_index =
1819                             (adapter->tx_pool[pool].producer_index + 1) %
1820                             adapter->req_tx_entries_per_subcrq;
1821                 }
1822                 /* remove tx_comp scrq*/
1823                 next->tx_comp.first = 0;
1824
1825                 if (atomic_sub_return(next->tx_comp.num_comps, &scrq->used) <=
1826                     (adapter->req_tx_entries_per_subcrq / 2) &&
1827                     __netif_subqueue_stopped(adapter->netdev,
1828                                              scrq->pool_index)) {
1829                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
1830                         netdev_info(adapter->netdev, "Started queue %d\n",
1831                                     scrq->pool_index);
1832                 }
1833         }
1834
1835         enable_scrq_irq(adapter, scrq);
1836
1837         if (pending_scrq(adapter, scrq)) {
1838                 disable_scrq_irq(adapter, scrq);
1839                 goto restart_loop;
1840         }
1841
1842         return 0;
1843 }
1844
1845 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
1846 {
1847         struct ibmvnic_sub_crq_queue *scrq = instance;
1848         struct ibmvnic_adapter *adapter = scrq->adapter;
1849
1850         disable_scrq_irq(adapter, scrq);
1851         ibmvnic_complete_tx(adapter, scrq);
1852
1853         return IRQ_HANDLED;
1854 }
1855
1856 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
1857 {
1858         struct ibmvnic_sub_crq_queue *scrq = instance;
1859         struct ibmvnic_adapter *adapter = scrq->adapter;
1860
1861         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
1862                 disable_scrq_irq(adapter, scrq);
1863                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
1864         }
1865
1866         return IRQ_HANDLED;
1867 }
1868
1869 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
1870 {
1871         struct device *dev = &adapter->vdev->dev;
1872         struct ibmvnic_sub_crq_queue *scrq;
1873         int i = 0, j = 0;
1874         int rc = 0;
1875
1876         for (i = 0; i < adapter->req_tx_queues; i++) {
1877                 scrq = adapter->tx_scrq[i];
1878                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
1879
1880                 if (!scrq->irq) {
1881                         rc = -EINVAL;
1882                         dev_err(dev, "Error mapping irq\n");
1883                         goto req_tx_irq_failed;
1884                 }
1885
1886                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
1887                                  0, "ibmvnic_tx", scrq);
1888
1889                 if (rc) {
1890                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
1891                                 scrq->irq, rc);
1892                         irq_dispose_mapping(scrq->irq);
1893                         goto req_rx_irq_failed;
1894                 }
1895         }
1896
1897         for (i = 0; i < adapter->req_rx_queues; i++) {
1898                 scrq = adapter->rx_scrq[i];
1899                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
1900                 if (!scrq->irq) {
1901                         rc = -EINVAL;
1902                         dev_err(dev, "Error mapping irq\n");
1903                         goto req_rx_irq_failed;
1904                 }
1905                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
1906                                  0, "ibmvnic_rx", scrq);
1907                 if (rc) {
1908                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
1909                                 scrq->irq, rc);
1910                         irq_dispose_mapping(scrq->irq);
1911                         goto req_rx_irq_failed;
1912                 }
1913         }
1914         return rc;
1915
1916 req_rx_irq_failed:
1917         for (j = 0; j < i; j++) {
1918                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
1919                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
1920         }
1921         i = adapter->req_tx_queues;
1922 req_tx_irq_failed:
1923         for (j = 0; j < i; j++) {
1924                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
1925                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
1926         }
1927         release_sub_crqs(adapter);
1928         return rc;
1929 }
1930
1931 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
1932 {
1933         struct device *dev = &adapter->vdev->dev;
1934         struct ibmvnic_sub_crq_queue **allqueues;
1935         int registered_queues = 0;
1936         int total_queues;
1937         int more = 0;
1938         int i;
1939
1940         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
1941
1942         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
1943         if (!allqueues)
1944                 return -1;
1945
1946         for (i = 0; i < total_queues; i++) {
1947                 allqueues[i] = init_sub_crq_queue(adapter);
1948                 if (!allqueues[i]) {
1949                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
1950                         break;
1951                 }
1952                 registered_queues++;
1953         }
1954
1955         /* Make sure we were able to register the minimum number of queues */
1956         if (registered_queues <
1957             adapter->min_tx_queues + adapter->min_rx_queues) {
1958                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
1959                 goto tx_failed;
1960         }
1961
1962         /* Distribute the failed allocated queues*/
1963         for (i = 0; i < total_queues - registered_queues + more ; i++) {
1964                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
1965                 switch (i % 3) {
1966                 case 0:
1967                         if (adapter->req_rx_queues > adapter->min_rx_queues)
1968                                 adapter->req_rx_queues--;
1969                         else
1970                                 more++;
1971                         break;
1972                 case 1:
1973                         if (adapter->req_tx_queues > adapter->min_tx_queues)
1974                                 adapter->req_tx_queues--;
1975                         else
1976                                 more++;
1977                         break;
1978                 }
1979         }
1980
1981         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
1982                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
1983         if (!adapter->tx_scrq)
1984                 goto tx_failed;
1985
1986         for (i = 0; i < adapter->req_tx_queues; i++) {
1987                 adapter->tx_scrq[i] = allqueues[i];
1988                 adapter->tx_scrq[i]->pool_index = i;
1989         }
1990
1991         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
1992                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
1993         if (!adapter->rx_scrq)
1994                 goto rx_failed;
1995
1996         for (i = 0; i < adapter->req_rx_queues; i++) {
1997                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
1998                 adapter->rx_scrq[i]->scrq_num = i;
1999         }
2000
2001         kfree(allqueues);
2002         return 0;
2003
2004 rx_failed:
2005         kfree(adapter->tx_scrq);
2006         adapter->tx_scrq = NULL;
2007 tx_failed:
2008         for (i = 0; i < registered_queues; i++)
2009                 release_sub_crq_queue(adapter, allqueues[i]);
2010         kfree(allqueues);
2011         return -1;
2012 }
2013
2014 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2015 {
2016         struct device *dev = &adapter->vdev->dev;
2017         union ibmvnic_crq crq;
2018
2019         if (!retry) {
2020                 /* Sub-CRQ entries are 32 byte long */
2021                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2022
2023                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
2024                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
2025                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2026                         return;
2027                 }
2028
2029                 /* Get the minimum between the queried max and the entries
2030                  * that fit in our PAGE_SIZE
2031                  */
2032                 adapter->req_tx_entries_per_subcrq =
2033                     adapter->max_tx_entries_per_subcrq > entries_page ?
2034                     entries_page : adapter->max_tx_entries_per_subcrq;
2035                 adapter->req_rx_add_entries_per_subcrq =
2036                     adapter->max_rx_add_entries_per_subcrq > entries_page ?
2037                     entries_page : adapter->max_rx_add_entries_per_subcrq;
2038
2039                 adapter->req_tx_queues = adapter->opt_tx_comp_sub_queues;
2040                 adapter->req_rx_queues = adapter->opt_rx_comp_queues;
2041                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2042
2043                 adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2044         }
2045
2046         memset(&crq, 0, sizeof(crq));
2047         crq.request_capability.first = IBMVNIC_CRQ_CMD;
2048         crq.request_capability.cmd = REQUEST_CAPABILITY;
2049
2050         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2051         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2052         atomic_inc(&adapter->running_cap_crqs);
2053         ibmvnic_send_crq(adapter, &crq);
2054
2055         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2056         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2057         atomic_inc(&adapter->running_cap_crqs);
2058         ibmvnic_send_crq(adapter, &crq);
2059
2060         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2061         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2062         atomic_inc(&adapter->running_cap_crqs);
2063         ibmvnic_send_crq(adapter, &crq);
2064
2065         crq.request_capability.capability =
2066             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2067         crq.request_capability.number =
2068             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2069         atomic_inc(&adapter->running_cap_crqs);
2070         ibmvnic_send_crq(adapter, &crq);
2071
2072         crq.request_capability.capability =
2073             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2074         crq.request_capability.number =
2075             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2076         atomic_inc(&adapter->running_cap_crqs);
2077         ibmvnic_send_crq(adapter, &crq);
2078
2079         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2080         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2081         atomic_inc(&adapter->running_cap_crqs);
2082         ibmvnic_send_crq(adapter, &crq);
2083
2084         if (adapter->netdev->flags & IFF_PROMISC) {
2085                 if (adapter->promisc_supported) {
2086                         crq.request_capability.capability =
2087                             cpu_to_be16(PROMISC_REQUESTED);
2088                         crq.request_capability.number = cpu_to_be64(1);
2089                         atomic_inc(&adapter->running_cap_crqs);
2090                         ibmvnic_send_crq(adapter, &crq);
2091                 }
2092         } else {
2093                 crq.request_capability.capability =
2094                     cpu_to_be16(PROMISC_REQUESTED);
2095                 crq.request_capability.number = cpu_to_be64(0);
2096                 atomic_inc(&adapter->running_cap_crqs);
2097                 ibmvnic_send_crq(adapter, &crq);
2098         }
2099 }
2100
2101 static int pending_scrq(struct ibmvnic_adapter *adapter,
2102                         struct ibmvnic_sub_crq_queue *scrq)
2103 {
2104         union sub_crq *entry = &scrq->msgs[scrq->cur];
2105
2106         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP ||
2107             adapter->state == VNIC_CLOSING)
2108                 return 1;
2109         else
2110                 return 0;
2111 }
2112
2113 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2114                                         struct ibmvnic_sub_crq_queue *scrq)
2115 {
2116         union sub_crq *entry;
2117         unsigned long flags;
2118
2119         spin_lock_irqsave(&scrq->lock, flags);
2120         entry = &scrq->msgs[scrq->cur];
2121         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2122                 if (++scrq->cur == scrq->size)
2123                         scrq->cur = 0;
2124         } else {
2125                 entry = NULL;
2126         }
2127         spin_unlock_irqrestore(&scrq->lock, flags);
2128
2129         return entry;
2130 }
2131
2132 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2133 {
2134         struct ibmvnic_crq_queue *queue = &adapter->crq;
2135         union ibmvnic_crq *crq;
2136
2137         crq = &queue->msgs[queue->cur];
2138         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2139                 if (++queue->cur == queue->size)
2140                         queue->cur = 0;
2141         } else {
2142                 crq = NULL;
2143         }
2144
2145         return crq;
2146 }
2147
2148 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2149                        union sub_crq *sub_crq)
2150 {
2151         unsigned int ua = adapter->vdev->unit_address;
2152         struct device *dev = &adapter->vdev->dev;
2153         u64 *u64_crq = (u64 *)sub_crq;
2154         int rc;
2155
2156         netdev_dbg(adapter->netdev,
2157                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2158                    (unsigned long int)cpu_to_be64(remote_handle),
2159                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2160                    (unsigned long int)cpu_to_be64(u64_crq[1]),
2161                    (unsigned long int)cpu_to_be64(u64_crq[2]),
2162                    (unsigned long int)cpu_to_be64(u64_crq[3]));
2163
2164         /* Make sure the hypervisor sees the complete request */
2165         mb();
2166
2167         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2168                                 cpu_to_be64(remote_handle),
2169                                 cpu_to_be64(u64_crq[0]),
2170                                 cpu_to_be64(u64_crq[1]),
2171                                 cpu_to_be64(u64_crq[2]),
2172                                 cpu_to_be64(u64_crq[3]));
2173
2174         if (rc) {
2175                 if (rc == H_CLOSED)
2176                         dev_warn(dev, "CRQ Queue closed\n");
2177                 dev_err(dev, "Send error (rc=%d)\n", rc);
2178         }
2179
2180         return rc;
2181 }
2182
2183 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2184                                 u64 remote_handle, u64 ioba, u64 num_entries)
2185 {
2186         unsigned int ua = adapter->vdev->unit_address;
2187         struct device *dev = &adapter->vdev->dev;
2188         int rc;
2189
2190         /* Make sure the hypervisor sees the complete request */
2191         mb();
2192         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2193                                 cpu_to_be64(remote_handle),
2194                                 ioba, num_entries);
2195
2196         if (rc) {
2197                 if (rc == H_CLOSED)
2198                         dev_warn(dev, "CRQ Queue closed\n");
2199                 dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2200         }
2201
2202         return rc;
2203 }
2204
2205 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2206                             union ibmvnic_crq *crq)
2207 {
2208         unsigned int ua = adapter->vdev->unit_address;
2209         struct device *dev = &adapter->vdev->dev;
2210         u64 *u64_crq = (u64 *)crq;
2211         int rc;
2212
2213         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2214                    (unsigned long int)cpu_to_be64(u64_crq[0]),
2215                    (unsigned long int)cpu_to_be64(u64_crq[1]));
2216
2217         /* Make sure the hypervisor sees the complete request */
2218         mb();
2219
2220         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2221                                 cpu_to_be64(u64_crq[0]),
2222                                 cpu_to_be64(u64_crq[1]));
2223
2224         if (rc) {
2225                 if (rc == H_CLOSED)
2226                         dev_warn(dev, "CRQ Queue closed\n");
2227                 dev_warn(dev, "Send error (rc=%d)\n", rc);
2228         }
2229
2230         return rc;
2231 }
2232
2233 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2234 {
2235         union ibmvnic_crq crq;
2236
2237         memset(&crq, 0, sizeof(crq));
2238         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2239         crq.generic.cmd = IBMVNIC_CRQ_INIT;
2240         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2241
2242         return ibmvnic_send_crq(adapter, &crq);
2243 }
2244
2245 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2246 {
2247         union ibmvnic_crq crq;
2248
2249         memset(&crq, 0, sizeof(crq));
2250         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2251         crq.version_exchange.cmd = VERSION_EXCHANGE;
2252         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2253
2254         return ibmvnic_send_crq(adapter, &crq);
2255 }
2256
2257 static void send_login(struct ibmvnic_adapter *adapter)
2258 {
2259         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
2260         struct ibmvnic_login_buffer *login_buffer;
2261         struct device *dev = &adapter->vdev->dev;
2262         dma_addr_t rsp_buffer_token;
2263         dma_addr_t buffer_token;
2264         size_t rsp_buffer_size;
2265         union ibmvnic_crq crq;
2266         size_t buffer_size;
2267         __be64 *tx_list_p;
2268         __be64 *rx_list_p;
2269         int i;
2270
2271         buffer_size =
2272             sizeof(struct ibmvnic_login_buffer) +
2273             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues);
2274
2275         login_buffer = kmalloc(buffer_size, GFP_ATOMIC);
2276         if (!login_buffer)
2277                 goto buf_alloc_failed;
2278
2279         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
2280                                       DMA_TO_DEVICE);
2281         if (dma_mapping_error(dev, buffer_token)) {
2282                 dev_err(dev, "Couldn't map login buffer\n");
2283                 goto buf_map_failed;
2284         }
2285
2286         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
2287                           sizeof(u64) * adapter->req_tx_queues +
2288                           sizeof(u64) * adapter->req_rx_queues +
2289                           sizeof(u64) * adapter->req_rx_queues +
2290                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
2291
2292         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
2293         if (!login_rsp_buffer)
2294                 goto buf_rsp_alloc_failed;
2295
2296         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
2297                                           rsp_buffer_size, DMA_FROM_DEVICE);
2298         if (dma_mapping_error(dev, rsp_buffer_token)) {
2299                 dev_err(dev, "Couldn't map login rsp buffer\n");
2300                 goto buf_rsp_map_failed;
2301         }
2302
2303         adapter->login_buf = login_buffer;
2304         adapter->login_buf_token = buffer_token;
2305         adapter->login_buf_sz = buffer_size;
2306         adapter->login_rsp_buf = login_rsp_buffer;
2307         adapter->login_rsp_buf_token = rsp_buffer_token;
2308         adapter->login_rsp_buf_sz = rsp_buffer_size;
2309
2310         login_buffer->len = cpu_to_be32(buffer_size);
2311         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
2312         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
2313         login_buffer->off_txcomp_subcrqs =
2314             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
2315         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
2316         login_buffer->off_rxcomp_subcrqs =
2317             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
2318                         sizeof(u64) * adapter->req_tx_queues);
2319         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
2320         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
2321
2322         tx_list_p = (__be64 *)((char *)login_buffer +
2323                                       sizeof(struct ibmvnic_login_buffer));
2324         rx_list_p = (__be64 *)((char *)login_buffer +
2325                                       sizeof(struct ibmvnic_login_buffer) +
2326                                       sizeof(u64) * adapter->req_tx_queues);
2327
2328         for (i = 0; i < adapter->req_tx_queues; i++) {
2329                 if (adapter->tx_scrq[i]) {
2330                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
2331                                                    crq_num);
2332                 }
2333         }
2334
2335         for (i = 0; i < adapter->req_rx_queues; i++) {
2336                 if (adapter->rx_scrq[i]) {
2337                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
2338                                                    crq_num);
2339                 }
2340         }
2341
2342         netdev_dbg(adapter->netdev, "Login Buffer:\n");
2343         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
2344                 netdev_dbg(adapter->netdev, "%016lx\n",
2345                            ((unsigned long int *)(adapter->login_buf))[i]);
2346         }
2347
2348         memset(&crq, 0, sizeof(crq));
2349         crq.login.first = IBMVNIC_CRQ_CMD;
2350         crq.login.cmd = LOGIN;
2351         crq.login.ioba = cpu_to_be32(buffer_token);
2352         crq.login.len = cpu_to_be32(buffer_size);
2353         ibmvnic_send_crq(adapter, &crq);
2354
2355         return;
2356
2357 buf_rsp_map_failed:
2358         kfree(login_rsp_buffer);
2359 buf_rsp_alloc_failed:
2360         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
2361 buf_map_failed:
2362         kfree(login_buffer);
2363 buf_alloc_failed:
2364         return;
2365 }
2366
2367 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
2368                              u32 len, u8 map_id)
2369 {
2370         union ibmvnic_crq crq;
2371
2372         memset(&crq, 0, sizeof(crq));
2373         crq.request_map.first = IBMVNIC_CRQ_CMD;
2374         crq.request_map.cmd = REQUEST_MAP;
2375         crq.request_map.map_id = map_id;
2376         crq.request_map.ioba = cpu_to_be32(addr);
2377         crq.request_map.len = cpu_to_be32(len);
2378         ibmvnic_send_crq(adapter, &crq);
2379 }
2380
2381 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
2382 {
2383         union ibmvnic_crq crq;
2384
2385         memset(&crq, 0, sizeof(crq));
2386         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
2387         crq.request_unmap.cmd = REQUEST_UNMAP;
2388         crq.request_unmap.map_id = map_id;
2389         ibmvnic_send_crq(adapter, &crq);
2390 }
2391
2392 static void send_map_query(struct ibmvnic_adapter *adapter)
2393 {
2394         union ibmvnic_crq crq;
2395
2396         memset(&crq, 0, sizeof(crq));
2397         crq.query_map.first = IBMVNIC_CRQ_CMD;
2398         crq.query_map.cmd = QUERY_MAP;
2399         ibmvnic_send_crq(adapter, &crq);
2400 }
2401
2402 /* Send a series of CRQs requesting various capabilities of the VNIC server */
2403 static void send_cap_queries(struct ibmvnic_adapter *adapter)
2404 {
2405         union ibmvnic_crq crq;
2406
2407         atomic_set(&adapter->running_cap_crqs, 0);
2408         memset(&crq, 0, sizeof(crq));
2409         crq.query_capability.first = IBMVNIC_CRQ_CMD;
2410         crq.query_capability.cmd = QUERY_CAPABILITY;
2411
2412         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
2413         atomic_inc(&adapter->running_cap_crqs);
2414         ibmvnic_send_crq(adapter, &crq);
2415
2416         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
2417         atomic_inc(&adapter->running_cap_crqs);
2418         ibmvnic_send_crq(adapter, &crq);
2419
2420         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
2421         atomic_inc(&adapter->running_cap_crqs);
2422         ibmvnic_send_crq(adapter, &crq);
2423
2424         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
2425         atomic_inc(&adapter->running_cap_crqs);
2426         ibmvnic_send_crq(adapter, &crq);
2427
2428         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
2429         atomic_inc(&adapter->running_cap_crqs);
2430         ibmvnic_send_crq(adapter, &crq);
2431
2432         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
2433         atomic_inc(&adapter->running_cap_crqs);
2434         ibmvnic_send_crq(adapter, &crq);
2435
2436         crq.query_capability.capability =
2437             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
2438         atomic_inc(&adapter->running_cap_crqs);
2439         ibmvnic_send_crq(adapter, &crq);
2440
2441         crq.query_capability.capability =
2442             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
2443         atomic_inc(&adapter->running_cap_crqs);
2444         ibmvnic_send_crq(adapter, &crq);
2445
2446         crq.query_capability.capability =
2447             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
2448         atomic_inc(&adapter->running_cap_crqs);
2449         ibmvnic_send_crq(adapter, &crq);
2450
2451         crq.query_capability.capability =
2452             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
2453         atomic_inc(&adapter->running_cap_crqs);
2454         ibmvnic_send_crq(adapter, &crq);
2455
2456         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
2457         atomic_inc(&adapter->running_cap_crqs);
2458         ibmvnic_send_crq(adapter, &crq);
2459
2460         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
2461         atomic_inc(&adapter->running_cap_crqs);
2462         ibmvnic_send_crq(adapter, &crq);
2463
2464         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
2465         atomic_inc(&adapter->running_cap_crqs);
2466         ibmvnic_send_crq(adapter, &crq);
2467
2468         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
2469         atomic_inc(&adapter->running_cap_crqs);
2470         ibmvnic_send_crq(adapter, &crq);
2471
2472         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
2473         atomic_inc(&adapter->running_cap_crqs);
2474         ibmvnic_send_crq(adapter, &crq);
2475
2476         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
2477         atomic_inc(&adapter->running_cap_crqs);
2478         ibmvnic_send_crq(adapter, &crq);
2479
2480         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
2481         atomic_inc(&adapter->running_cap_crqs);
2482         ibmvnic_send_crq(adapter, &crq);
2483
2484         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
2485         atomic_inc(&adapter->running_cap_crqs);
2486         ibmvnic_send_crq(adapter, &crq);
2487
2488         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
2489         atomic_inc(&adapter->running_cap_crqs);
2490         ibmvnic_send_crq(adapter, &crq);
2491
2492         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
2493         atomic_inc(&adapter->running_cap_crqs);
2494         ibmvnic_send_crq(adapter, &crq);
2495
2496         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
2497         atomic_inc(&adapter->running_cap_crqs);
2498         ibmvnic_send_crq(adapter, &crq);
2499
2500         crq.query_capability.capability =
2501                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
2502         atomic_inc(&adapter->running_cap_crqs);
2503         ibmvnic_send_crq(adapter, &crq);
2504
2505         crq.query_capability.capability =
2506                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
2507         atomic_inc(&adapter->running_cap_crqs);
2508         ibmvnic_send_crq(adapter, &crq);
2509
2510         crq.query_capability.capability =
2511                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
2512         atomic_inc(&adapter->running_cap_crqs);
2513         ibmvnic_send_crq(adapter, &crq);
2514
2515         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
2516         atomic_inc(&adapter->running_cap_crqs);
2517         ibmvnic_send_crq(adapter, &crq);
2518 }
2519
2520 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
2521 {
2522         struct device *dev = &adapter->vdev->dev;
2523         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
2524         union ibmvnic_crq crq;
2525         int i;
2526
2527         dma_unmap_single(dev, adapter->ip_offload_tok,
2528                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
2529
2530         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
2531         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
2532                 netdev_dbg(adapter->netdev, "%016lx\n",
2533                            ((unsigned long int *)(buf))[i]);
2534
2535         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
2536         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
2537         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
2538                    buf->tcp_ipv4_chksum);
2539         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
2540                    buf->tcp_ipv6_chksum);
2541         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
2542                    buf->udp_ipv4_chksum);
2543         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
2544                    buf->udp_ipv6_chksum);
2545         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
2546                    buf->large_tx_ipv4);
2547         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
2548                    buf->large_tx_ipv6);
2549         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
2550                    buf->large_rx_ipv4);
2551         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
2552                    buf->large_rx_ipv6);
2553         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
2554                    buf->max_ipv4_header_size);
2555         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
2556                    buf->max_ipv6_header_size);
2557         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
2558                    buf->max_tcp_header_size);
2559         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
2560                    buf->max_udp_header_size);
2561         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
2562                    buf->max_large_tx_size);
2563         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
2564                    buf->max_large_rx_size);
2565         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
2566                    buf->ipv6_extension_header);
2567         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
2568                    buf->tcp_pseudosum_req);
2569         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
2570                    buf->num_ipv6_ext_headers);
2571         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
2572                    buf->off_ipv6_ext_headers);
2573
2574         adapter->ip_offload_ctrl_tok =
2575             dma_map_single(dev, &adapter->ip_offload_ctrl,
2576                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
2577
2578         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
2579                 dev_err(dev, "Couldn't map ip offload control buffer\n");
2580                 return;
2581         }
2582
2583         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
2584         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
2585         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
2586         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
2587         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
2588
2589         /* large_tx/rx disabled for now, additional features needed */
2590         adapter->ip_offload_ctrl.large_tx_ipv4 = 0;
2591         adapter->ip_offload_ctrl.large_tx_ipv6 = 0;
2592         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
2593         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
2594
2595         adapter->netdev->features = NETIF_F_GSO;
2596
2597         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
2598                 adapter->netdev->features |= NETIF_F_IP_CSUM;
2599
2600         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
2601                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
2602
2603         if ((adapter->netdev->features &
2604             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
2605                 adapter->netdev->features |= NETIF_F_RXCSUM;
2606
2607         memset(&crq, 0, sizeof(crq));
2608         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
2609         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
2610         crq.control_ip_offload.len =
2611             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
2612         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
2613         ibmvnic_send_crq(adapter, &crq);
2614 }
2615
2616 static void handle_error_info_rsp(union ibmvnic_crq *crq,
2617                                   struct ibmvnic_adapter *adapter)
2618 {
2619         struct device *dev = &adapter->vdev->dev;
2620         struct ibmvnic_error_buff *error_buff, *tmp;
2621         unsigned long flags;
2622         bool found = false;
2623         int i;
2624
2625         if (!crq->request_error_rsp.rc.code) {
2626                 dev_info(dev, "Request Error Rsp returned with rc=%x\n",
2627                          crq->request_error_rsp.rc.code);
2628                 return;
2629         }
2630
2631         spin_lock_irqsave(&adapter->error_list_lock, flags);
2632         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
2633                 if (error_buff->error_id == crq->request_error_rsp.error_id) {
2634                         found = true;
2635                         list_del(&error_buff->list);
2636                         break;
2637                 }
2638         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2639
2640         if (!found) {
2641                 dev_err(dev, "Couldn't find error id %x\n",
2642                         be32_to_cpu(crq->request_error_rsp.error_id));
2643                 return;
2644         }
2645
2646         dev_err(dev, "Detailed info for error id %x:",
2647                 be32_to_cpu(crq->request_error_rsp.error_id));
2648
2649         for (i = 0; i < error_buff->len; i++) {
2650                 pr_cont("%02x", (int)error_buff->buff[i]);
2651                 if (i % 8 == 7)
2652                         pr_cont(" ");
2653         }
2654         pr_cont("\n");
2655
2656         dma_unmap_single(dev, error_buff->dma, error_buff->len,
2657                          DMA_FROM_DEVICE);
2658         kfree(error_buff->buff);
2659         kfree(error_buff);
2660 }
2661
2662 static void request_error_information(struct ibmvnic_adapter *adapter,
2663                                       union ibmvnic_crq *err_crq)
2664 {
2665         struct device *dev = &adapter->vdev->dev;
2666         struct net_device *netdev = adapter->netdev;
2667         struct ibmvnic_error_buff *error_buff;
2668         unsigned long timeout = msecs_to_jiffies(30000);
2669         union ibmvnic_crq crq;
2670         unsigned long flags;
2671         int rc, detail_len;
2672
2673         error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
2674         if (!error_buff)
2675                 return;
2676
2677         detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
2678         error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
2679         if (!error_buff->buff) {
2680                 kfree(error_buff);
2681                 return;
2682         }
2683
2684         error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
2685                                          DMA_FROM_DEVICE);
2686         if (dma_mapping_error(dev, error_buff->dma)) {
2687                 netdev_err(netdev, "Couldn't map error buffer\n");
2688                 kfree(error_buff->buff);
2689                 kfree(error_buff);
2690                 return;
2691         }
2692
2693         error_buff->len = detail_len;
2694         error_buff->error_id = err_crq->error_indication.error_id;
2695
2696         spin_lock_irqsave(&adapter->error_list_lock, flags);
2697         list_add_tail(&error_buff->list, &adapter->errors);
2698         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2699
2700         memset(&crq, 0, sizeof(crq));
2701         crq.request_error_info.first = IBMVNIC_CRQ_CMD;
2702         crq.request_error_info.cmd = REQUEST_ERROR_INFO;
2703         crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
2704         crq.request_error_info.len = cpu_to_be32(detail_len);
2705         crq.request_error_info.error_id = err_crq->error_indication.error_id;
2706
2707         rc = ibmvnic_send_crq(adapter, &crq);
2708         if (rc) {
2709                 netdev_err(netdev, "failed to request error information\n");
2710                 goto err_info_fail;
2711         }
2712
2713         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
2714                 netdev_err(netdev, "timeout waiting for error information\n");
2715                 goto err_info_fail;
2716         }
2717
2718         return;
2719
2720 err_info_fail:
2721         spin_lock_irqsave(&adapter->error_list_lock, flags);
2722         list_del(&error_buff->list);
2723         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2724
2725         kfree(error_buff->buff);
2726         kfree(error_buff);
2727 }
2728
2729 static void handle_error_indication(union ibmvnic_crq *crq,
2730                                     struct ibmvnic_adapter *adapter)
2731 {
2732         struct device *dev = &adapter->vdev->dev;
2733
2734         dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
2735                 crq->error_indication.flags
2736                         & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
2737                 be32_to_cpu(crq->error_indication.error_id),
2738                 be16_to_cpu(crq->error_indication.error_cause));
2739
2740         if (be32_to_cpu(crq->error_indication.error_id))
2741                 request_error_information(adapter, crq);
2742
2743         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
2744                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
2745 }
2746
2747 static void handle_change_mac_rsp(union ibmvnic_crq *crq,
2748                                   struct ibmvnic_adapter *adapter)
2749 {
2750         struct net_device *netdev = adapter->netdev;
2751         struct device *dev = &adapter->vdev->dev;
2752         long rc;
2753
2754         rc = crq->change_mac_addr_rsp.rc.code;
2755         if (rc) {
2756                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
2757                 return;
2758         }
2759         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
2760                ETH_ALEN);
2761 }
2762
2763 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
2764                                    struct ibmvnic_adapter *adapter)
2765 {
2766         struct device *dev = &adapter->vdev->dev;
2767         u64 *req_value;
2768         char *name;
2769
2770         atomic_dec(&adapter->running_cap_crqs);
2771         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
2772         case REQ_TX_QUEUES:
2773                 req_value = &adapter->req_tx_queues;
2774                 name = "tx";
2775                 break;
2776         case REQ_RX_QUEUES:
2777                 req_value = &adapter->req_rx_queues;
2778                 name = "rx";
2779                 break;
2780         case REQ_RX_ADD_QUEUES:
2781                 req_value = &adapter->req_rx_add_queues;
2782                 name = "rx_add";
2783                 break;
2784         case REQ_TX_ENTRIES_PER_SUBCRQ:
2785                 req_value = &adapter->req_tx_entries_per_subcrq;
2786                 name = "tx_entries_per_subcrq";
2787                 break;
2788         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
2789                 req_value = &adapter->req_rx_add_entries_per_subcrq;
2790                 name = "rx_add_entries_per_subcrq";
2791                 break;
2792         case REQ_MTU:
2793                 req_value = &adapter->req_mtu;
2794                 name = "mtu";
2795                 break;
2796         case PROMISC_REQUESTED:
2797                 req_value = &adapter->promisc;
2798                 name = "promisc";
2799                 break;
2800         default:
2801                 dev_err(dev, "Got invalid cap request rsp %d\n",
2802                         crq->request_capability.capability);
2803                 return;
2804         }
2805
2806         switch (crq->request_capability_rsp.rc.code) {
2807         case SUCCESS:
2808                 break;
2809         case PARTIALSUCCESS:
2810                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
2811                          *req_value,
2812                          (long int)be64_to_cpu(crq->request_capability_rsp.
2813                                                number), name);
2814                 release_sub_crqs(adapter);
2815                 *req_value = be64_to_cpu(crq->request_capability_rsp.number);
2816                 ibmvnic_send_req_caps(adapter, 1);
2817                 return;
2818         default:
2819                 dev_err(dev, "Error %d in request cap rsp\n",
2820                         crq->request_capability_rsp.rc.code);
2821                 return;
2822         }
2823
2824         /* Done receiving requested capabilities, query IP offload support */
2825         if (atomic_read(&adapter->running_cap_crqs) == 0) {
2826                 union ibmvnic_crq newcrq;
2827                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
2828                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
2829                     &adapter->ip_offload_buf;
2830
2831                 adapter->wait_capability = false;
2832                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
2833                                                          buf_sz,
2834                                                          DMA_FROM_DEVICE);
2835
2836                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
2837                         if (!firmware_has_feature(FW_FEATURE_CMO))
2838                                 dev_err(dev, "Couldn't map offload buffer\n");
2839                         return;
2840                 }
2841
2842                 memset(&newcrq, 0, sizeof(newcrq));
2843                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
2844                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
2845                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
2846                 newcrq.query_ip_offload.ioba =
2847                     cpu_to_be32(adapter->ip_offload_tok);
2848
2849                 ibmvnic_send_crq(adapter, &newcrq);
2850         }
2851 }
2852
2853 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
2854                             struct ibmvnic_adapter *adapter)
2855 {
2856         struct device *dev = &adapter->vdev->dev;
2857         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
2858         struct ibmvnic_login_buffer *login = adapter->login_buf;
2859         int i;
2860
2861         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
2862                          DMA_BIDIRECTIONAL);
2863         dma_unmap_single(dev, adapter->login_rsp_buf_token,
2864                          adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
2865
2866         /* If the number of queues requested can't be allocated by the
2867          * server, the login response will return with code 1. We will need
2868          * to resend the login buffer with fewer queues requested.
2869          */
2870         if (login_rsp_crq->generic.rc.code) {
2871                 adapter->renegotiate = true;
2872                 complete(&adapter->init_done);
2873                 return 0;
2874         }
2875
2876         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
2877         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
2878                 netdev_dbg(adapter->netdev, "%016lx\n",
2879                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
2880         }
2881
2882         /* Sanity checks */
2883         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
2884             (be32_to_cpu(login->num_rxcomp_subcrqs) *
2885              adapter->req_rx_add_queues !=
2886              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
2887                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
2888                 ibmvnic_remove(adapter->vdev);
2889                 return -EIO;
2890         }
2891         complete(&adapter->init_done);
2892
2893         return 0;
2894 }
2895
2896 static void handle_request_map_rsp(union ibmvnic_crq *crq,
2897                                    struct ibmvnic_adapter *adapter)
2898 {
2899         struct device *dev = &adapter->vdev->dev;
2900         u8 map_id = crq->request_map_rsp.map_id;
2901         int tx_subcrqs;
2902         int rx_subcrqs;
2903         long rc;
2904         int i;
2905
2906         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
2907         rx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
2908
2909         rc = crq->request_map_rsp.rc.code;
2910         if (rc) {
2911                 dev_err(dev, "Error %ld in REQUEST_MAP_RSP\n", rc);
2912                 adapter->map_id--;
2913                 /* need to find and zero tx/rx_pool map_id */
2914                 for (i = 0; i < tx_subcrqs; i++) {
2915                         if (adapter->tx_pool[i].long_term_buff.map_id == map_id)
2916                                 adapter->tx_pool[i].long_term_buff.map_id = 0;
2917                 }
2918                 for (i = 0; i < rx_subcrqs; i++) {
2919                         if (adapter->rx_pool[i].long_term_buff.map_id == map_id)
2920                                 adapter->rx_pool[i].long_term_buff.map_id = 0;
2921                 }
2922         }
2923         complete(&adapter->fw_done);
2924 }
2925
2926 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
2927                                      struct ibmvnic_adapter *adapter)
2928 {
2929         struct device *dev = &adapter->vdev->dev;
2930         long rc;
2931
2932         rc = crq->request_unmap_rsp.rc.code;
2933         if (rc)
2934                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
2935 }
2936
2937 static void handle_query_map_rsp(union ibmvnic_crq *crq,
2938                                  struct ibmvnic_adapter *adapter)
2939 {
2940         struct net_device *netdev = adapter->netdev;
2941         struct device *dev = &adapter->vdev->dev;
2942         long rc;
2943
2944         rc = crq->query_map_rsp.rc.code;
2945         if (rc) {
2946                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
2947                 return;
2948         }
2949         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
2950                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
2951                    crq->query_map_rsp.free_pages);
2952 }
2953
2954 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
2955                                  struct ibmvnic_adapter *adapter)
2956 {
2957         struct net_device *netdev = adapter->netdev;
2958         struct device *dev = &adapter->vdev->dev;
2959         long rc;
2960
2961         atomic_dec(&adapter->running_cap_crqs);
2962         netdev_dbg(netdev, "Outstanding queries: %d\n",
2963                    atomic_read(&adapter->running_cap_crqs));
2964         rc = crq->query_capability.rc.code;
2965         if (rc) {
2966                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
2967                 goto out;
2968         }
2969
2970         switch (be16_to_cpu(crq->query_capability.capability)) {
2971         case MIN_TX_QUEUES:
2972                 adapter->min_tx_queues =
2973                     be64_to_cpu(crq->query_capability.number);
2974                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
2975                            adapter->min_tx_queues);
2976                 break;
2977         case MIN_RX_QUEUES:
2978                 adapter->min_rx_queues =
2979                     be64_to_cpu(crq->query_capability.number);
2980                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
2981                            adapter->min_rx_queues);
2982                 break;
2983         case MIN_RX_ADD_QUEUES:
2984                 adapter->min_rx_add_queues =
2985                     be64_to_cpu(crq->query_capability.number);
2986                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
2987                            adapter->min_rx_add_queues);
2988                 break;
2989         case MAX_TX_QUEUES:
2990                 adapter->max_tx_queues =
2991                     be64_to_cpu(crq->query_capability.number);
2992                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
2993                            adapter->max_tx_queues);
2994                 break;
2995         case MAX_RX_QUEUES:
2996                 adapter->max_rx_queues =
2997                     be64_to_cpu(crq->query_capability.number);
2998                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
2999                            adapter->max_rx_queues);
3000                 break;
3001         case MAX_RX_ADD_QUEUES:
3002                 adapter->max_rx_add_queues =
3003                     be64_to_cpu(crq->query_capability.number);
3004                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3005                            adapter->max_rx_add_queues);
3006                 break;
3007         case MIN_TX_ENTRIES_PER_SUBCRQ:
3008                 adapter->min_tx_entries_per_subcrq =
3009                     be64_to_cpu(crq->query_capability.number);
3010                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3011                            adapter->min_tx_entries_per_subcrq);
3012                 break;
3013         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3014                 adapter->min_rx_add_entries_per_subcrq =
3015                     be64_to_cpu(crq->query_capability.number);
3016                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3017                            adapter->min_rx_add_entries_per_subcrq);
3018                 break;
3019         case MAX_TX_ENTRIES_PER_SUBCRQ:
3020                 adapter->max_tx_entries_per_subcrq =
3021                     be64_to_cpu(crq->query_capability.number);
3022                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3023                            adapter->max_tx_entries_per_subcrq);
3024                 break;
3025         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3026                 adapter->max_rx_add_entries_per_subcrq =
3027                     be64_to_cpu(crq->query_capability.number);
3028                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3029                            adapter->max_rx_add_entries_per_subcrq);
3030                 break;
3031         case TCP_IP_OFFLOAD:
3032                 adapter->tcp_ip_offload =
3033                     be64_to_cpu(crq->query_capability.number);
3034                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3035                            adapter->tcp_ip_offload);
3036                 break;
3037         case PROMISC_SUPPORTED:
3038                 adapter->promisc_supported =
3039                     be64_to_cpu(crq->query_capability.number);
3040                 netdev_dbg(netdev, "promisc_supported = %lld\n",
3041                            adapter->promisc_supported);
3042                 break;
3043         case MIN_MTU:
3044                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3045                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3046                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3047                 break;
3048         case MAX_MTU:
3049                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3050                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3051                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3052                 break;
3053         case MAX_MULTICAST_FILTERS:
3054                 adapter->max_multicast_filters =
3055                     be64_to_cpu(crq->query_capability.number);
3056                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3057                            adapter->max_multicast_filters);
3058                 break;
3059         case VLAN_HEADER_INSERTION:
3060                 adapter->vlan_header_insertion =
3061                     be64_to_cpu(crq->query_capability.number);
3062                 if (adapter->vlan_header_insertion)
3063                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3064                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3065                            adapter->vlan_header_insertion);
3066                 break;
3067         case RX_VLAN_HEADER_INSERTION:
3068                 adapter->rx_vlan_header_insertion =
3069                     be64_to_cpu(crq->query_capability.number);
3070                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3071                            adapter->rx_vlan_header_insertion);
3072                 break;
3073         case MAX_TX_SG_ENTRIES:
3074                 adapter->max_tx_sg_entries =
3075                     be64_to_cpu(crq->query_capability.number);
3076                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3077                            adapter->max_tx_sg_entries);
3078                 break;
3079         case RX_SG_SUPPORTED:
3080                 adapter->rx_sg_supported =
3081                     be64_to_cpu(crq->query_capability.number);
3082                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3083                            adapter->rx_sg_supported);
3084                 break;
3085         case OPT_TX_COMP_SUB_QUEUES:
3086                 adapter->opt_tx_comp_sub_queues =
3087                     be64_to_cpu(crq->query_capability.number);
3088                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3089                            adapter->opt_tx_comp_sub_queues);
3090                 break;
3091         case OPT_RX_COMP_QUEUES:
3092                 adapter->opt_rx_comp_queues =
3093                     be64_to_cpu(crq->query_capability.number);
3094                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3095                            adapter->opt_rx_comp_queues);
3096                 break;
3097         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3098                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
3099                     be64_to_cpu(crq->query_capability.number);
3100                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3101                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
3102                 break;
3103         case OPT_TX_ENTRIES_PER_SUBCRQ:
3104                 adapter->opt_tx_entries_per_subcrq =
3105                     be64_to_cpu(crq->query_capability.number);
3106                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3107                            adapter->opt_tx_entries_per_subcrq);
3108                 break;
3109         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3110                 adapter->opt_rxba_entries_per_subcrq =
3111                     be64_to_cpu(crq->query_capability.number);
3112                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3113                            adapter->opt_rxba_entries_per_subcrq);
3114                 break;
3115         case TX_RX_DESC_REQ:
3116                 adapter->tx_rx_desc_req = crq->query_capability.number;
3117                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3118                            adapter->tx_rx_desc_req);
3119                 break;
3120
3121         default:
3122                 netdev_err(netdev, "Got invalid cap rsp %d\n",
3123                            crq->query_capability.capability);
3124         }
3125
3126 out:
3127         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3128                 adapter->wait_capability = false;
3129                 ibmvnic_send_req_caps(adapter, 0);
3130         }
3131 }
3132
3133 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
3134                                struct ibmvnic_adapter *adapter)
3135 {
3136         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
3137         struct net_device *netdev = adapter->netdev;
3138         struct device *dev = &adapter->vdev->dev;
3139         u64 *u64_crq = (u64 *)crq;
3140         long rc;
3141
3142         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
3143                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3144                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3145         switch (gen_crq->first) {
3146         case IBMVNIC_CRQ_INIT_RSP:
3147                 switch (gen_crq->cmd) {
3148                 case IBMVNIC_CRQ_INIT:
3149                         dev_info(dev, "Partner initialized\n");
3150                         break;
3151                 case IBMVNIC_CRQ_INIT_COMPLETE:
3152                         dev_info(dev, "Partner initialization complete\n");
3153                         send_version_xchg(adapter);
3154                         break;
3155                 default:
3156                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
3157                 }
3158                 return;
3159         case IBMVNIC_CRQ_XPORT_EVENT:
3160                 netif_carrier_off(netdev);
3161                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
3162                         dev_info(dev, "Migrated, re-enabling adapter\n");
3163                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
3164                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
3165                         dev_info(dev, "Backing device failover detected\n");
3166                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
3167                 } else {
3168                         /* The adapter lost the connection */
3169                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
3170                                 gen_crq->cmd);
3171                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3172                 }
3173                 return;
3174         case IBMVNIC_CRQ_CMD_RSP:
3175                 break;
3176         default:
3177                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
3178                         gen_crq->first);
3179                 return;
3180         }
3181
3182         switch (gen_crq->cmd) {
3183         case VERSION_EXCHANGE_RSP:
3184                 rc = crq->version_exchange_rsp.rc.code;
3185                 if (rc) {
3186                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
3187                         break;
3188                 }
3189                 dev_info(dev, "Partner protocol version is %d\n",
3190                          crq->version_exchange_rsp.version);
3191                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
3192                     ibmvnic_version)
3193                         ibmvnic_version =
3194                             be16_to_cpu(crq->version_exchange_rsp.version);
3195                 send_cap_queries(adapter);
3196                 break;
3197         case QUERY_CAPABILITY_RSP:
3198                 handle_query_cap_rsp(crq, adapter);
3199                 break;
3200         case QUERY_MAP_RSP:
3201                 handle_query_map_rsp(crq, adapter);
3202                 break;
3203         case REQUEST_MAP_RSP:
3204                 handle_request_map_rsp(crq, adapter);
3205                 break;
3206         case REQUEST_UNMAP_RSP:
3207                 handle_request_unmap_rsp(crq, adapter);
3208                 break;
3209         case REQUEST_CAPABILITY_RSP:
3210                 handle_request_cap_rsp(crq, adapter);
3211                 break;
3212         case LOGIN_RSP:
3213                 netdev_dbg(netdev, "Got Login Response\n");
3214                 handle_login_rsp(crq, adapter);
3215                 break;
3216         case LOGICAL_LINK_STATE_RSP:
3217                 netdev_dbg(netdev,
3218                            "Got Logical Link State Response, state: %d rc: %d\n",
3219                            crq->logical_link_state_rsp.link_state,
3220                            crq->logical_link_state_rsp.rc.code);
3221                 adapter->logical_link_state =
3222                     crq->logical_link_state_rsp.link_state;
3223                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
3224                 complete(&adapter->init_done);
3225                 break;
3226         case LINK_STATE_INDICATION:
3227                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
3228                 adapter->phys_link_state =
3229                     crq->link_state_indication.phys_link_state;
3230                 adapter->logical_link_state =
3231                     crq->link_state_indication.logical_link_state;
3232                 break;
3233         case CHANGE_MAC_ADDR_RSP:
3234                 netdev_dbg(netdev, "Got MAC address change Response\n");
3235                 handle_change_mac_rsp(crq, adapter);
3236                 break;
3237         case ERROR_INDICATION:
3238                 netdev_dbg(netdev, "Got Error Indication\n");
3239                 handle_error_indication(crq, adapter);
3240                 break;
3241         case REQUEST_ERROR_RSP:
3242                 netdev_dbg(netdev, "Got Error Detail Response\n");
3243                 handle_error_info_rsp(crq, adapter);
3244                 break;
3245         case REQUEST_STATISTICS_RSP:
3246                 netdev_dbg(netdev, "Got Statistics Response\n");
3247                 complete(&adapter->stats_done);
3248                 break;
3249         case QUERY_IP_OFFLOAD_RSP:
3250                 netdev_dbg(netdev, "Got Query IP offload Response\n");
3251                 handle_query_ip_offload_rsp(adapter);
3252                 break;
3253         case MULTICAST_CTRL_RSP:
3254                 netdev_dbg(netdev, "Got multicast control Response\n");
3255                 break;
3256         case CONTROL_IP_OFFLOAD_RSP:
3257                 netdev_dbg(netdev, "Got Control IP offload Response\n");
3258                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
3259                                  sizeof(adapter->ip_offload_ctrl),
3260                                  DMA_TO_DEVICE);
3261                 complete(&adapter->init_done);
3262                 break;
3263         case COLLECT_FW_TRACE_RSP:
3264                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
3265                 complete(&adapter->fw_done);
3266                 break;
3267         default:
3268                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
3269                            gen_crq->cmd);
3270         }
3271 }
3272
3273 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
3274 {
3275         struct ibmvnic_adapter *adapter = instance;
3276
3277         tasklet_schedule(&adapter->tasklet);
3278         return IRQ_HANDLED;
3279 }
3280
3281 static void ibmvnic_tasklet(void *data)
3282 {
3283         struct ibmvnic_adapter *adapter = data;
3284         struct ibmvnic_crq_queue *queue = &adapter->crq;
3285         union ibmvnic_crq *crq;
3286         unsigned long flags;
3287         bool done = false;
3288
3289         spin_lock_irqsave(&queue->lock, flags);
3290         while (!done) {
3291                 /* Pull all the valid messages off the CRQ */
3292                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
3293                         ibmvnic_handle_crq(crq, adapter);
3294                         crq->generic.first = 0;
3295                 }
3296
3297                 /* remain in tasklet until all
3298                  * capabilities responses are received
3299                  */
3300                 if (!adapter->wait_capability)
3301                         done = true;
3302         }
3303         /* if capabilities CRQ's were sent in this tasklet, the following
3304          * tasklet must wait until all responses are received
3305          */
3306         if (atomic_read(&adapter->running_cap_crqs) != 0)
3307                 adapter->wait_capability = true;
3308         spin_unlock_irqrestore(&queue->lock, flags);
3309 }
3310
3311 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
3312 {
3313         struct vio_dev *vdev = adapter->vdev;
3314         int rc;
3315
3316         do {
3317                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
3318         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
3319
3320         if (rc)
3321                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
3322
3323         return rc;
3324 }
3325
3326 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
3327 {
3328         struct ibmvnic_crq_queue *crq = &adapter->crq;
3329         struct device *dev = &adapter->vdev->dev;
3330         struct vio_dev *vdev = adapter->vdev;
3331         int rc;
3332
3333         /* Close the CRQ */
3334         do {
3335                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3336         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3337
3338         /* Clean out the queue */
3339         memset(crq->msgs, 0, PAGE_SIZE);
3340         crq->cur = 0;
3341
3342         /* And re-open it again */
3343         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3344                                 crq->msg_token, PAGE_SIZE);
3345
3346         if (rc == H_CLOSED)
3347                 /* Adapter is good, but other end is not ready */
3348                 dev_warn(dev, "Partner adapter not ready\n");
3349         else if (rc != 0)
3350                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
3351
3352         return rc;
3353 }
3354
3355 static void release_crq_queue(struct ibmvnic_adapter *adapter)
3356 {
3357         struct ibmvnic_crq_queue *crq = &adapter->crq;
3358         struct vio_dev *vdev = adapter->vdev;
3359         long rc;
3360
3361         if (!crq->msgs)
3362                 return;
3363
3364         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
3365         free_irq(vdev->irq, adapter);
3366         tasklet_kill(&adapter->tasklet);
3367         do {
3368                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3369         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3370
3371         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
3372                          DMA_BIDIRECTIONAL);
3373         free_page((unsigned long)crq->msgs);
3374         crq->msgs = NULL;
3375 }
3376
3377 static int init_crq_queue(struct ibmvnic_adapter *adapter)
3378 {
3379         struct ibmvnic_crq_queue *crq = &adapter->crq;
3380         struct device *dev = &adapter->vdev->dev;
3381         struct vio_dev *vdev = adapter->vdev;
3382         int rc, retrc = -ENOMEM;
3383
3384         if (crq->msgs)
3385                 return 0;
3386
3387         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
3388         /* Should we allocate more than one page? */
3389
3390         if (!crq->msgs)
3391                 return -ENOMEM;
3392
3393         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
3394         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
3395                                         DMA_BIDIRECTIONAL);
3396         if (dma_mapping_error(dev, crq->msg_token))
3397                 goto map_failed;
3398
3399         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3400                                 crq->msg_token, PAGE_SIZE);
3401
3402         if (rc == H_RESOURCE)
3403                 /* maybe kexecing and resource is busy. try a reset */
3404                 rc = ibmvnic_reset_crq(adapter);
3405         retrc = rc;
3406
3407         if (rc == H_CLOSED) {
3408                 dev_warn(dev, "Partner adapter not ready\n");
3409         } else if (rc) {
3410                 dev_warn(dev, "Error %d opening adapter\n", rc);
3411                 goto reg_crq_failed;
3412         }
3413
3414         retrc = 0;
3415
3416         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
3417                      (unsigned long)adapter);
3418
3419         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
3420         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
3421                          adapter);
3422         if (rc) {
3423                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
3424                         vdev->irq, rc);
3425                 goto req_irq_failed;
3426         }
3427
3428         rc = vio_enable_interrupts(vdev);
3429         if (rc) {
3430                 dev_err(dev, "Error %d enabling interrupts\n", rc);
3431                 goto req_irq_failed;
3432         }
3433
3434         crq->cur = 0;
3435         spin_lock_init(&crq->lock);
3436
3437         return retrc;
3438
3439 req_irq_failed:
3440         tasklet_kill(&adapter->tasklet);
3441         do {
3442                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3443         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3444 reg_crq_failed:
3445         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
3446 map_failed:
3447         free_page((unsigned long)crq->msgs);
3448         crq->msgs = NULL;
3449         return retrc;
3450 }
3451
3452 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
3453 {
3454         struct device *dev = &adapter->vdev->dev;
3455         unsigned long timeout = msecs_to_jiffies(30000);
3456         int rc;
3457
3458         rc = init_crq_queue(adapter);
3459         if (rc) {
3460                 dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
3461                 return rc;
3462         }
3463
3464         init_completion(&adapter->init_done);
3465         ibmvnic_send_crq_init(adapter);
3466         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3467                 dev_err(dev, "Initialization sequence timed out\n");
3468                 release_crq_queue(adapter);
3469                 return -1;
3470         }
3471
3472         rc = init_sub_crqs(adapter);
3473         if (rc) {
3474                 dev_err(dev, "Initialization of sub crqs failed\n");
3475                 release_crq_queue(adapter);
3476         }
3477
3478         return rc;
3479 }
3480
3481 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
3482 {
3483         struct ibmvnic_adapter *adapter;
3484         struct net_device *netdev;
3485         unsigned char *mac_addr_p;
3486         int rc;
3487
3488         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
3489                 dev->unit_address);
3490
3491         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
3492                                                         VETH_MAC_ADDR, NULL);
3493         if (!mac_addr_p) {
3494                 dev_err(&dev->dev,
3495                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
3496                         __FILE__, __LINE__);
3497                 return 0;
3498         }
3499
3500         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
3501                                    IBMVNIC_MAX_TX_QUEUES);
3502         if (!netdev)
3503                 return -ENOMEM;
3504
3505         adapter = netdev_priv(netdev);
3506         adapter->state = VNIC_PROBING;
3507         dev_set_drvdata(&dev->dev, netdev);
3508         adapter->vdev = dev;
3509         adapter->netdev = netdev;
3510
3511         ether_addr_copy(adapter->mac_addr, mac_addr_p);
3512         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
3513         netdev->irq = dev->irq;
3514         netdev->netdev_ops = &ibmvnic_netdev_ops;
3515         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
3516         SET_NETDEV_DEV(netdev, &dev->dev);
3517
3518         spin_lock_init(&adapter->stats_lock);
3519
3520         INIT_LIST_HEAD(&adapter->errors);
3521         spin_lock_init(&adapter->error_list_lock);
3522
3523         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
3524         INIT_LIST_HEAD(&adapter->rwi_list);
3525         mutex_init(&adapter->reset_lock);
3526         mutex_init(&adapter->rwi_lock);
3527         adapter->resetting = false;
3528
3529         rc = ibmvnic_init(adapter);
3530         if (rc) {
3531                 free_netdev(netdev);
3532                 return rc;
3533         }
3534
3535         netdev->mtu = adapter->req_mtu - ETH_HLEN;
3536
3537         rc = register_netdev(netdev);
3538         if (rc) {
3539                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
3540                 free_netdev(netdev);
3541                 return rc;
3542         }
3543         dev_info(&dev->dev, "ibmvnic registered\n");
3544
3545         adapter->state = VNIC_PROBED;
3546         return 0;
3547 }
3548
3549 static int ibmvnic_remove(struct vio_dev *dev)
3550 {
3551         struct net_device *netdev = dev_get_drvdata(&dev->dev);
3552         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3553
3554         adapter->state = VNIC_REMOVING;
3555         unregister_netdev(netdev);
3556         mutex_lock(&adapter->reset_lock);
3557
3558         release_resources(adapter);
3559         release_sub_crqs(adapter);
3560         release_crq_queue(adapter);
3561
3562         adapter->state = VNIC_REMOVED;
3563
3564         mutex_unlock(&adapter->reset_lock);
3565         free_netdev(netdev);
3566         dev_set_drvdata(&dev->dev, NULL);
3567
3568         return 0;
3569 }
3570
3571 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
3572 {
3573         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
3574         struct ibmvnic_adapter *adapter;
3575         struct iommu_table *tbl;
3576         unsigned long ret = 0;
3577         int i;
3578
3579         tbl = get_iommu_table_base(&vdev->dev);
3580
3581         /* netdev inits at probe time along with the structures we need below*/
3582         if (!netdev)
3583                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
3584
3585         adapter = netdev_priv(netdev);
3586
3587         ret += PAGE_SIZE; /* the crq message queue */
3588         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
3589
3590         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
3591                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
3592
3593         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
3594              i++)
3595                 ret += adapter->rx_pool[i].size *
3596                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
3597
3598         return ret;
3599 }
3600
3601 static int ibmvnic_resume(struct device *dev)
3602 {
3603         struct net_device *netdev = dev_get_drvdata(dev);
3604         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3605         int i;
3606
3607         /* kick the interrupt handlers just in case we lost an interrupt */
3608         for (i = 0; i < adapter->req_rx_queues; i++)
3609                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
3610                                      adapter->rx_scrq[i]);
3611
3612         return 0;
3613 }
3614
3615 static struct vio_device_id ibmvnic_device_table[] = {
3616         {"network", "IBM,vnic"},
3617         {"", "" }
3618 };
3619 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
3620
3621 static const struct dev_pm_ops ibmvnic_pm_ops = {
3622         .resume = ibmvnic_resume
3623 };
3624
3625 static struct vio_driver ibmvnic_driver = {
3626         .id_table       = ibmvnic_device_table,
3627         .probe          = ibmvnic_probe,
3628         .remove         = ibmvnic_remove,
3629         .get_desired_dma = ibmvnic_get_desired_dma,
3630         .name           = ibmvnic_driver_name,
3631         .pm             = &ibmvnic_pm_ops,
3632 };
3633
3634 /* module functions */
3635 static int __init ibmvnic_module_init(void)
3636 {
3637         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
3638                 IBMVNIC_DRIVER_VERSION);
3639
3640         return vio_register_driver(&ibmvnic_driver);
3641 }
3642
3643 static void __exit ibmvnic_module_exit(void)
3644 {
3645         vio_unregister_driver(&ibmvnic_driver);
3646 }
3647
3648 module_init(ibmvnic_module_init);
3649 module_exit(ibmvnic_module_exit);