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/if_arp.h>
  63 #include <linux/in.h>
  64 #include <linux/ip.h>
  65 #include <linux/ipv6.h>
  66 #include <linux/irq.h>
  67 #include <linux/kthread.h>
  68 #include <linux/seq_file.h>
  69 #include <linux/interrupt.h>
  70 #include <net/net_namespace.h>
  71 #include <asm/hvcall.h>
  72 #include <linux/atomic.h>
  73 #include <asm/vio.h>
  74 #include <asm/iommu.h>
  75 #include <linux/uaccess.h>
  76 #include <asm/firmware.h>
  77 #include <linux/workqueue.h>
  78 #include <linux/if_vlan.h>
  79 #include <linux/utsname.h>
  80
  81 #include "ibmvnic.h"
  82
  83 static const char ibmvnic_driver_name[] = "ibmvnic";
  84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  85
  86 MODULE_AUTHOR("Santiago Leon");
  87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  88 MODULE_LICENSE("GPL");
  89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  90
  91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  92 static int ibmvnic_remove(struct vio_dev *);
  93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  99                        union sub_crq *sub_crq);
 100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
 101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 102 static int enable_scrq_irq(struct ibmvnic_adapter *,
 103                            struct ibmvnic_sub_crq_queue *);
 104 static int disable_scrq_irq(struct ibmvnic_adapter *,
 105                             struct ibmvnic_sub_crq_queue *);
 106 static int pending_scrq(struct ibmvnic_adapter *,
 107                         struct ibmvnic_sub_crq_queue *);
 108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 109                                         struct ibmvnic_sub_crq_queue *);
 110 static int ibmvnic_poll(struct napi_struct *napi, int data);
 111 static void send_map_query(struct ibmvnic_adapter *adapter);
 112 static void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 113 static void send_request_unmap(struct ibmvnic_adapter *, u8);
 114 static int send_login(struct ibmvnic_adapter *adapter);
 115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 116 static int init_sub_crqs(struct ibmvnic_adapter *);
 117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 118 static int ibmvnic_init(struct ibmvnic_adapter *);
 119 static void release_crq_queue(struct ibmvnic_adapter *);
 120 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
 121 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 122
 123 struct ibmvnic_stat {
 124         char name[ETH_GSTRING_LEN];
 125         int offset;
 126 };
 127
 128 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 129                              offsetof(struct ibmvnic_statistics, stat))
 130 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 131
 132 static const struct ibmvnic_stat ibmvnic_stats[] = {
 133         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 134         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 135         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 136         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 137         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 138         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 139         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 140         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 141         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 142         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 143         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 144         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 145         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 146         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 147         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 148         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 149         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 150         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 151         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 152         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 153         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 154         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 155 };
 156
 157 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 158                           unsigned long length, unsigned long *number,
 159                           unsigned long *irq)
 160 {
 161         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 162         long rc;
 163
 164         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 165         *number = retbuf[0];
 166         *irq = retbuf[1];
 167
 168         return rc;
 169 }
 170
 171 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 172                                 struct ibmvnic_long_term_buff *ltb, int size)
 173 {
 174         struct device *dev = &adapter->vdev->dev;
 175
 176         ltb->size = size;
 177         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 178                                        GFP_KERNEL);
 179
 180         if (!ltb->buff) {
 181                 dev_err(dev, "Couldn't alloc long term buffer\n");
 182                 return -ENOMEM;
 183         }
 184         ltb->map_id = adapter->map_id;
 185         adapter->map_id++;
 186
 187         init_completion(&adapter->fw_done);
 188         send_request_map(adapter, ltb->addr,
 189                          ltb->size, ltb->map_id);
 190         wait_for_completion(&adapter->fw_done);
 191
 192         if (adapter->fw_done_rc) {
 193                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 194                         adapter->fw_done_rc);
 195                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 196                 return -1;
 197         }
 198         return 0;
 199 }
 200
 201 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 202                                 struct ibmvnic_long_term_buff *ltb)
 203 {
 204         struct device *dev = &adapter->vdev->dev;
 205
 206         if (!ltb->buff)
 207                 return;
 208
 209         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 210             adapter->reset_reason != VNIC_RESET_MOBILITY)
 211                 send_request_unmap(adapter, ltb->map_id);
 212         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 213 }
 214
 215 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 216                                 struct ibmvnic_long_term_buff *ltb)
 217 {
 218         memset(ltb->buff, 0, ltb->size);
 219
 220         init_completion(&adapter->fw_done);
 221         send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 222         wait_for_completion(&adapter->fw_done);
 223
 224         if (adapter->fw_done_rc) {
 225                 dev_info(&adapter->vdev->dev,
 226                          "Reset failed, attempting to free and reallocate buffer\n");
 227                 free_long_term_buff(adapter, ltb);
 228                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 229         }
 230         return 0;
 231 }
 232
 233 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 234 {
 235         int i;
 236
 237         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 238              i++)
 239                 adapter->rx_pool[i].active = 0;
 240 }
 241
 242 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 243                               struct ibmvnic_rx_pool *pool)
 244 {
 245         int count = pool->size - atomic_read(&pool->available);
 246         struct device *dev = &adapter->vdev->dev;
 247         int buffers_added = 0;
 248         unsigned long lpar_rc;
 249         union sub_crq sub_crq;
 250         struct sk_buff *skb;
 251         unsigned int offset;
 252         dma_addr_t dma_addr;
 253         unsigned char *dst;
 254         u64 *handle_array;
 255         int shift = 0;
 256         int index;
 257         int i;
 258
 259         if (!pool->active)
 260                 return;
 261
 262         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 263                                       be32_to_cpu(adapter->login_rsp_buf->
 264                                       off_rxadd_subcrqs));
 265
 266         for (i = 0; i < count; ++i) {
 267                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 268                 if (!skb) {
 269                         dev_err(dev, "Couldn't replenish rx buff\n");
 270                         adapter->replenish_no_mem++;
 271                         break;
 272                 }
 273
 274                 index = pool->free_map[pool->next_free];
 275
 276                 if (pool->rx_buff[index].skb)
 277                         dev_err(dev, "Inconsistent free_map!\n");
 278
 279                 /* Copy the skb to the long term mapped DMA buffer */
 280                 offset = index * pool->buff_size;
 281                 dst = pool->long_term_buff.buff + offset;
 282                 memset(dst, 0, pool->buff_size);
 283                 dma_addr = pool->long_term_buff.addr + offset;
 284                 pool->rx_buff[index].data = dst;
 285
 286                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 287                 pool->rx_buff[index].dma = dma_addr;
 288                 pool->rx_buff[index].skb = skb;
 289                 pool->rx_buff[index].pool_index = pool->index;
 290                 pool->rx_buff[index].size = pool->buff_size;
 291
 292                 memset(&sub_crq, 0, sizeof(sub_crq));
 293                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 294                 sub_crq.rx_add.correlator =
 295                     cpu_to_be64((u64)&pool->rx_buff[index]);
 296                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 297                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 298
 299                 /* The length field of the sCRQ is defined to be 24 bits so the
 300                  * buffer size needs to be left shifted by a byte before it is
 301                  * converted to big endian to prevent the last byte from being
 302                  * truncated.
 303                  */
 304 #ifdef __LITTLE_ENDIAN__
 305                 shift = 8;
 306 #endif
 307                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 308
 309                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 310                                       &sub_crq);
 311                 if (lpar_rc != H_SUCCESS)
 312                         goto failure;
 313
 314                 buffers_added++;
 315                 adapter->replenish_add_buff_success++;
 316                 pool->next_free = (pool->next_free + 1) % pool->size;
 317         }
 318         atomic_add(buffers_added, &pool->available);
 319         return;
 320
 321 failure:
 322         dev_info(dev, "replenish pools failure\n");
 323         pool->free_map[pool->next_free] = index;
 324         pool->rx_buff[index].skb = NULL;
 325
 326         dev_kfree_skb_any(skb);
 327         adapter->replenish_add_buff_failure++;
 328         atomic_add(buffers_added, &pool->available);
 329
 330         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 331                 /* Disable buffer pool replenishment and report carrier off if
 332                  * queue is closed or pending failover.
 333                  * Firmware guarantees that a signal will be sent to the
 334                  * driver, triggering a reset.
 335                  */
 336                 deactivate_rx_pools(adapter);
 337                 netif_carrier_off(adapter->netdev);
 338         }
 339 }
 340
 341 static void replenish_pools(struct ibmvnic_adapter *adapter)
 342 {
 343         int i;
 344
 345         adapter->replenish_task_cycles++;
 346         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 347              i++) {
 348                 if (adapter->rx_pool[i].active)
 349                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 350         }
 351 }
 352
 353 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 354 {
 355         kfree(adapter->tx_stats_buffers);
 356         kfree(adapter->rx_stats_buffers);
 357         adapter->tx_stats_buffers = NULL;
 358         adapter->rx_stats_buffers = NULL;
 359 }
 360
 361 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 362 {
 363         adapter->tx_stats_buffers =
 364                                 kcalloc(IBMVNIC_MAX_QUEUES,
 365                                         sizeof(struct ibmvnic_tx_queue_stats),
 366                                         GFP_KERNEL);
 367         if (!adapter->tx_stats_buffers)
 368                 return -ENOMEM;
 369
 370         adapter->rx_stats_buffers =
 371                                 kcalloc(IBMVNIC_MAX_QUEUES,
 372                                         sizeof(struct ibmvnic_rx_queue_stats),
 373                                         GFP_KERNEL);
 374         if (!adapter->rx_stats_buffers)
 375                 return -ENOMEM;
 376
 377         return 0;
 378 }
 379
 380 static void release_stats_token(struct ibmvnic_adapter *adapter)
 381 {
 382         struct device *dev = &adapter->vdev->dev;
 383
 384         if (!adapter->stats_token)
 385                 return;
 386
 387         dma_unmap_single(dev, adapter->stats_token,
 388                          sizeof(struct ibmvnic_statistics),
 389                          DMA_FROM_DEVICE);
 390         adapter->stats_token = 0;
 391 }
 392
 393 static int init_stats_token(struct ibmvnic_adapter *adapter)
 394 {
 395         struct device *dev = &adapter->vdev->dev;
 396         dma_addr_t stok;
 397
 398         stok = dma_map_single(dev, &adapter->stats,
 399                               sizeof(struct ibmvnic_statistics),
 400                               DMA_FROM_DEVICE);
 401         if (dma_mapping_error(dev, stok)) {
 402                 dev_err(dev, "Couldn't map stats buffer\n");
 403                 return -1;
 404         }
 405
 406         adapter->stats_token = stok;
 407         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 408         return 0;
 409 }
 410
 411 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 412 {
 413         struct ibmvnic_rx_pool *rx_pool;
 414         int rx_scrqs;
 415         int i, j, rc;
 416         u64 *size_array;
 417
 418         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 419                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 420
 421         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 422         for (i = 0; i < rx_scrqs; i++) {
 423                 rx_pool = &adapter->rx_pool[i];
 424
 425                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 426
 427                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 428                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 429                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 430                         alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 431                                              rx_pool->size *
 432                                              rx_pool->buff_size);
 433                 } else {
 434                         rc = reset_long_term_buff(adapter,
 435                                                   &rx_pool->long_term_buff);
 436                 }
 437
 438                 if (rc)
 439                         return rc;
 440
 441                 for (j = 0; j < rx_pool->size; j++)
 442                         rx_pool->free_map[j] = j;
 443
 444                 memset(rx_pool->rx_buff, 0,
 445                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 446
 447                 atomic_set(&rx_pool->available, 0);
 448                 rx_pool->next_alloc = 0;
 449                 rx_pool->next_free = 0;
 450                 rx_pool->active = 1;
 451         }
 452
 453         return 0;
 454 }
 455
 456 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 457 {
 458         struct ibmvnic_rx_pool *rx_pool;
 459         int i, j;
 460
 461         if (!adapter->rx_pool)
 462                 return;
 463
 464         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 465                 rx_pool = &adapter->rx_pool[i];
 466
 467                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 468
 469                 kfree(rx_pool->free_map);
 470                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 471
 472                 if (!rx_pool->rx_buff)
 473                         continue;
 474
 475                 for (j = 0; j < rx_pool->size; j++) {
 476                         if (rx_pool->rx_buff[j].skb) {
 477                                 dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
 478                                 rx_pool->rx_buff[i].skb = NULL;
 479                         }
 480                 }
 481
 482                 kfree(rx_pool->rx_buff);
 483         }
 484
 485         kfree(adapter->rx_pool);
 486         adapter->rx_pool = NULL;
 487         adapter->num_active_rx_pools = 0;
 488 }
 489
 490 static int init_rx_pools(struct net_device *netdev)
 491 {
 492         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 493         struct device *dev = &adapter->vdev->dev;
 494         struct ibmvnic_rx_pool *rx_pool;
 495         int rxadd_subcrqs;
 496         u64 *size_array;
 497         int i, j;
 498
 499         rxadd_subcrqs =
 500                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 501         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 502                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 503
 504         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 505                                    sizeof(struct ibmvnic_rx_pool),
 506                                    GFP_KERNEL);
 507         if (!adapter->rx_pool) {
 508                 dev_err(dev, "Failed to allocate rx pools\n");
 509                 return -1;
 510         }
 511
 512         adapter->num_active_rx_pools = rxadd_subcrqs;
 513
 514         for (i = 0; i < rxadd_subcrqs; i++) {
 515                 rx_pool = &adapter->rx_pool[i];
 516
 517                 netdev_dbg(adapter->netdev,
 518                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 519                            i, adapter->req_rx_add_entries_per_subcrq,
 520                            be64_to_cpu(size_array[i]));
 521
 522                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 523                 rx_pool->index = i;
 524                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 525                 rx_pool->active = 1;
 526
 527                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 528                                             GFP_KERNEL);
 529                 if (!rx_pool->free_map) {
 530                         release_rx_pools(adapter);
 531                         return -1;
 532                 }
 533
 534                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 535                                            sizeof(struct ibmvnic_rx_buff),
 536                                            GFP_KERNEL);
 537                 if (!rx_pool->rx_buff) {
 538                         dev_err(dev, "Couldn't alloc rx buffers\n");
 539                         release_rx_pools(adapter);
 540                         return -1;
 541                 }
 542
 543                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 544                                          rx_pool->size * rx_pool->buff_size)) {
 545                         release_rx_pools(adapter);
 546                         return -1;
 547                 }
 548
 549                 for (j = 0; j < rx_pool->size; ++j)
 550                         rx_pool->free_map[j] = j;
 551
 552                 atomic_set(&rx_pool->available, 0);
 553                 rx_pool->next_alloc = 0;
 554                 rx_pool->next_free = 0;
 555         }
 556
 557         return 0;
 558 }
 559
 560 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 561                              struct ibmvnic_tx_pool *tx_pool)
 562 {
 563         int rc, i;
 564
 565         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 566         if (rc)
 567                 return rc;
 568
 569         memset(tx_pool->tx_buff, 0,
 570                tx_pool->num_buffers *
 571                sizeof(struct ibmvnic_tx_buff));
 572
 573         for (i = 0; i < tx_pool->num_buffers; i++)
 574                 tx_pool->free_map[i] = i;
 575
 576         tx_pool->consumer_index = 0;
 577         tx_pool->producer_index = 0;
 578
 579         return 0;
 580 }
 581
 582 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 583 {
 584         int tx_scrqs;
 585         int i, rc;
 586
 587         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 588         for (i = 0; i < tx_scrqs; i++) {
 589                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 590                 if (rc)
 591                         return rc;
 592                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 593                 if (rc)
 594                         return rc;
 595         }
 596
 597         return 0;
 598 }
 599
 600 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 601 {
 602         if (!adapter->vpd)
 603                 return;
 604
 605         kfree(adapter->vpd->buff);
 606         kfree(adapter->vpd);
 607
 608         adapter->vpd = NULL;
 609 }
 610
 611 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 612                                 struct ibmvnic_tx_pool *tx_pool)
 613 {
 614         kfree(tx_pool->tx_buff);
 615         kfree(tx_pool->free_map);
 616         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 617 }
 618
 619 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 620 {
 621         int i;
 622
 623         if (!adapter->tx_pool)
 624                 return;
 625
 626         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 627                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 628                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 629         }
 630
 631         kfree(adapter->tx_pool);
 632         adapter->tx_pool = NULL;
 633         kfree(adapter->tso_pool);
 634         adapter->tso_pool = NULL;
 635         adapter->num_active_tx_pools = 0;
 636 }
 637
 638 static int init_one_tx_pool(struct net_device *netdev,
 639                             struct ibmvnic_tx_pool *tx_pool,
 640                             int num_entries, int buf_size)
 641 {
 642         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 643         int i;
 644
 645         tx_pool->tx_buff = kcalloc(num_entries,
 646                                    sizeof(struct ibmvnic_tx_buff),
 647                                    GFP_KERNEL);
 648         if (!tx_pool->tx_buff)
 649                 return -1;
 650
 651         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 652                                  num_entries * buf_size))
 653                 return -1;
 654
 655         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 656         if (!tx_pool->free_map)
 657                 return -1;
 658
 659         for (i = 0; i < num_entries; i++)
 660                 tx_pool->free_map[i] = i;
 661
 662         tx_pool->consumer_index = 0;
 663         tx_pool->producer_index = 0;
 664         tx_pool->num_buffers = num_entries;
 665         tx_pool->buf_size = buf_size;
 666
 667         return 0;
 668 }
 669
 670 static int init_tx_pools(struct net_device *netdev)
 671 {
 672         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 673         int tx_subcrqs;
 674         int i, rc;
 675
 676         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 677         adapter->tx_pool = kcalloc(tx_subcrqs,
 678                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 679         if (!adapter->tx_pool)
 680                 return -1;
 681
 682         adapter->tso_pool = kcalloc(tx_subcrqs,
 683                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 684         if (!adapter->tso_pool)
 685                 return -1;
 686
 687         adapter->num_active_tx_pools = tx_subcrqs;
 688
 689         for (i = 0; i < tx_subcrqs; i++) {
 690                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 691                                       adapter->req_tx_entries_per_subcrq,
 692                                       adapter->req_mtu + VLAN_HLEN);
 693                 if (rc) {
 694                         release_tx_pools(adapter);
 695                         return rc;
 696                 }
 697
 698                 init_one_tx_pool(netdev, &adapter->tso_pool[i],
 699                                  IBMVNIC_TSO_BUFS,
 700                                  IBMVNIC_TSO_BUF_SZ);
 701                 if (rc) {
 702                         release_tx_pools(adapter);
 703                         return rc;
 704                 }
 705         }
 706
 707         return 0;
 708 }
 709
 710 static void release_error_buffers(struct ibmvnic_adapter *adapter)
 711 {
 712         struct device *dev = &adapter->vdev->dev;
 713         struct ibmvnic_error_buff *error_buff, *tmp;
 714         unsigned long flags;
 715
 716         spin_lock_irqsave(&adapter->error_list_lock, flags);
 717         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
 718                 list_del(&error_buff->list);
 719                 dma_unmap_single(dev, error_buff->dma, error_buff->len,
 720                                  DMA_FROM_DEVICE);
 721                 kfree(error_buff->buff);
 722                 kfree(error_buff);
 723         }
 724         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
 725 }
 726
 727 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 728 {
 729         int i;
 730
 731         if (adapter->napi_enabled)
 732                 return;
 733
 734         for (i = 0; i < adapter->req_rx_queues; i++)
 735                 napi_enable(&adapter->napi[i]);
 736
 737         adapter->napi_enabled = true;
 738 }
 739
 740 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 741 {
 742         int i;
 743
 744         if (!adapter->napi_enabled)
 745                 return;
 746
 747         for (i = 0; i < adapter->req_rx_queues; i++) {
 748                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 749                 napi_disable(&adapter->napi[i]);
 750         }
 751
 752         adapter->napi_enabled = false;
 753 }
 754
 755 static int init_napi(struct ibmvnic_adapter *adapter)
 756 {
 757         int i;
 758
 759         adapter->napi = kcalloc(adapter->req_rx_queues,
 760                                 sizeof(struct napi_struct), GFP_KERNEL);
 761         if (!adapter->napi)
 762                 return -ENOMEM;
 763
 764         for (i = 0; i < adapter->req_rx_queues; i++) {
 765                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 766                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 767                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 768         }
 769
 770         adapter->num_active_rx_napi = adapter->req_rx_queues;
 771         return 0;
 772 }
 773
 774 static void release_napi(struct ibmvnic_adapter *adapter)
 775 {
 776         int i;
 777
 778         if (!adapter->napi)
 779                 return;
 780
 781         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 782                 if (&adapter->napi[i]) {
 783                         netdev_dbg(adapter->netdev,
 784                                    "Releasing napi[%d]\n", i);
 785                         netif_napi_del(&adapter->napi[i]);
 786                 }
 787         }
 788
 789         kfree(adapter->napi);
 790         adapter->napi = NULL;
 791         adapter->num_active_rx_napi = 0;
 792 }
 793
 794 static int ibmvnic_login(struct net_device *netdev)
 795 {
 796         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 797         unsigned long timeout = msecs_to_jiffies(30000);
 798         int retry_count = 0;
 799         int rc;
 800
 801         do {
 802                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 803                         netdev_warn(netdev, "Login attempts exceeded\n");
 804                         return -1;
 805                 }
 806
 807                 adapter->init_done_rc = 0;
 808                 reinit_completion(&adapter->init_done);
 809                 rc = send_login(adapter);
 810                 if (rc) {
 811                         netdev_warn(netdev, "Unable to login\n");
 812                         return rc;
 813                 }
 814
 815                 if (!wait_for_completion_timeout(&adapter->init_done,
 816                                                  timeout)) {
 817                         netdev_warn(netdev, "Login timed out\n");
 818                         return -1;
 819                 }
 820
 821                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 822                         retry_count++;
 823                         release_sub_crqs(adapter, 1);
 824
 825                         adapter->init_done_rc = 0;
 826                         reinit_completion(&adapter->init_done);
 827                         send_cap_queries(adapter);
 828                         if (!wait_for_completion_timeout(&adapter->init_done,
 829                                                          timeout)) {
 830                                 netdev_warn(netdev,
 831                                             "Capabilities query timed out\n");
 832                                 return -1;
 833                         }
 834
 835                         rc = init_sub_crqs(adapter);
 836                         if (rc) {
 837                                 netdev_warn(netdev,
 838                                             "SCRQ initialization failed\n");
 839                                 return -1;
 840                         }
 841
 842                         rc = init_sub_crq_irqs(adapter);
 843                         if (rc) {
 844                                 netdev_warn(netdev,
 845                                             "SCRQ irq initialization failed\n");
 846                                 return -1;
 847                         }
 848                 } else if (adapter->init_done_rc) {
 849                         netdev_warn(netdev, "Adapter login failed\n");
 850                         return -1;
 851                 }
 852         } while (adapter->init_done_rc == PARTIALSUCCESS);
 853
 854         /* handle pending MAC address changes after successful login */
 855         if (adapter->mac_change_pending) {
 856                 __ibmvnic_set_mac(netdev, &adapter->desired.mac);
 857                 adapter->mac_change_pending = false;
 858         }
 859
 860         return 0;
 861 }
 862
 863 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 864 {
 865         kfree(adapter->login_buf);
 866         adapter->login_buf = NULL;
 867 }
 868
 869 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 870 {
 871         kfree(adapter->login_rsp_buf);
 872         adapter->login_rsp_buf = NULL;
 873 }
 874
 875 static void release_resources(struct ibmvnic_adapter *adapter)
 876 {
 877         release_vpd_data(adapter);
 878
 879         release_tx_pools(adapter);
 880         release_rx_pools(adapter);
 881
 882         release_error_buffers(adapter);
 883         release_napi(adapter);
 884         release_login_rsp_buffer(adapter);
 885 }
 886
 887 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 888 {
 889         struct net_device *netdev = adapter->netdev;
 890         unsigned long timeout = msecs_to_jiffies(30000);
 891         union ibmvnic_crq crq;
 892         bool resend;
 893         int rc;
 894
 895         netdev_dbg(netdev, "setting link state %d\n", link_state);
 896
 897         memset(&crq, 0, sizeof(crq));
 898         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 899         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 900         crq.logical_link_state.link_state = link_state;
 901
 902         do {
 903                 resend = false;
 904
 905                 reinit_completion(&adapter->init_done);
 906                 rc = ibmvnic_send_crq(adapter, &crq);
 907                 if (rc) {
 908                         netdev_err(netdev, "Failed to set link state\n");
 909                         return rc;
 910                 }
 911
 912                 if (!wait_for_completion_timeout(&adapter->init_done,
 913                                                  timeout)) {
 914                         netdev_err(netdev, "timeout setting link state\n");
 915                         return -1;
 916                 }
 917
 918                 if (adapter->init_done_rc == 1) {
 919                         /* Partuial success, delay and re-send */
 920                         mdelay(1000);
 921                         resend = true;
 922                 }
 923         } while (resend);
 924
 925         return 0;
 926 }
 927
 928 static int set_real_num_queues(struct net_device *netdev)
 929 {
 930         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 931         int rc;
 932
 933         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 934                    adapter->req_tx_queues, adapter->req_rx_queues);
 935
 936         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 937         if (rc) {
 938                 netdev_err(netdev, "failed to set the number of tx queues\n");
 939                 return rc;
 940         }
 941
 942         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 943         if (rc)
 944                 netdev_err(netdev, "failed to set the number of rx queues\n");
 945
 946         return rc;
 947 }
 948
 949 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 950 {
 951         struct device *dev = &adapter->vdev->dev;
 952         union ibmvnic_crq crq;
 953         int len = 0;
 954
 955         if (adapter->vpd->buff)
 956                 len = adapter->vpd->len;
 957
 958         init_completion(&adapter->fw_done);
 959         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 960         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 961         ibmvnic_send_crq(adapter, &crq);
 962         wait_for_completion(&adapter->fw_done);
 963
 964         if (!adapter->vpd->len)
 965                 return -ENODATA;
 966
 967         if (!adapter->vpd->buff)
 968                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 969         else if (adapter->vpd->len != len)
 970                 adapter->vpd->buff =
 971                         krealloc(adapter->vpd->buff,
 972                                  adapter->vpd->len, GFP_KERNEL);
 973
 974         if (!adapter->vpd->buff) {
 975                 dev_err(dev, "Could allocate VPD buffer\n");
 976                 return -ENOMEM;
 977         }
 978
 979         adapter->vpd->dma_addr =
 980                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 981                                DMA_FROM_DEVICE);
 982         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 983                 dev_err(dev, "Could not map VPD buffer\n");
 984                 kfree(adapter->vpd->buff);
 985                 adapter->vpd->buff = NULL;
 986                 return -ENOMEM;
 987         }
 988
 989         reinit_completion(&adapter->fw_done);
 990         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 991         crq.get_vpd.cmd = GET_VPD;
 992         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 993         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
 994         ibmvnic_send_crq(adapter, &crq);
 995         wait_for_completion(&adapter->fw_done);
 996
 997         return 0;
 998 }
 999
1000 static int init_resources(struct ibmvnic_adapter *adapter)
1001 {
1002         struct net_device *netdev = adapter->netdev;
1003         int rc;
1004
1005         rc = set_real_num_queues(netdev);
1006         if (rc)
1007                 return rc;
1008
1009         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1010         if (!adapter->vpd)
1011                 return -ENOMEM;
1012
1013         /* Vital Product Data (VPD) */
1014         rc = ibmvnic_get_vpd(adapter);
1015         if (rc) {
1016                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1017                 return rc;
1018         }
1019
1020         adapter->map_id = 1;
1021
1022         rc = init_napi(adapter);
1023         if (rc)
1024                 return rc;
1025
1026         send_map_query(adapter);
1027
1028         rc = init_rx_pools(netdev);
1029         if (rc)
1030                 return rc;
1031
1032         rc = init_tx_pools(netdev);
1033         return rc;
1034 }
1035
1036 static int __ibmvnic_open(struct net_device *netdev)
1037 {
1038         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1039         enum vnic_state prev_state = adapter->state;
1040         int i, rc;
1041
1042         adapter->state = VNIC_OPENING;
1043         replenish_pools(adapter);
1044         ibmvnic_napi_enable(adapter);
1045
1046         /* We're ready to receive frames, enable the sub-crq interrupts and
1047          * set the logical link state to up
1048          */
1049         for (i = 0; i < adapter->req_rx_queues; i++) {
1050                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1051                 if (prev_state == VNIC_CLOSED)
1052                         enable_irq(adapter->rx_scrq[i]->irq);
1053                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1054         }
1055
1056         for (i = 0; i < adapter->req_tx_queues; i++) {
1057                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1058                 if (prev_state == VNIC_CLOSED)
1059                         enable_irq(adapter->tx_scrq[i]->irq);
1060                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1061         }
1062
1063         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1064         if (rc) {
1065                 for (i = 0; i < adapter->req_rx_queues; i++)
1066                         napi_disable(&adapter->napi[i]);
1067                 release_resources(adapter);
1068                 return rc;
1069         }
1070
1071         netif_tx_start_all_queues(netdev);
1072
1073         if (prev_state == VNIC_CLOSED) {
1074                 for (i = 0; i < adapter->req_rx_queues; i++)
1075                         napi_schedule(&adapter->napi[i]);
1076         }
1077
1078         adapter->state = VNIC_OPEN;
1079         return rc;
1080 }
1081
1082 static int ibmvnic_open(struct net_device *netdev)
1083 {
1084         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1085         int rc;
1086
1087         /* If device failover is pending, just set device state and return.
1088          * Device operation will be handled by reset routine.
1089          */
1090         if (adapter->failover_pending) {
1091                 adapter->state = VNIC_OPEN;
1092                 return 0;
1093         }
1094
1095         mutex_lock(&adapter->reset_lock);
1096
1097         if (adapter->state != VNIC_CLOSED) {
1098                 rc = ibmvnic_login(netdev);
1099                 if (rc) {
1100                         mutex_unlock(&adapter->reset_lock);
1101                         return rc;
1102                 }
1103
1104                 rc = init_resources(adapter);
1105                 if (rc) {
1106                         netdev_err(netdev, "failed to initialize resources\n");
1107                         release_resources(adapter);
1108                         mutex_unlock(&adapter->reset_lock);
1109                         return rc;
1110                 }
1111         }
1112
1113         rc = __ibmvnic_open(netdev);
1114         netif_carrier_on(netdev);
1115
1116         mutex_unlock(&adapter->reset_lock);
1117
1118         return rc;
1119 }
1120
1121 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1122 {
1123         struct ibmvnic_rx_pool *rx_pool;
1124         struct ibmvnic_rx_buff *rx_buff;
1125         u64 rx_entries;
1126         int rx_scrqs;
1127         int i, j;
1128
1129         if (!adapter->rx_pool)
1130                 return;
1131
1132         rx_scrqs = adapter->num_active_rx_pools;
1133         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1134
1135         /* Free any remaining skbs in the rx buffer pools */
1136         for (i = 0; i < rx_scrqs; i++) {
1137                 rx_pool = &adapter->rx_pool[i];
1138                 if (!rx_pool || !rx_pool->rx_buff)
1139                         continue;
1140
1141                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1142                 for (j = 0; j < rx_entries; j++) {
1143                         rx_buff = &rx_pool->rx_buff[j];
1144                         if (rx_buff && rx_buff->skb) {
1145                                 dev_kfree_skb_any(rx_buff->skb);
1146                                 rx_buff->skb = NULL;
1147                         }
1148                 }
1149         }
1150 }
1151
1152 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1153                               struct ibmvnic_tx_pool *tx_pool)
1154 {
1155         struct ibmvnic_tx_buff *tx_buff;
1156         u64 tx_entries;
1157         int i;
1158
1159         if (!tx_pool || !tx_pool->tx_buff)
1160                 return;
1161
1162         tx_entries = tx_pool->num_buffers;
1163
1164         for (i = 0; i < tx_entries; i++) {
1165                 tx_buff = &tx_pool->tx_buff[i];
1166                 if (tx_buff && tx_buff->skb) {
1167                         dev_kfree_skb_any(tx_buff->skb);
1168                         tx_buff->skb = NULL;
1169                 }
1170         }
1171 }
1172
1173 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1174 {
1175         int tx_scrqs;
1176         int i;
1177
1178         if (!adapter->tx_pool || !adapter->tso_pool)
1179                 return;
1180
1181         tx_scrqs = adapter->num_active_tx_pools;
1182
1183         /* Free any remaining skbs in the tx buffer pools */
1184         for (i = 0; i < tx_scrqs; i++) {
1185                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1186                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1187                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1188         }
1189 }
1190
1191 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1192 {
1193         struct net_device *netdev = adapter->netdev;
1194         int i;
1195
1196         if (adapter->tx_scrq) {
1197                 for (i = 0; i < adapter->req_tx_queues; i++)
1198                         if (adapter->tx_scrq[i]->irq) {
1199                                 netdev_dbg(netdev,
1200                                            "Disabling tx_scrq[%d] irq\n", i);
1201                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1202                                 disable_irq(adapter->tx_scrq[i]->irq);
1203                         }
1204         }
1205
1206         if (adapter->rx_scrq) {
1207                 for (i = 0; i < adapter->req_rx_queues; i++) {
1208                         if (adapter->rx_scrq[i]->irq) {
1209                                 netdev_dbg(netdev,
1210                                            "Disabling rx_scrq[%d] irq\n", i);
1211                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1212                                 disable_irq(adapter->rx_scrq[i]->irq);
1213                         }
1214                 }
1215         }
1216 }
1217
1218 static void ibmvnic_cleanup(struct net_device *netdev)
1219 {
1220         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1221
1222         /* ensure that transmissions are stopped if called by do_reset */
1223         if (adapter->resetting)
1224                 netif_tx_disable(netdev);
1225         else
1226                 netif_tx_stop_all_queues(netdev);
1227
1228         ibmvnic_napi_disable(adapter);
1229         ibmvnic_disable_irqs(adapter);
1230
1231         clean_rx_pools(adapter);
1232         clean_tx_pools(adapter);
1233 }
1234
1235 static int __ibmvnic_close(struct net_device *netdev)
1236 {
1237         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1238         int rc = 0;
1239
1240         adapter->state = VNIC_CLOSING;
1241         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1242         if (rc)
1243                 return rc;
1244         adapter->state = VNIC_CLOSED;
1245         return 0;
1246 }
1247
1248 static int ibmvnic_close(struct net_device *netdev)
1249 {
1250         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1251         int rc;
1252
1253         /* If device failover is pending, just set device state and return.
1254          * Device operation will be handled by reset routine.
1255          */
1256         if (adapter->failover_pending) {
1257                 adapter->state = VNIC_CLOSED;
1258                 return 0;
1259         }
1260
1261         mutex_lock(&adapter->reset_lock);
1262         rc = __ibmvnic_close(netdev);
1263         ibmvnic_cleanup(netdev);
1264         mutex_unlock(&adapter->reset_lock);
1265
1266         return rc;
1267 }
1268
1269 /**
1270  * build_hdr_data - creates L2/L3/L4 header data buffer
1271  * @hdr_field - bitfield determining needed headers
1272  * @skb - socket buffer
1273  * @hdr_len - array of header lengths
1274  * @tot_len - total length of data
1275  *
1276  * Reads hdr_field to determine which headers are needed by firmware.
1277  * Builds a buffer containing these headers.  Saves individual header
1278  * lengths and total buffer length to be used to build descriptors.
1279  */
1280 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1281                           int *hdr_len, u8 *hdr_data)
1282 {
1283         int len = 0;
1284         u8 *hdr;
1285
1286         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1287                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1288         else
1289                 hdr_len[0] = sizeof(struct ethhdr);
1290
1291         if (skb->protocol == htons(ETH_P_IP)) {
1292                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1293                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1294                         hdr_len[2] = tcp_hdrlen(skb);
1295                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1296                         hdr_len[2] = sizeof(struct udphdr);
1297         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1298                 hdr_len[1] = sizeof(struct ipv6hdr);
1299                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1300                         hdr_len[2] = tcp_hdrlen(skb);
1301                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1302                         hdr_len[2] = sizeof(struct udphdr);
1303         } else if (skb->protocol == htons(ETH_P_ARP)) {
1304                 hdr_len[1] = arp_hdr_len(skb->dev);
1305                 hdr_len[2] = 0;
1306         }
1307
1308         memset(hdr_data, 0, 120);
1309         if ((hdr_field >> 6) & 1) {
1310                 hdr = skb_mac_header(skb);
1311                 memcpy(hdr_data, hdr, hdr_len[0]);
1312                 len += hdr_len[0];
1313         }
1314
1315         if ((hdr_field >> 5) & 1) {
1316                 hdr = skb_network_header(skb);
1317                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1318                 len += hdr_len[1];
1319         }
1320
1321         if ((hdr_field >> 4) & 1) {
1322                 hdr = skb_transport_header(skb);
1323                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1324                 len += hdr_len[2];
1325         }
1326         return len;
1327 }
1328
1329 /**
1330  * create_hdr_descs - create header and header extension descriptors
1331  * @hdr_field - bitfield determining needed headers
1332  * @data - buffer containing header data
1333  * @len - length of data buffer
1334  * @hdr_len - array of individual header lengths
1335  * @scrq_arr - descriptor array
1336  *
1337  * Creates header and, if needed, header extension descriptors and
1338  * places them in a descriptor array, scrq_arr
1339  */
1340
1341 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1342                             union sub_crq *scrq_arr)
1343 {
1344         union sub_crq hdr_desc;
1345         int tmp_len = len;
1346         int num_descs = 0;
1347         u8 *data, *cur;
1348         int tmp;
1349
1350         while (tmp_len > 0) {
1351                 cur = hdr_data + len - tmp_len;
1352
1353                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1354                 if (cur != hdr_data) {
1355                         data = hdr_desc.hdr_ext.data;
1356                         tmp = tmp_len > 29 ? 29 : tmp_len;
1357                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1358                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1359                         hdr_desc.hdr_ext.len = tmp;
1360                 } else {
1361                         data = hdr_desc.hdr.data;
1362                         tmp = tmp_len > 24 ? 24 : tmp_len;
1363                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1364                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1365                         hdr_desc.hdr.len = tmp;
1366                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1367                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1368                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1369                         hdr_desc.hdr.flag = hdr_field << 1;
1370                 }
1371                 memcpy(data, cur, tmp);
1372                 tmp_len -= tmp;
1373                 *scrq_arr = hdr_desc;
1374                 scrq_arr++;
1375                 num_descs++;
1376         }
1377
1378         return num_descs;
1379 }
1380
1381 /**
1382  * build_hdr_descs_arr - build a header descriptor array
1383  * @skb - socket buffer
1384  * @num_entries - number of descriptors to be sent
1385  * @subcrq - first TX descriptor
1386  * @hdr_field - bit field determining which headers will be sent
1387  *
1388  * This function will build a TX descriptor array with applicable
1389  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1390  */
1391
1392 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1393                                 int *num_entries, u8 hdr_field)
1394 {
1395         int hdr_len[3] = {0, 0, 0};
1396         int tot_len;
1397         u8 *hdr_data = txbuff->hdr_data;
1398
1399         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1400                                  txbuff->hdr_data);
1401         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1402                          txbuff->indir_arr + 1);
1403 }
1404
1405 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1406                                     struct net_device *netdev)
1407 {
1408         /* For some backing devices, mishandling of small packets
1409          * can result in a loss of connection or TX stall. Device
1410          * architects recommend that no packet should be smaller
1411          * than the minimum MTU value provided to the driver, so
1412          * pad any packets to that length
1413          */
1414         if (skb->len < netdev->min_mtu)
1415                 return skb_put_padto(skb, netdev->min_mtu);
1416
1417         return 0;
1418 }
1419
1420 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1421 {
1422         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1423         int queue_num = skb_get_queue_mapping(skb);
1424         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1425         struct device *dev = &adapter->vdev->dev;
1426         struct ibmvnic_tx_buff *tx_buff = NULL;
1427         struct ibmvnic_sub_crq_queue *tx_scrq;
1428         struct ibmvnic_tx_pool *tx_pool;
1429         unsigned int tx_send_failed = 0;
1430         unsigned int tx_map_failed = 0;
1431         unsigned int tx_dropped = 0;
1432         unsigned int tx_packets = 0;
1433         unsigned int tx_bytes = 0;
1434         dma_addr_t data_dma_addr;
1435         struct netdev_queue *txq;
1436         unsigned long lpar_rc;
1437         union sub_crq tx_crq;
1438         unsigned int offset;
1439         int num_entries = 1;
1440         unsigned char *dst;
1441         u64 *handle_array;
1442         int index = 0;
1443         u8 proto = 0;
1444         int ret = 0;
1445
1446         if (adapter->resetting) {
1447                 if (!netif_subqueue_stopped(netdev, skb))
1448                         netif_stop_subqueue(netdev, queue_num);
1449                 dev_kfree_skb_any(skb);
1450
1451                 tx_send_failed++;
1452                 tx_dropped++;
1453                 ret = NETDEV_TX_OK;
1454                 goto out;
1455         }
1456
1457         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1458                 tx_dropped++;
1459                 tx_send_failed++;
1460                 ret = NETDEV_TX_OK;
1461                 goto out;
1462         }
1463         if (skb_is_gso(skb))
1464                 tx_pool = &adapter->tso_pool[queue_num];
1465         else
1466                 tx_pool = &adapter->tx_pool[queue_num];
1467
1468         tx_scrq = adapter->tx_scrq[queue_num];
1469         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1470         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1471                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1472
1473         index = tx_pool->free_map[tx_pool->consumer_index];
1474
1475         if (index == IBMVNIC_INVALID_MAP) {
1476                 dev_kfree_skb_any(skb);
1477                 tx_send_failed++;
1478                 tx_dropped++;
1479                 ret = NETDEV_TX_OK;
1480                 goto out;
1481         }
1482
1483         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1484
1485         offset = index * tx_pool->buf_size;
1486         dst = tx_pool->long_term_buff.buff + offset;
1487         memset(dst, 0, tx_pool->buf_size);
1488         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1489
1490         if (skb_shinfo(skb)->nr_frags) {
1491                 int cur, i;
1492
1493                 /* Copy the head */
1494                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1495                 cur = skb_headlen(skb);
1496
1497                 /* Copy the frags */
1498                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1499                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1500
1501                         memcpy(dst + cur,
1502                                page_address(skb_frag_page(frag)) +
1503                                frag->page_offset, skb_frag_size(frag));
1504                         cur += skb_frag_size(frag);
1505                 }
1506         } else {
1507                 skb_copy_from_linear_data(skb, dst, skb->len);
1508         }
1509
1510         tx_pool->consumer_index =
1511             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1512
1513         tx_buff = &tx_pool->tx_buff[index];
1514         tx_buff->skb = skb;
1515         tx_buff->data_dma[0] = data_dma_addr;
1516         tx_buff->data_len[0] = skb->len;
1517         tx_buff->index = index;
1518         tx_buff->pool_index = queue_num;
1519         tx_buff->last_frag = true;
1520
1521         memset(&tx_crq, 0, sizeof(tx_crq));
1522         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1523         tx_crq.v1.type = IBMVNIC_TX_DESC;
1524         tx_crq.v1.n_crq_elem = 1;
1525         tx_crq.v1.n_sge = 1;
1526         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1527
1528         if (skb_is_gso(skb))
1529                 tx_crq.v1.correlator =
1530                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1531         else
1532                 tx_crq.v1.correlator = cpu_to_be32(index);
1533         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1534         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1535         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1536
1537         if (adapter->vlan_header_insertion) {
1538                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1539                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1540         }
1541
1542         if (skb->protocol == htons(ETH_P_IP)) {
1543                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1544                 proto = ip_hdr(skb)->protocol;
1545         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1546                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1547                 proto = ipv6_hdr(skb)->nexthdr;
1548         }
1549
1550         if (proto == IPPROTO_TCP)
1551                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1552         else if (proto == IPPROTO_UDP)
1553                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1554
1555         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1556                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1557                 hdrs += 2;
1558         }
1559         if (skb_is_gso(skb)) {
1560                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1561                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1562                 hdrs += 2;
1563         }
1564         /* determine if l2/3/4 headers are sent to firmware */
1565         if ((*hdrs >> 7) & 1) {
1566                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1567                 tx_crq.v1.n_crq_elem = num_entries;
1568                 tx_buff->num_entries = num_entries;
1569                 tx_buff->indir_arr[0] = tx_crq;
1570                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1571                                                     sizeof(tx_buff->indir_arr),
1572                                                     DMA_TO_DEVICE);
1573                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1574                         dev_kfree_skb_any(skb);
1575                         tx_buff->skb = NULL;
1576                         if (!firmware_has_feature(FW_FEATURE_CMO))
1577                                 dev_err(dev, "tx: unable to map descriptor array\n");
1578                         tx_map_failed++;
1579                         tx_dropped++;
1580                         ret = NETDEV_TX_OK;
1581                         goto tx_err_out;
1582                 }
1583                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1584                                                (u64)tx_buff->indir_dma,
1585                                                (u64)num_entries);
1586         } else {
1587                 tx_buff->num_entries = num_entries;
1588                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1589                                       &tx_crq);
1590         }
1591         if (lpar_rc != H_SUCCESS) {
1592                 dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1593                 dev_kfree_skb_any(skb);
1594                 tx_buff->skb = NULL;
1595
1596                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1597                         /* Disable TX and report carrier off if queue is closed
1598                          * or pending failover.
1599                          * Firmware guarantees that a signal will be sent to the
1600                          * driver, triggering a reset or some other action.
1601                          */
1602                         netif_tx_stop_all_queues(netdev);
1603                         netif_carrier_off(netdev);
1604                 }
1605
1606                 tx_send_failed++;
1607                 tx_dropped++;
1608                 ret = NETDEV_TX_OK;
1609                 goto tx_err_out;
1610         }
1611
1612         if (atomic_add_return(num_entries, &tx_scrq->used)
1613                                         >= adapter->req_tx_entries_per_subcrq) {
1614                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1615                 netif_stop_subqueue(netdev, queue_num);
1616         }
1617
1618         tx_packets++;
1619         tx_bytes += skb->len;
1620         txq->trans_start = jiffies;
1621         ret = NETDEV_TX_OK;
1622         goto out;
1623
1624 tx_err_out:
1625         /* roll back consumer index and map array*/
1626         if (tx_pool->consumer_index == 0)
1627                 tx_pool->consumer_index =
1628                         tx_pool->num_buffers - 1;
1629         else
1630                 tx_pool->consumer_index--;
1631         tx_pool->free_map[tx_pool->consumer_index] = index;
1632 out:
1633         netdev->stats.tx_dropped += tx_dropped;
1634         netdev->stats.tx_bytes += tx_bytes;
1635         netdev->stats.tx_packets += tx_packets;
1636         adapter->tx_send_failed += tx_send_failed;
1637         adapter->tx_map_failed += tx_map_failed;
1638         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1639         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1640         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1641
1642         return ret;
1643 }
1644
1645 static void ibmvnic_set_multi(struct net_device *netdev)
1646 {
1647         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1648         struct netdev_hw_addr *ha;
1649         union ibmvnic_crq crq;
1650
1651         memset(&crq, 0, sizeof(crq));
1652         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1653         crq.request_capability.cmd = REQUEST_CAPABILITY;
1654
1655         if (netdev->flags & IFF_PROMISC) {
1656                 if (!adapter->promisc_supported)
1657                         return;
1658         } else {
1659                 if (netdev->flags & IFF_ALLMULTI) {
1660                         /* Accept all multicast */
1661                         memset(&crq, 0, sizeof(crq));
1662                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1663                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1664                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1665                         ibmvnic_send_crq(adapter, &crq);
1666                 } else if (netdev_mc_empty(netdev)) {
1667                         /* Reject all multicast */
1668                         memset(&crq, 0, sizeof(crq));
1669                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1670                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1671                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1672                         ibmvnic_send_crq(adapter, &crq);
1673                 } else {
1674                         /* Accept one or more multicast(s) */
1675                         netdev_for_each_mc_addr(ha, netdev) {
1676                                 memset(&crq, 0, sizeof(crq));
1677                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1678                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1679                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1680                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1681                                                 ha->addr);
1682                                 ibmvnic_send_crq(adapter, &crq);
1683                         }
1684                 }
1685         }
1686 }
1687
1688 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1689 {
1690         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1691         struct sockaddr *addr = p;
1692         union ibmvnic_crq crq;
1693
1694         if (!is_valid_ether_addr(addr->sa_data))
1695                 return -EADDRNOTAVAIL;
1696
1697         memset(&crq, 0, sizeof(crq));
1698         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1699         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1700         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1701
1702         init_completion(&adapter->fw_done);
1703         ibmvnic_send_crq(adapter, &crq);
1704         wait_for_completion(&adapter->fw_done);
1705         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1706         return adapter->fw_done_rc ? -EIO : 0;
1707 }
1708
1709 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1710 {
1711         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1712         struct sockaddr *addr = p;
1713         int rc;
1714
1715         if (adapter->state == VNIC_PROBED) {
1716                 memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1717                 adapter->mac_change_pending = true;
1718                 return 0;
1719         }
1720
1721         rc = __ibmvnic_set_mac(netdev, addr);
1722
1723         return rc;
1724 }
1725
1726 /**
1727  * do_reset returns zero if we are able to keep processing reset events, or
1728  * non-zero if we hit a fatal error and must halt.
1729  */
1730 static int do_reset(struct ibmvnic_adapter *adapter,
1731                     struct ibmvnic_rwi *rwi, u32 reset_state)
1732 {
1733         u64 old_num_rx_queues, old_num_tx_queues;
1734         struct net_device *netdev = adapter->netdev;
1735         int i, rc;
1736
1737         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1738                    rwi->reset_reason);
1739
1740         netif_carrier_off(netdev);
1741         adapter->reset_reason = rwi->reset_reason;
1742
1743         old_num_rx_queues = adapter->req_rx_queues;
1744         old_num_tx_queues = adapter->req_tx_queues;
1745
1746         ibmvnic_cleanup(netdev);
1747
1748         if (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1749             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1750                 rc = __ibmvnic_close(netdev);
1751                 if (rc)
1752                         return rc;
1753         }
1754
1755         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1756             adapter->wait_for_reset) {
1757                 release_resources(adapter);
1758                 release_sub_crqs(adapter, 1);
1759                 release_crq_queue(adapter);
1760         }
1761
1762         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1763                 /* remove the closed state so when we call open it appears
1764                  * we are coming from the probed state.
1765                  */
1766                 adapter->state = VNIC_PROBED;
1767
1768                 if (adapter->wait_for_reset) {
1769                         rc = init_crq_queue(adapter);
1770                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1771                         rc = ibmvnic_reenable_crq_queue(adapter);
1772                         release_sub_crqs(adapter, 1);
1773                 } else {
1774                         rc = ibmvnic_reset_crq(adapter);
1775                         if (!rc)
1776                                 rc = vio_enable_interrupts(adapter->vdev);
1777                 }
1778
1779                 if (rc) {
1780                         netdev_err(adapter->netdev,
1781                                    "Couldn't initialize crq. rc=%d\n", rc);
1782                         return rc;
1783                 }
1784
1785                 rc = ibmvnic_init(adapter);
1786                 if (rc)
1787                         return IBMVNIC_INIT_FAILED;
1788
1789                 /* If the adapter was in PROBE state prior to the reset,
1790                  * exit here.
1791                  */
1792                 if (reset_state == VNIC_PROBED)
1793                         return 0;
1794
1795                 rc = ibmvnic_login(netdev);
1796                 if (rc) {
1797                         adapter->state = VNIC_PROBED;
1798                         return 0;
1799                 }
1800
1801                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1802                     adapter->wait_for_reset) {
1803                         rc = init_resources(adapter);
1804                         if (rc)
1805                                 return rc;
1806                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1807                            adapter->req_tx_queues != old_num_tx_queues) {
1808                         adapter->map_id = 1;
1809                         release_rx_pools(adapter);
1810                         release_tx_pools(adapter);
1811                         init_rx_pools(netdev);
1812                         init_tx_pools(netdev);
1813
1814                         release_napi(adapter);
1815                         init_napi(adapter);
1816                 } else {
1817                         rc = reset_tx_pools(adapter);
1818                         if (rc)
1819                                 return rc;
1820
1821                         rc = reset_rx_pools(adapter);
1822                         if (rc)
1823                                 return rc;
1824                 }
1825                 ibmvnic_disable_irqs(adapter);
1826         }
1827         adapter->state = VNIC_CLOSED;
1828
1829         if (reset_state == VNIC_CLOSED)
1830                 return 0;
1831
1832         rc = __ibmvnic_open(netdev);
1833         if (rc) {
1834                 if (list_empty(&adapter->rwi_list))
1835                         adapter->state = VNIC_CLOSED;
1836                 else
1837                         adapter->state = reset_state;
1838
1839                 return 0;
1840         }
1841
1842         /* kick napi */
1843         for (i = 0; i < adapter->req_rx_queues; i++)
1844                 napi_schedule(&adapter->napi[i]);
1845
1846         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1847             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1848                 netdev_notify_peers(netdev);
1849
1850         netif_carrier_on(netdev);
1851
1852         return 0;
1853 }
1854
1855 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1856 {
1857         struct ibmvnic_rwi *rwi;
1858
1859         mutex_lock(&adapter->rwi_lock);
1860
1861         if (!list_empty(&adapter->rwi_list)) {
1862                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1863                                        list);
1864                 list_del(&rwi->list);
1865         } else {
1866                 rwi = NULL;
1867         }
1868
1869         mutex_unlock(&adapter->rwi_lock);
1870         return rwi;
1871 }
1872
1873 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1874 {
1875         struct ibmvnic_rwi *rwi;
1876
1877         rwi = get_next_rwi(adapter);
1878         while (rwi) {
1879                 kfree(rwi);
1880                 rwi = get_next_rwi(adapter);
1881         }
1882 }
1883
1884 static void __ibmvnic_reset(struct work_struct *work)
1885 {
1886         struct ibmvnic_rwi *rwi;
1887         struct ibmvnic_adapter *adapter;
1888         struct net_device *netdev;
1889         u32 reset_state;
1890         int rc = 0;
1891
1892         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1893         netdev = adapter->netdev;
1894
1895         mutex_lock(&adapter->reset_lock);
1896         adapter->resetting = true;
1897         reset_state = adapter->state;
1898
1899         rwi = get_next_rwi(adapter);
1900         while (rwi) {
1901                 rc = do_reset(adapter, rwi, reset_state);
1902                 kfree(rwi);
1903                 if (rc && rc != IBMVNIC_INIT_FAILED)
1904                         break;
1905
1906                 rwi = get_next_rwi(adapter);
1907         }
1908
1909         if (adapter->wait_for_reset) {
1910                 adapter->wait_for_reset = false;
1911                 adapter->reset_done_rc = rc;
1912                 complete(&adapter->reset_done);
1913         }
1914
1915         if (rc) {
1916                 netdev_dbg(adapter->netdev, "Reset failed\n");
1917                 free_all_rwi(adapter);
1918                 mutex_unlock(&adapter->reset_lock);
1919                 return;
1920         }
1921
1922         adapter->resetting = false;
1923         mutex_unlock(&adapter->reset_lock);
1924 }
1925
1926 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
1927                          enum ibmvnic_reset_reason reason)
1928 {
1929         struct ibmvnic_rwi *rwi, *tmp;
1930         struct net_device *netdev = adapter->netdev;
1931         struct list_head *entry;
1932         int ret;
1933
1934         if (adapter->state == VNIC_REMOVING ||
1935             adapter->state == VNIC_REMOVED ||
1936             adapter->failover_pending) {
1937                 ret = EBUSY;
1938                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
1939                 goto err;
1940         }
1941
1942         if (adapter->state == VNIC_PROBING) {
1943                 netdev_warn(netdev, "Adapter reset during probe\n");
1944                 ret = adapter->init_done_rc = EAGAIN;
1945                 goto err;
1946         }
1947
1948         mutex_lock(&adapter->rwi_lock);
1949
1950         list_for_each(entry, &adapter->rwi_list) {
1951                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
1952                 if (tmp->reset_reason == reason) {
1953                         netdev_dbg(netdev, "Skipping matching reset\n");
1954                         mutex_unlock(&adapter->rwi_lock);
1955                         ret = EBUSY;
1956                         goto err;
1957                 }
1958         }
1959
1960         rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1961         if (!rwi) {
1962                 mutex_unlock(&adapter->rwi_lock);
1963                 ibmvnic_close(netdev);
1964                 ret = ENOMEM;
1965                 goto err;
1966         }
1967
1968         rwi->reset_reason = reason;
1969         list_add_tail(&rwi->list, &adapter->rwi_list);
1970         mutex_unlock(&adapter->rwi_lock);
1971
1972         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1973         schedule_work(&adapter->ibmvnic_reset);
1974
1975         return 0;
1976 err:
1977         if (adapter->wait_for_reset)
1978                 adapter->wait_for_reset = false;
1979         return -ret;
1980 }
1981
1982 static void ibmvnic_tx_timeout(struct net_device *dev)
1983 {
1984         struct ibmvnic_adapter *adapter = netdev_priv(dev);
1985
1986         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1987 }
1988
1989 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1990                                   struct ibmvnic_rx_buff *rx_buff)
1991 {
1992         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1993
1994         rx_buff->skb = NULL;
1995
1996         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1997         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1998
1999         atomic_dec(&pool->available);
2000 }
2001
2002 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2003 {
2004         struct net_device *netdev = napi->dev;
2005         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2006         int scrq_num = (int)(napi - adapter->napi);
2007         int frames_processed = 0;
2008
2009 restart_poll:
2010         while (frames_processed < budget) {
2011                 struct sk_buff *skb;
2012                 struct ibmvnic_rx_buff *rx_buff;
2013                 union sub_crq *next;
2014                 u32 length;
2015                 u16 offset;
2016                 u8 flags = 0;
2017
2018                 if (unlikely(adapter->resetting &&
2019                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2020                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2021                         napi_complete_done(napi, frames_processed);
2022                         return frames_processed;
2023                 }
2024
2025                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2026                         break;
2027                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2028                 rx_buff =
2029                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2030                                                           rx_comp.correlator);
2031                 /* do error checking */
2032                 if (next->rx_comp.rc) {
2033                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2034                                    be16_to_cpu(next->rx_comp.rc));
2035                         /* free the entry */
2036                         next->rx_comp.first = 0;
2037                         dev_kfree_skb_any(rx_buff->skb);
2038                         remove_buff_from_pool(adapter, rx_buff);
2039                         continue;
2040                 } else if (!rx_buff->skb) {
2041                         /* free the entry */
2042                         next->rx_comp.first = 0;
2043                         remove_buff_from_pool(adapter, rx_buff);
2044                         continue;
2045                 }
2046
2047                 length = be32_to_cpu(next->rx_comp.len);
2048                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2049                 flags = next->rx_comp.flags;
2050                 skb = rx_buff->skb;
2051                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2052                                         length);
2053
2054                 /* VLAN Header has been stripped by the system firmware and
2055                  * needs to be inserted by the driver
2056                  */
2057                 if (adapter->rx_vlan_header_insertion &&
2058                     (flags & IBMVNIC_VLAN_STRIPPED))
2059                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2060                                                ntohs(next->rx_comp.vlan_tci));
2061
2062                 /* free the entry */
2063                 next->rx_comp.first = 0;
2064                 remove_buff_from_pool(adapter, rx_buff);
2065
2066                 skb_put(skb, length);
2067                 skb->protocol = eth_type_trans(skb, netdev);
2068                 skb_record_rx_queue(skb, scrq_num);
2069
2070                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2071                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2072                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2073                 }
2074
2075                 length = skb->len;
2076                 napi_gro_receive(napi, skb); /* send it up */
2077                 netdev->stats.rx_packets++;
2078                 netdev->stats.rx_bytes += length;
2079                 adapter->rx_stats_buffers[scrq_num].packets++;
2080                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2081                 frames_processed++;
2082         }
2083
2084         if (adapter->state != VNIC_CLOSING)
2085                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2086
2087         if (frames_processed < budget) {
2088                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2089                 napi_complete_done(napi, frames_processed);
2090                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2091                     napi_reschedule(napi)) {
2092                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2093                         goto restart_poll;
2094                 }
2095         }
2096         return frames_processed;
2097 }
2098
2099 #ifdef CONFIG_NET_POLL_CONTROLLER
2100 static void ibmvnic_netpoll_controller(struct net_device *dev)
2101 {
2102         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2103         int i;
2104
2105         replenish_pools(netdev_priv(dev));
2106         for (i = 0; i < adapter->req_rx_queues; i++)
2107                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
2108                                      adapter->rx_scrq[i]);
2109 }
2110 #endif
2111
2112 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2113 {
2114         int rc, ret;
2115
2116         adapter->fallback.mtu = adapter->req_mtu;
2117         adapter->fallback.rx_queues = adapter->req_rx_queues;
2118         adapter->fallback.tx_queues = adapter->req_tx_queues;
2119         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2120         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2121
2122         init_completion(&adapter->reset_done);
2123         adapter->wait_for_reset = true;
2124         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2125         if (rc)
2126                 return rc;
2127         wait_for_completion(&adapter->reset_done);
2128
2129         ret = 0;
2130         if (adapter->reset_done_rc) {
2131                 ret = -EIO;
2132                 adapter->desired.mtu = adapter->fallback.mtu;
2133                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2134                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2135                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2136                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2137
2138                 init_completion(&adapter->reset_done);
2139                 adapter->wait_for_reset = true;
2140                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2141                 if (rc)
2142                         return ret;
2143                 wait_for_completion(&adapter->reset_done);
2144         }
2145         adapter->wait_for_reset = false;
2146
2147         return ret;
2148 }
2149
2150 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2151 {
2152         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2153
2154         adapter->desired.mtu = new_mtu + ETH_HLEN;
2155
2156         return wait_for_reset(adapter);
2157 }
2158
2159 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2160                                                 struct net_device *dev,
2161                                                 netdev_features_t features)
2162 {
2163         /* Some backing hardware adapters can not
2164          * handle packets with a MSS less than 224
2165          * or with only one segment.
2166          */
2167         if (skb_is_gso(skb)) {
2168                 if (skb_shinfo(skb)->gso_size < 224 ||
2169                     skb_shinfo(skb)->gso_segs == 1)
2170                         features &= ~NETIF_F_GSO_MASK;
2171         }
2172
2173         return features;
2174 }
2175
2176 static const struct net_device_ops ibmvnic_netdev_ops = {
2177         .ndo_open               = ibmvnic_open,
2178         .ndo_stop               = ibmvnic_close,
2179         .ndo_start_xmit         = ibmvnic_xmit,
2180         .ndo_set_rx_mode        = ibmvnic_set_multi,
2181         .ndo_set_mac_address    = ibmvnic_set_mac,
2182         .ndo_validate_addr      = eth_validate_addr,
2183         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2184 #ifdef CONFIG_NET_POLL_CONTROLLER
2185         .ndo_poll_controller    = ibmvnic_netpoll_controller,
2186 #endif
2187         .ndo_change_mtu         = ibmvnic_change_mtu,
2188         .ndo_features_check     = ibmvnic_features_check,
2189 };
2190
2191 /* ethtool functions */
2192
2193 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2194                                       struct ethtool_link_ksettings *cmd)
2195 {
2196         u32 supported, advertising;
2197
2198         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2199                           SUPPORTED_FIBRE);
2200         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2201                             ADVERTISED_FIBRE);
2202         cmd->base.speed = SPEED_1000;
2203         cmd->base.duplex = DUPLEX_FULL;
2204         cmd->base.port = PORT_FIBRE;
2205         cmd->base.phy_address = 0;
2206         cmd->base.autoneg = AUTONEG_ENABLE;
2207
2208         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2209                                                 supported);
2210         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2211                                                 advertising);
2212
2213         return 0;
2214 }
2215
2216 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2217                                 struct ethtool_drvinfo *info)
2218 {
2219         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2220
2221         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2222         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2223         strlcpy(info->fw_version, adapter->fw_version,
2224                 sizeof(info->fw_version));
2225 }
2226
2227 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2228 {
2229         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2230
2231         return adapter->msg_enable;
2232 }
2233
2234 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2235 {
2236         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2237
2238         adapter->msg_enable = data;
2239 }
2240
2241 static u32 ibmvnic_get_link(struct net_device *netdev)
2242 {
2243         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2244
2245         /* Don't need to send a query because we request a logical link up at
2246          * init and then we wait for link state indications
2247          */
2248         return adapter->logical_link_state;
2249 }
2250
2251 static void ibmvnic_get_ringparam(struct net_device *netdev,
2252                                   struct ethtool_ringparam *ring)
2253 {
2254         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2255
2256         ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2257         ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2258         ring->rx_mini_max_pending = 0;
2259         ring->rx_jumbo_max_pending = 0;
2260         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2261         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2262         ring->rx_mini_pending = 0;
2263         ring->rx_jumbo_pending = 0;
2264 }
2265
2266 static int ibmvnic_set_ringparam(struct net_device *netdev,
2267                                  struct ethtool_ringparam *ring)
2268 {
2269         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2270
2271         if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2272             ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2273                 netdev_err(netdev, "Invalid request.\n");
2274                 netdev_err(netdev, "Max tx buffers = %llu\n",
2275                            adapter->max_rx_add_entries_per_subcrq);
2276                 netdev_err(netdev, "Max rx buffers = %llu\n",
2277                            adapter->max_tx_entries_per_subcrq);
2278                 return -EINVAL;
2279         }
2280
2281         adapter->desired.rx_entries = ring->rx_pending;
2282         adapter->desired.tx_entries = ring->tx_pending;
2283
2284         return wait_for_reset(adapter);
2285 }
2286
2287 static void ibmvnic_get_channels(struct net_device *netdev,
2288                                  struct ethtool_channels *channels)
2289 {
2290         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2291
2292         channels->max_rx = adapter->max_rx_queues;
2293         channels->max_tx = adapter->max_tx_queues;
2294         channels->max_other = 0;
2295         channels->max_combined = 0;
2296         channels->rx_count = adapter->req_rx_queues;
2297         channels->tx_count = adapter->req_tx_queues;
2298         channels->other_count = 0;
2299         channels->combined_count = 0;
2300 }
2301
2302 static int ibmvnic_set_channels(struct net_device *netdev,
2303                                 struct ethtool_channels *channels)
2304 {
2305         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2306
2307         adapter->desired.rx_queues = channels->rx_count;
2308         adapter->desired.tx_queues = channels->tx_count;
2309
2310         return wait_for_reset(adapter);
2311 }
2312
2313 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2314 {
2315         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2316         int i;
2317
2318         if (stringset != ETH_SS_STATS)
2319                 return;
2320
2321         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2322                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2323
2324         for (i = 0; i < adapter->req_tx_queues; i++) {
2325                 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2326                 data += ETH_GSTRING_LEN;
2327
2328                 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2329                 data += ETH_GSTRING_LEN;
2330
2331                 snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2332                 data += ETH_GSTRING_LEN;
2333         }
2334
2335         for (i = 0; i < adapter->req_rx_queues; i++) {
2336                 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2337                 data += ETH_GSTRING_LEN;
2338
2339                 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2340                 data += ETH_GSTRING_LEN;
2341
2342                 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2343                 data += ETH_GSTRING_LEN;
2344         }
2345 }
2346
2347 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2348 {
2349         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2350
2351         switch (sset) {
2352         case ETH_SS_STATS:
2353                 return ARRAY_SIZE(ibmvnic_stats) +
2354                        adapter->req_tx_queues * NUM_TX_STATS +
2355                        adapter->req_rx_queues * NUM_RX_STATS;
2356         default:
2357                 return -EOPNOTSUPP;
2358         }
2359 }
2360
2361 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2362                                       struct ethtool_stats *stats, u64 *data)
2363 {
2364         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2365         union ibmvnic_crq crq;
2366         int i, j;
2367
2368         memset(&crq, 0, sizeof(crq));
2369         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2370         crq.request_statistics.cmd = REQUEST_STATISTICS;
2371         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2372         crq.request_statistics.len =
2373             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2374
2375         /* Wait for data to be written */
2376         init_completion(&adapter->stats_done);
2377         ibmvnic_send_crq(adapter, &crq);
2378         wait_for_completion(&adapter->stats_done);
2379
2380         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2381                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2382                                                 ibmvnic_stats[i].offset));
2383
2384         for (j = 0; j < adapter->req_tx_queues; j++) {
2385                 data[i] = adapter->tx_stats_buffers[j].packets;
2386                 i++;
2387                 data[i] = adapter->tx_stats_buffers[j].bytes;
2388                 i++;
2389                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2390                 i++;
2391         }
2392
2393         for (j = 0; j < adapter->req_rx_queues; j++) {
2394                 data[i] = adapter->rx_stats_buffers[j].packets;
2395                 i++;
2396                 data[i] = adapter->rx_stats_buffers[j].bytes;
2397                 i++;
2398                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2399                 i++;
2400         }
2401 }
2402
2403 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2404         .get_drvinfo            = ibmvnic_get_drvinfo,
2405         .get_msglevel           = ibmvnic_get_msglevel,
2406         .set_msglevel           = ibmvnic_set_msglevel,
2407         .get_link               = ibmvnic_get_link,
2408         .get_ringparam          = ibmvnic_get_ringparam,
2409         .set_ringparam          = ibmvnic_set_ringparam,
2410         .get_channels           = ibmvnic_get_channels,
2411         .set_channels           = ibmvnic_set_channels,
2412         .get_strings            = ibmvnic_get_strings,
2413         .get_sset_count         = ibmvnic_get_sset_count,
2414         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2415         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2416 };
2417
2418 /* Routines for managing CRQs/sCRQs  */
2419
2420 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2421                                    struct ibmvnic_sub_crq_queue *scrq)
2422 {
2423         int rc;
2424
2425         if (scrq->irq) {
2426                 free_irq(scrq->irq, scrq);
2427                 irq_dispose_mapping(scrq->irq);
2428                 scrq->irq = 0;
2429         }
2430
2431         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2432         atomic_set(&scrq->used, 0);
2433         scrq->cur = 0;
2434
2435         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2436                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2437         return rc;
2438 }
2439
2440 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2441 {
2442         int i, rc;
2443
2444         for (i = 0; i < adapter->req_tx_queues; i++) {
2445                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2446                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2447                 if (rc)
2448                         return rc;
2449         }
2450
2451         for (i = 0; i < adapter->req_rx_queues; i++) {
2452                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2453                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2454                 if (rc)
2455                         return rc;
2456         }
2457
2458         return rc;
2459 }
2460
2461 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2462                                   struct ibmvnic_sub_crq_queue *scrq,
2463                                   bool do_h_free)
2464 {
2465         struct device *dev = &adapter->vdev->dev;
2466         long rc;
2467
2468         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2469
2470         if (do_h_free) {
2471                 /* Close the sub-crqs */
2472                 do {
2473                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2474                                                 adapter->vdev->unit_address,
2475                                                 scrq->crq_num);
2476                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2477
2478                 if (rc) {
2479                         netdev_err(adapter->netdev,
2480                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2481                                    scrq->crq_num, rc);
2482                 }
2483         }
2484
2485         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2486                          DMA_BIDIRECTIONAL);
2487         free_pages((unsigned long)scrq->msgs, 2);
2488         kfree(scrq);
2489 }
2490
2491 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2492                                                         *adapter)
2493 {
2494         struct device *dev = &adapter->vdev->dev;
2495         struct ibmvnic_sub_crq_queue *scrq;
2496         int rc;
2497
2498         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2499         if (!scrq)
2500                 return NULL;
2501
2502         scrq->msgs =
2503                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2504         if (!scrq->msgs) {
2505                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2506                 goto zero_page_failed;
2507         }
2508
2509         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2510                                          DMA_BIDIRECTIONAL);
2511         if (dma_mapping_error(dev, scrq->msg_token)) {
2512                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2513                 goto map_failed;
2514         }
2515
2516         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2517                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2518
2519         if (rc == H_RESOURCE)
2520                 rc = ibmvnic_reset_crq(adapter);
2521
2522         if (rc == H_CLOSED) {
2523                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2524         } else if (rc) {
2525                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2526                 goto reg_failed;
2527         }
2528
2529         scrq->adapter = adapter;
2530         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2531         spin_lock_init(&scrq->lock);
2532
2533         netdev_dbg(adapter->netdev,
2534                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2535                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2536
2537         return scrq;
2538
2539 reg_failed:
2540         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2541                          DMA_BIDIRECTIONAL);
2542 map_failed:
2543         free_pages((unsigned long)scrq->msgs, 2);
2544 zero_page_failed:
2545         kfree(scrq);
2546
2547         return NULL;
2548 }
2549
2550 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2551 {
2552         int i;
2553
2554         if (adapter->tx_scrq) {
2555                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2556                         if (!adapter->tx_scrq[i])
2557                                 continue;
2558
2559                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2560                                    i);
2561                         if (adapter->tx_scrq[i]->irq) {
2562                                 free_irq(adapter->tx_scrq[i]->irq,
2563                                          adapter->tx_scrq[i]);
2564                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2565                                 adapter->tx_scrq[i]->irq = 0;
2566                         }
2567
2568                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2569                                               do_h_free);
2570                 }
2571
2572                 kfree(adapter->tx_scrq);
2573                 adapter->tx_scrq = NULL;
2574                 adapter->num_active_tx_scrqs = 0;
2575         }
2576
2577         if (adapter->rx_scrq) {
2578                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2579                         if (!adapter->rx_scrq[i])
2580                                 continue;
2581
2582                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2583                                    i);
2584                         if (adapter->rx_scrq[i]->irq) {
2585                                 free_irq(adapter->rx_scrq[i]->irq,
2586                                          adapter->rx_scrq[i]);
2587                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2588                                 adapter->rx_scrq[i]->irq = 0;
2589                         }
2590
2591                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2592                                               do_h_free);
2593                 }
2594
2595                 kfree(adapter->rx_scrq);
2596                 adapter->rx_scrq = NULL;
2597                 adapter->num_active_rx_scrqs = 0;
2598         }
2599 }
2600
2601 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2602                             struct ibmvnic_sub_crq_queue *scrq)
2603 {
2604         struct device *dev = &adapter->vdev->dev;
2605         unsigned long rc;
2606
2607         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2608                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2609         if (rc)
2610                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2611                         scrq->hw_irq, rc);
2612         return rc;
2613 }
2614
2615 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2616                            struct ibmvnic_sub_crq_queue *scrq)
2617 {
2618         struct device *dev = &adapter->vdev->dev;
2619         unsigned long rc;
2620         u64 val;
2621
2622         if (scrq->hw_irq > 0x100000000ULL) {
2623                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2624                 return 1;
2625         }
2626
2627         val = (0xff000000) | scrq->hw_irq;
2628         rc = plpar_hcall_norets(H_EOI, val);
2629         if (rc)
2630                 dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2631                         val, rc);
2632
2633         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2634                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2635         if (rc)
2636                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2637                         scrq->hw_irq, rc);
2638         return rc;
2639 }
2640
2641 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2642                                struct ibmvnic_sub_crq_queue *scrq)
2643 {
2644         struct device *dev = &adapter->vdev->dev;
2645         struct ibmvnic_tx_pool *tx_pool;
2646         struct ibmvnic_tx_buff *txbuff;
2647         union sub_crq *next;
2648         int index;
2649         int i, j;
2650         u8 *first;
2651
2652 restart_loop:
2653         while (pending_scrq(adapter, scrq)) {
2654                 unsigned int pool = scrq->pool_index;
2655                 int num_entries = 0;
2656
2657                 next = ibmvnic_next_scrq(adapter, scrq);
2658                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2659                         if (next->tx_comp.rcs[i]) {
2660                                 dev_err(dev, "tx error %x\n",
2661                                         next->tx_comp.rcs[i]);
2662                                 continue;
2663                         }
2664                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2665                         if (index & IBMVNIC_TSO_POOL_MASK) {
2666                                 tx_pool = &adapter->tso_pool[pool];
2667                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2668                         } else {
2669                                 tx_pool = &adapter->tx_pool[pool];
2670                         }
2671
2672                         txbuff = &tx_pool->tx_buff[index];
2673
2674                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2675                                 if (!txbuff->data_dma[j])
2676                                         continue;
2677
2678                                 txbuff->data_dma[j] = 0;
2679                         }
2680                         /* if sub_crq was sent indirectly */
2681                         first = &txbuff->indir_arr[0].generic.first;
2682                         if (*first == IBMVNIC_CRQ_CMD) {
2683                                 dma_unmap_single(dev, txbuff->indir_dma,
2684                                                  sizeof(txbuff->indir_arr),
2685                                                  DMA_TO_DEVICE);
2686                                 *first = 0;
2687                         }
2688
2689                         if (txbuff->last_frag) {
2690                                 dev_kfree_skb_any(txbuff->skb);
2691                                 txbuff->skb = NULL;
2692                         }
2693
2694                         num_entries += txbuff->num_entries;
2695
2696                         tx_pool->free_map[tx_pool->producer_index] = index;
2697                         tx_pool->producer_index =
2698                                 (tx_pool->producer_index + 1) %
2699                                         tx_pool->num_buffers;
2700                 }
2701                 /* remove tx_comp scrq*/
2702                 next->tx_comp.first = 0;
2703
2704                 if (atomic_sub_return(num_entries, &scrq->used) <=
2705                     (adapter->req_tx_entries_per_subcrq / 2) &&
2706                     __netif_subqueue_stopped(adapter->netdev,
2707                                              scrq->pool_index)) {
2708                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2709                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2710                                    scrq->pool_index);
2711                 }
2712         }
2713
2714         enable_scrq_irq(adapter, scrq);
2715
2716         if (pending_scrq(adapter, scrq)) {
2717                 disable_scrq_irq(adapter, scrq);
2718                 goto restart_loop;
2719         }
2720
2721         return 0;
2722 }
2723
2724 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2725 {
2726         struct ibmvnic_sub_crq_queue *scrq = instance;
2727         struct ibmvnic_adapter *adapter = scrq->adapter;
2728
2729         disable_scrq_irq(adapter, scrq);
2730         ibmvnic_complete_tx(adapter, scrq);
2731
2732         return IRQ_HANDLED;
2733 }
2734
2735 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2736 {
2737         struct ibmvnic_sub_crq_queue *scrq = instance;
2738         struct ibmvnic_adapter *adapter = scrq->adapter;
2739
2740         /* When booting a kdump kernel we can hit pending interrupts
2741          * prior to completing driver initialization.
2742          */
2743         if (unlikely(adapter->state != VNIC_OPEN))
2744                 return IRQ_NONE;
2745
2746         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2747
2748         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2749                 disable_scrq_irq(adapter, scrq);
2750                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2751         }
2752
2753         return IRQ_HANDLED;
2754 }
2755
2756 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2757 {
2758         struct device *dev = &adapter->vdev->dev;
2759         struct ibmvnic_sub_crq_queue *scrq;
2760         int i = 0, j = 0;
2761         int rc = 0;
2762
2763         for (i = 0; i < adapter->req_tx_queues; i++) {
2764                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2765                            i);
2766                 scrq = adapter->tx_scrq[i];
2767                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2768
2769                 if (!scrq->irq) {
2770                         rc = -EINVAL;
2771                         dev_err(dev, "Error mapping irq\n");
2772                         goto req_tx_irq_failed;
2773                 }
2774
2775                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2776                                  0, "ibmvnic_tx", scrq);
2777
2778                 if (rc) {
2779                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2780                                 scrq->irq, rc);
2781                         irq_dispose_mapping(scrq->irq);
2782                         goto req_tx_irq_failed;
2783                 }
2784         }
2785
2786         for (i = 0; i < adapter->req_rx_queues; i++) {
2787                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2788                            i);
2789                 scrq = adapter->rx_scrq[i];
2790                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2791                 if (!scrq->irq) {
2792                         rc = -EINVAL;
2793                         dev_err(dev, "Error mapping irq\n");
2794                         goto req_rx_irq_failed;
2795                 }
2796                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2797                                  0, "ibmvnic_rx", scrq);
2798                 if (rc) {
2799                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2800                                 scrq->irq, rc);
2801                         irq_dispose_mapping(scrq->irq);
2802                         goto req_rx_irq_failed;
2803                 }
2804         }
2805         return rc;
2806
2807 req_rx_irq_failed:
2808         for (j = 0; j < i; j++) {
2809                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2810                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2811         }
2812         i = adapter->req_tx_queues;
2813 req_tx_irq_failed:
2814         for (j = 0; j < i; j++) {
2815                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2816                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2817         }
2818         release_sub_crqs(adapter, 1);
2819         return rc;
2820 }
2821
2822 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2823 {
2824         struct device *dev = &adapter->vdev->dev;
2825         struct ibmvnic_sub_crq_queue **allqueues;
2826         int registered_queues = 0;
2827         int total_queues;
2828         int more = 0;
2829         int i;
2830
2831         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2832
2833         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2834         if (!allqueues)
2835                 return -1;
2836
2837         for (i = 0; i < total_queues; i++) {
2838                 allqueues[i] = init_sub_crq_queue(adapter);
2839                 if (!allqueues[i]) {
2840                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2841                         break;
2842                 }
2843                 registered_queues++;
2844         }
2845
2846         /* Make sure we were able to register the minimum number of queues */
2847         if (registered_queues <
2848             adapter->min_tx_queues + adapter->min_rx_queues) {
2849                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2850                 goto tx_failed;
2851         }
2852
2853         /* Distribute the failed allocated queues*/
2854         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2855                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2856                 switch (i % 3) {
2857                 case 0:
2858                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2859                                 adapter->req_rx_queues--;
2860                         else
2861                                 more++;
2862                         break;
2863                 case 1:
2864                         if (adapter->req_tx_queues > adapter->min_tx_queues)
2865                                 adapter->req_tx_queues--;
2866                         else
2867                                 more++;
2868                         break;
2869                 }
2870         }
2871
2872         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2873                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
2874         if (!adapter->tx_scrq)
2875                 goto tx_failed;
2876
2877         for (i = 0; i < adapter->req_tx_queues; i++) {
2878                 adapter->tx_scrq[i] = allqueues[i];
2879                 adapter->tx_scrq[i]->pool_index = i;
2880                 adapter->num_active_tx_scrqs++;
2881         }
2882
2883         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2884                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
2885         if (!adapter->rx_scrq)
2886                 goto rx_failed;
2887
2888         for (i = 0; i < adapter->req_rx_queues; i++) {
2889                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2890                 adapter->rx_scrq[i]->scrq_num = i;
2891                 adapter->num_active_rx_scrqs++;
2892         }
2893
2894         kfree(allqueues);
2895         return 0;
2896
2897 rx_failed:
2898         kfree(adapter->tx_scrq);
2899         adapter->tx_scrq = NULL;
2900 tx_failed:
2901         for (i = 0; i < registered_queues; i++)
2902                 release_sub_crq_queue(adapter, allqueues[i], 1);
2903         kfree(allqueues);
2904         return -1;
2905 }
2906
2907 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2908 {
2909         struct device *dev = &adapter->vdev->dev;
2910         union ibmvnic_crq crq;
2911         int max_entries;
2912
2913         if (!retry) {
2914                 /* Sub-CRQ entries are 32 byte long */
2915                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2916
2917                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
2918                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
2919                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2920                         return;
2921                 }
2922
2923                 if (adapter->desired.mtu)
2924                         adapter->req_mtu = adapter->desired.mtu;
2925                 else
2926                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2927
2928                 if (!adapter->desired.tx_entries)
2929                         adapter->desired.tx_entries =
2930                                         adapter->max_tx_entries_per_subcrq;
2931                 if (!adapter->desired.rx_entries)
2932                         adapter->desired.rx_entries =
2933                                         adapter->max_rx_add_entries_per_subcrq;
2934
2935                 max_entries = IBMVNIC_MAX_LTB_SIZE /
2936                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
2937
2938                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2939                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
2940                         adapter->desired.tx_entries = max_entries;
2941                 }
2942
2943                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2944                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
2945                         adapter->desired.rx_entries = max_entries;
2946                 }
2947
2948                 if (adapter->desired.tx_entries)
2949                         adapter->req_tx_entries_per_subcrq =
2950                                         adapter->desired.tx_entries;
2951                 else
2952                         adapter->req_tx_entries_per_subcrq =
2953                                         adapter->max_tx_entries_per_subcrq;
2954
2955                 if (adapter->desired.rx_entries)
2956                         adapter->req_rx_add_entries_per_subcrq =
2957                                         adapter->desired.rx_entries;
2958                 else
2959                         adapter->req_rx_add_entries_per_subcrq =
2960                                         adapter->max_rx_add_entries_per_subcrq;
2961
2962                 if (adapter->desired.tx_queues)
2963                         adapter->req_tx_queues =
2964                                         adapter->desired.tx_queues;
2965                 else
2966                         adapter->req_tx_queues =
2967                                         adapter->opt_tx_comp_sub_queues;
2968
2969                 if (adapter->desired.rx_queues)
2970                         adapter->req_rx_queues =
2971                                         adapter->desired.rx_queues;
2972                 else
2973                         adapter->req_rx_queues =
2974                                         adapter->opt_rx_comp_queues;
2975
2976                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2977         }
2978
2979         memset(&crq, 0, sizeof(crq));
2980         crq.request_capability.first = IBMVNIC_CRQ_CMD;
2981         crq.request_capability.cmd = REQUEST_CAPABILITY;
2982
2983         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2984         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2985         atomic_inc(&adapter->running_cap_crqs);
2986         ibmvnic_send_crq(adapter, &crq);
2987
2988         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2989         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2990         atomic_inc(&adapter->running_cap_crqs);
2991         ibmvnic_send_crq(adapter, &crq);
2992
2993         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2994         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2995         atomic_inc(&adapter->running_cap_crqs);
2996         ibmvnic_send_crq(adapter, &crq);
2997
2998         crq.request_capability.capability =
2999             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3000         crq.request_capability.number =
3001             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3002         atomic_inc(&adapter->running_cap_crqs);
3003         ibmvnic_send_crq(adapter, &crq);
3004
3005         crq.request_capability.capability =
3006             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3007         crq.request_capability.number =
3008             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3009         atomic_inc(&adapter->running_cap_crqs);
3010         ibmvnic_send_crq(adapter, &crq);
3011
3012         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3013         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3014         atomic_inc(&adapter->running_cap_crqs);
3015         ibmvnic_send_crq(adapter, &crq);
3016
3017         if (adapter->netdev->flags & IFF_PROMISC) {
3018                 if (adapter->promisc_supported) {
3019                         crq.request_capability.capability =
3020                             cpu_to_be16(PROMISC_REQUESTED);
3021                         crq.request_capability.number = cpu_to_be64(1);
3022                         atomic_inc(&adapter->running_cap_crqs);
3023                         ibmvnic_send_crq(adapter, &crq);
3024                 }
3025         } else {
3026                 crq.request_capability.capability =
3027                     cpu_to_be16(PROMISC_REQUESTED);
3028                 crq.request_capability.number = cpu_to_be64(0);
3029                 atomic_inc(&adapter->running_cap_crqs);
3030                 ibmvnic_send_crq(adapter, &crq);
3031         }
3032 }
3033
3034 static int pending_scrq(struct ibmvnic_adapter *adapter,
3035                         struct ibmvnic_sub_crq_queue *scrq)
3036 {
3037         union sub_crq *entry = &scrq->msgs[scrq->cur];
3038
3039         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3040                 return 1;
3041         else
3042                 return 0;
3043 }
3044
3045 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3046                                         struct ibmvnic_sub_crq_queue *scrq)
3047 {
3048         union sub_crq *entry;
3049         unsigned long flags;
3050
3051         spin_lock_irqsave(&scrq->lock, flags);
3052         entry = &scrq->msgs[scrq->cur];
3053         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3054                 if (++scrq->cur == scrq->size)
3055                         scrq->cur = 0;
3056         } else {
3057                 entry = NULL;
3058         }
3059         spin_unlock_irqrestore(&scrq->lock, flags);
3060
3061         return entry;
3062 }
3063
3064 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3065 {
3066         struct ibmvnic_crq_queue *queue = &adapter->crq;
3067         union ibmvnic_crq *crq;
3068
3069         crq = &queue->msgs[queue->cur];
3070         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3071                 if (++queue->cur == queue->size)
3072                         queue->cur = 0;
3073         } else {
3074                 crq = NULL;
3075         }
3076
3077         return crq;
3078 }
3079
3080 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3081                        union sub_crq *sub_crq)
3082 {
3083         unsigned int ua = adapter->vdev->unit_address;
3084         struct device *dev = &adapter->vdev->dev;
3085         u64 *u64_crq = (u64 *)sub_crq;
3086         int rc;
3087
3088         netdev_dbg(adapter->netdev,
3089                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3090                    (unsigned long int)cpu_to_be64(remote_handle),
3091                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3092                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3093                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3094                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3095
3096         /* Make sure the hypervisor sees the complete request */
3097         mb();
3098
3099         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3100                                 cpu_to_be64(remote_handle),
3101                                 cpu_to_be64(u64_crq[0]),
3102                                 cpu_to_be64(u64_crq[1]),
3103                                 cpu_to_be64(u64_crq[2]),
3104                                 cpu_to_be64(u64_crq[3]));
3105
3106         if (rc) {
3107                 if (rc == H_CLOSED)
3108                         dev_warn(dev, "CRQ Queue closed\n");
3109                 dev_err(dev, "Send error (rc=%d)\n", rc);
3110         }
3111
3112         return rc;
3113 }
3114
3115 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3116                                 u64 remote_handle, u64 ioba, u64 num_entries)
3117 {
3118         unsigned int ua = adapter->vdev->unit_address;
3119         struct device *dev = &adapter->vdev->dev;
3120         int rc;
3121
3122         /* Make sure the hypervisor sees the complete request */
3123         mb();
3124         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3125                                 cpu_to_be64(remote_handle),
3126                                 ioba, num_entries);
3127
3128         if (rc) {
3129                 if (rc == H_CLOSED)
3130                         dev_warn(dev, "CRQ Queue closed\n");
3131                 dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
3132         }
3133
3134         return rc;
3135 }
3136
3137 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3138                             union ibmvnic_crq *crq)
3139 {
3140         unsigned int ua = adapter->vdev->unit_address;
3141         struct device *dev = &adapter->vdev->dev;
3142         u64 *u64_crq = (u64 *)crq;
3143         int rc;
3144
3145         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3146                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3147                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3148
3149         /* Make sure the hypervisor sees the complete request */
3150         mb();
3151
3152         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3153                                 cpu_to_be64(u64_crq[0]),
3154                                 cpu_to_be64(u64_crq[1]));
3155
3156         if (rc) {
3157                 if (rc == H_CLOSED) {
3158                         dev_warn(dev, "CRQ Queue closed\n");
3159                         if (adapter->resetting)
3160                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3161                 }
3162
3163                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3164         }
3165
3166         return rc;
3167 }
3168
3169 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3170 {
3171         union ibmvnic_crq crq;
3172
3173         memset(&crq, 0, sizeof(crq));
3174         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3175         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3176         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3177
3178         return ibmvnic_send_crq(adapter, &crq);
3179 }
3180
3181 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3182 {
3183         union ibmvnic_crq crq;
3184
3185         memset(&crq, 0, sizeof(crq));
3186         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3187         crq.version_exchange.cmd = VERSION_EXCHANGE;
3188         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3189
3190         return ibmvnic_send_crq(adapter, &crq);
3191 }
3192
3193 struct vnic_login_client_data {
3194         u8      type;
3195         __be16  len;
3196         char    name[];
3197 } __packed;
3198
3199 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3200 {
3201         int len;
3202
3203         /* Calculate the amount of buffer space needed for the
3204          * vnic client data in the login buffer. There are four entries,
3205          * OS name, LPAR name, device name, and a null last entry.
3206          */
3207         len = 4 * sizeof(struct vnic_login_client_data);
3208         len += 6; /* "Linux" plus NULL */
3209         len += strlen(utsname()->nodename) + 1;
3210         len += strlen(adapter->netdev->name) + 1;
3211
3212         return len;
3213 }
3214
3215 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3216                                  struct vnic_login_client_data *vlcd)
3217 {
3218         const char *os_name = "Linux";
3219         int len;
3220
3221         /* Type 1 - LPAR OS */
3222         vlcd->type = 1;
3223         len = strlen(os_name) + 1;
3224         vlcd->len = cpu_to_be16(len);
3225         strncpy(vlcd->name, os_name, len);
3226         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3227
3228         /* Type 2 - LPAR name */
3229         vlcd->type = 2;
3230         len = strlen(utsname()->nodename) + 1;
3231         vlcd->len = cpu_to_be16(len);
3232         strncpy(vlcd->name, utsname()->nodename, len);
3233         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3234
3235         /* Type 3 - device name */
3236         vlcd->type = 3;
3237         len = strlen(adapter->netdev->name) + 1;
3238         vlcd->len = cpu_to_be16(len);
3239         strncpy(vlcd->name, adapter->netdev->name, len);
3240 }
3241
3242 static int send_login(struct ibmvnic_adapter *adapter)
3243 {
3244         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3245         struct ibmvnic_login_buffer *login_buffer;
3246         struct device *dev = &adapter->vdev->dev;
3247         dma_addr_t rsp_buffer_token;
3248         dma_addr_t buffer_token;
3249         size_t rsp_buffer_size;
3250         union ibmvnic_crq crq;
3251         size_t buffer_size;
3252         __be64 *tx_list_p;
3253         __be64 *rx_list_p;
3254         int client_data_len;
3255         struct vnic_login_client_data *vlcd;
3256         int i;
3257
3258         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3259                 netdev_err(adapter->netdev,
3260                            "RX or TX queues are not allocated, device login failed\n");
3261                 return -1;
3262         }
3263
3264         release_login_rsp_buffer(adapter);
3265         client_data_len = vnic_client_data_len(adapter);
3266
3267         buffer_size =
3268             sizeof(struct ibmvnic_login_buffer) +
3269             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3270             client_data_len;
3271
3272         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3273         if (!login_buffer)
3274                 goto buf_alloc_failed;
3275
3276         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3277                                       DMA_TO_DEVICE);
3278         if (dma_mapping_error(dev, buffer_token)) {
3279                 dev_err(dev, "Couldn't map login buffer\n");
3280                 goto buf_map_failed;
3281         }
3282
3283         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3284                           sizeof(u64) * adapter->req_tx_queues +
3285                           sizeof(u64) * adapter->req_rx_queues +
3286                           sizeof(u64) * adapter->req_rx_queues +
3287                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3288
3289         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3290         if (!login_rsp_buffer)
3291                 goto buf_rsp_alloc_failed;
3292
3293         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3294                                           rsp_buffer_size, DMA_FROM_DEVICE);
3295         if (dma_mapping_error(dev, rsp_buffer_token)) {
3296                 dev_err(dev, "Couldn't map login rsp buffer\n");
3297                 goto buf_rsp_map_failed;
3298         }
3299
3300         adapter->login_buf = login_buffer;
3301         adapter->login_buf_token = buffer_token;
3302         adapter->login_buf_sz = buffer_size;
3303         adapter->login_rsp_buf = login_rsp_buffer;
3304         adapter->login_rsp_buf_token = rsp_buffer_token;
3305         adapter->login_rsp_buf_sz = rsp_buffer_size;
3306
3307         login_buffer->len = cpu_to_be32(buffer_size);
3308         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3309         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3310         login_buffer->off_txcomp_subcrqs =
3311             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3312         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3313         login_buffer->off_rxcomp_subcrqs =
3314             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3315                         sizeof(u64) * adapter->req_tx_queues);
3316         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3317         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3318
3319         tx_list_p = (__be64 *)((char *)login_buffer +
3320                                       sizeof(struct ibmvnic_login_buffer));
3321         rx_list_p = (__be64 *)((char *)login_buffer +
3322                                       sizeof(struct ibmvnic_login_buffer) +
3323                                       sizeof(u64) * adapter->req_tx_queues);
3324
3325         for (i = 0; i < adapter->req_tx_queues; i++) {
3326                 if (adapter->tx_scrq[i]) {
3327                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3328                                                    crq_num);
3329                 }
3330         }
3331
3332         for (i = 0; i < adapter->req_rx_queues; i++) {
3333                 if (adapter->rx_scrq[i]) {
3334                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3335                                                    crq_num);
3336                 }
3337         }
3338
3339         /* Insert vNIC login client data */
3340         vlcd = (struct vnic_login_client_data *)
3341                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3342         login_buffer->client_data_offset =
3343                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3344         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3345
3346         vnic_add_client_data(adapter, vlcd);
3347
3348         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3349         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3350                 netdev_dbg(adapter->netdev, "%016lx\n",
3351                            ((unsigned long int *)(adapter->login_buf))[i]);
3352         }
3353
3354         memset(&crq, 0, sizeof(crq));
3355         crq.login.first = IBMVNIC_CRQ_CMD;
3356         crq.login.cmd = LOGIN;
3357         crq.login.ioba = cpu_to_be32(buffer_token);
3358         crq.login.len = cpu_to_be32(buffer_size);
3359         ibmvnic_send_crq(adapter, &crq);
3360
3361         return 0;
3362
3363 buf_rsp_map_failed:
3364         kfree(login_rsp_buffer);
3365 buf_rsp_alloc_failed:
3366         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3367 buf_map_failed:
3368         kfree(login_buffer);
3369 buf_alloc_failed:
3370         return -1;
3371 }
3372
3373 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3374                              u32 len, u8 map_id)
3375 {
3376         union ibmvnic_crq crq;
3377
3378         memset(&crq, 0, sizeof(crq));
3379         crq.request_map.first = IBMVNIC_CRQ_CMD;
3380         crq.request_map.cmd = REQUEST_MAP;
3381         crq.request_map.map_id = map_id;
3382         crq.request_map.ioba = cpu_to_be32(addr);
3383         crq.request_map.len = cpu_to_be32(len);
3384         ibmvnic_send_crq(adapter, &crq);
3385 }
3386
3387 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3388 {
3389         union ibmvnic_crq crq;
3390
3391         memset(&crq, 0, sizeof(crq));
3392         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3393         crq.request_unmap.cmd = REQUEST_UNMAP;
3394         crq.request_unmap.map_id = map_id;
3395         ibmvnic_send_crq(adapter, &crq);
3396 }
3397
3398 static void send_map_query(struct ibmvnic_adapter *adapter)
3399 {
3400         union ibmvnic_crq crq;
3401
3402         memset(&crq, 0, sizeof(crq));
3403         crq.query_map.first = IBMVNIC_CRQ_CMD;
3404         crq.query_map.cmd = QUERY_MAP;
3405         ibmvnic_send_crq(adapter, &crq);
3406 }
3407
3408 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3409 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3410 {
3411         union ibmvnic_crq crq;
3412
3413         atomic_set(&adapter->running_cap_crqs, 0);
3414         memset(&crq, 0, sizeof(crq));
3415         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3416         crq.query_capability.cmd = QUERY_CAPABILITY;
3417
3418         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3419         atomic_inc(&adapter->running_cap_crqs);
3420         ibmvnic_send_crq(adapter, &crq);
3421
3422         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3423         atomic_inc(&adapter->running_cap_crqs);
3424         ibmvnic_send_crq(adapter, &crq);
3425
3426         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3427         atomic_inc(&adapter->running_cap_crqs);
3428         ibmvnic_send_crq(adapter, &crq);
3429
3430         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3431         atomic_inc(&adapter->running_cap_crqs);
3432         ibmvnic_send_crq(adapter, &crq);
3433
3434         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3435         atomic_inc(&adapter->running_cap_crqs);
3436         ibmvnic_send_crq(adapter, &crq);
3437
3438         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3439         atomic_inc(&adapter->running_cap_crqs);
3440         ibmvnic_send_crq(adapter, &crq);
3441
3442         crq.query_capability.capability =
3443             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3444         atomic_inc(&adapter->running_cap_crqs);
3445         ibmvnic_send_crq(adapter, &crq);
3446
3447         crq.query_capability.capability =
3448             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3449         atomic_inc(&adapter->running_cap_crqs);
3450         ibmvnic_send_crq(adapter, &crq);
3451
3452         crq.query_capability.capability =
3453             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3454         atomic_inc(&adapter->running_cap_crqs);
3455         ibmvnic_send_crq(adapter, &crq);
3456
3457         crq.query_capability.capability =
3458             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3459         atomic_inc(&adapter->running_cap_crqs);
3460         ibmvnic_send_crq(adapter, &crq);
3461
3462         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3463         atomic_inc(&adapter->running_cap_crqs);
3464         ibmvnic_send_crq(adapter, &crq);
3465
3466         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3467         atomic_inc(&adapter->running_cap_crqs);
3468         ibmvnic_send_crq(adapter, &crq);
3469
3470         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3471         atomic_inc(&adapter->running_cap_crqs);
3472         ibmvnic_send_crq(adapter, &crq);
3473
3474         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3475         atomic_inc(&adapter->running_cap_crqs);
3476         ibmvnic_send_crq(adapter, &crq);
3477
3478         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3479         atomic_inc(&adapter->running_cap_crqs);
3480         ibmvnic_send_crq(adapter, &crq);
3481
3482         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3483         atomic_inc(&adapter->running_cap_crqs);
3484         ibmvnic_send_crq(adapter, &crq);
3485
3486         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3487         atomic_inc(&adapter->running_cap_crqs);
3488         ibmvnic_send_crq(adapter, &crq);
3489
3490         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3491         atomic_inc(&adapter->running_cap_crqs);
3492         ibmvnic_send_crq(adapter, &crq);
3493
3494         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3495         atomic_inc(&adapter->running_cap_crqs);
3496         ibmvnic_send_crq(adapter, &crq);
3497
3498         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3499         atomic_inc(&adapter->running_cap_crqs);
3500         ibmvnic_send_crq(adapter, &crq);
3501
3502         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3503         atomic_inc(&adapter->running_cap_crqs);
3504         ibmvnic_send_crq(adapter, &crq);
3505
3506         crq.query_capability.capability =
3507                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3508         atomic_inc(&adapter->running_cap_crqs);
3509         ibmvnic_send_crq(adapter, &crq);
3510
3511         crq.query_capability.capability =
3512                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3513         atomic_inc(&adapter->running_cap_crqs);
3514         ibmvnic_send_crq(adapter, &crq);
3515
3516         crq.query_capability.capability =
3517                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3518         atomic_inc(&adapter->running_cap_crqs);
3519         ibmvnic_send_crq(adapter, &crq);
3520
3521         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3522         atomic_inc(&adapter->running_cap_crqs);
3523         ibmvnic_send_crq(adapter, &crq);
3524 }
3525
3526 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3527                                 struct ibmvnic_adapter *adapter)
3528 {
3529         struct device *dev = &adapter->vdev->dev;
3530
3531         if (crq->get_vpd_size_rsp.rc.code) {
3532                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3533                         crq->get_vpd_size_rsp.rc.code);
3534                 complete(&adapter->fw_done);
3535                 return;
3536         }
3537
3538         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3539         complete(&adapter->fw_done);
3540 }
3541
3542 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3543                            struct ibmvnic_adapter *adapter)
3544 {
3545         struct device *dev = &adapter->vdev->dev;
3546         unsigned char *substr = NULL;
3547         u8 fw_level_len = 0;
3548
3549         memset(adapter->fw_version, 0, 32);
3550
3551         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3552                          DMA_FROM_DEVICE);
3553
3554         if (crq->get_vpd_rsp.rc.code) {
3555                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3556                         crq->get_vpd_rsp.rc.code);
3557                 goto complete;
3558         }
3559
3560         /* get the position of the firmware version info
3561          * located after the ASCII 'RM' substring in the buffer
3562          */
3563         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3564         if (!substr) {
3565                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3566                 goto complete;
3567         }
3568
3569         /* get length of firmware level ASCII substring */
3570         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3571                 fw_level_len = *(substr + 2);
3572         } else {
3573                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3574                 goto complete;
3575         }
3576
3577         /* copy firmware version string from vpd into adapter */
3578         if ((substr + 3 + fw_level_len) <
3579             (adapter->vpd->buff + adapter->vpd->len)) {
3580                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3581         } else {
3582                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3583         }
3584
3585 complete:
3586         if (adapter->fw_version[0] == '\0')
3587                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3588         complete(&adapter->fw_done);
3589 }
3590
3591 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3592 {
3593         struct device *dev = &adapter->vdev->dev;
3594         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3595         union ibmvnic_crq crq;
3596         int i;
3597
3598         dma_unmap_single(dev, adapter->ip_offload_tok,
3599                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3600
3601         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3602         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3603                 netdev_dbg(adapter->netdev, "%016lx\n",
3604                            ((unsigned long int *)(buf))[i]);
3605
3606         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3607         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3608         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3609                    buf->tcp_ipv4_chksum);
3610         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3611                    buf->tcp_ipv6_chksum);
3612         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3613                    buf->udp_ipv4_chksum);
3614         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3615                    buf->udp_ipv6_chksum);
3616         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3617                    buf->large_tx_ipv4);
3618         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3619                    buf->large_tx_ipv6);
3620         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3621                    buf->large_rx_ipv4);
3622         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3623                    buf->large_rx_ipv6);
3624         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3625                    buf->max_ipv4_header_size);
3626         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3627                    buf->max_ipv6_header_size);
3628         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3629                    buf->max_tcp_header_size);
3630         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3631                    buf->max_udp_header_size);
3632         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3633                    buf->max_large_tx_size);
3634         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3635                    buf->max_large_rx_size);
3636         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3637                    buf->ipv6_extension_header);
3638         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3639                    buf->tcp_pseudosum_req);
3640         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3641                    buf->num_ipv6_ext_headers);
3642         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3643                    buf->off_ipv6_ext_headers);
3644
3645         adapter->ip_offload_ctrl_tok =
3646             dma_map_single(dev, &adapter->ip_offload_ctrl,
3647                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3648
3649         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3650                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3651                 return;
3652         }
3653
3654         adapter->ip_offload_ctrl.len =
3655             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3656         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3657         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3658         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3659         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3660         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3661         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3662         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3663         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3664         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3665
3666         /* large_rx disabled for now, additional features needed */
3667         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3668         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3669
3670         adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3671
3672         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3673                 adapter->netdev->features |= NETIF_F_IP_CSUM;
3674
3675         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3676                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3677
3678         if ((adapter->netdev->features &
3679             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3680                 adapter->netdev->features |= NETIF_F_RXCSUM;
3681
3682         if (buf->large_tx_ipv4)
3683                 adapter->netdev->features |= NETIF_F_TSO;
3684         if (buf->large_tx_ipv6)
3685                 adapter->netdev->features |= NETIF_F_TSO6;
3686
3687         adapter->netdev->hw_features |= adapter->netdev->features;
3688
3689         memset(&crq, 0, sizeof(crq));
3690         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3691         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3692         crq.control_ip_offload.len =
3693             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3694         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3695         ibmvnic_send_crq(adapter, &crq);
3696 }
3697
3698 static void handle_error_info_rsp(union ibmvnic_crq *crq,
3699                                   struct ibmvnic_adapter *adapter)
3700 {
3701         struct device *dev = &adapter->vdev->dev;
3702         struct ibmvnic_error_buff *error_buff, *tmp;
3703         unsigned long flags;
3704         bool found = false;
3705         int i;
3706
3707         if (!crq->request_error_rsp.rc.code) {
3708                 dev_info(dev, "Request Error Rsp returned with rc=%x\n",
3709                          crq->request_error_rsp.rc.code);
3710                 return;
3711         }
3712
3713         spin_lock_irqsave(&adapter->error_list_lock, flags);
3714         list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
3715                 if (error_buff->error_id == crq->request_error_rsp.error_id) {
3716                         found = true;
3717                         list_del(&error_buff->list);
3718                         break;
3719                 }
3720         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3721
3722         if (!found) {
3723                 dev_err(dev, "Couldn't find error id %x\n",
3724                         be32_to_cpu(crq->request_error_rsp.error_id));
3725                 return;
3726         }
3727
3728         dev_err(dev, "Detailed info for error id %x:",
3729                 be32_to_cpu(crq->request_error_rsp.error_id));
3730
3731         for (i = 0; i < error_buff->len; i++) {
3732                 pr_cont("%02x", (int)error_buff->buff[i]);
3733                 if (i % 8 == 7)
3734                         pr_cont(" ");
3735         }
3736         pr_cont("\n");
3737
3738         dma_unmap_single(dev, error_buff->dma, error_buff->len,
3739                          DMA_FROM_DEVICE);
3740         kfree(error_buff->buff);
3741         kfree(error_buff);
3742 }
3743
3744 static void request_error_information(struct ibmvnic_adapter *adapter,
3745                                       union ibmvnic_crq *err_crq)
3746 {
3747         struct device *dev = &adapter->vdev->dev;
3748         struct net_device *netdev = adapter->netdev;
3749         struct ibmvnic_error_buff *error_buff;
3750         unsigned long timeout = msecs_to_jiffies(30000);
3751         union ibmvnic_crq crq;
3752         unsigned long flags;
3753         int rc, detail_len;
3754
3755         error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3756         if (!error_buff)
3757                 return;
3758
3759         detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3760         error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3761         if (!error_buff->buff) {
3762                 kfree(error_buff);
3763                 return;
3764         }
3765
3766         error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3767                                          DMA_FROM_DEVICE);
3768         if (dma_mapping_error(dev, error_buff->dma)) {
3769                 netdev_err(netdev, "Couldn't map error buffer\n");
3770                 kfree(error_buff->buff);
3771                 kfree(error_buff);
3772                 return;
3773         }
3774
3775         error_buff->len = detail_len;
3776         error_buff->error_id = err_crq->error_indication.error_id;
3777
3778         spin_lock_irqsave(&adapter->error_list_lock, flags);
3779         list_add_tail(&error_buff->list, &adapter->errors);
3780         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3781
3782         memset(&crq, 0, sizeof(crq));
3783         crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3784         crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3785         crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3786         crq.request_error_info.len = cpu_to_be32(detail_len);
3787         crq.request_error_info.error_id = err_crq->error_indication.error_id;
3788
3789         rc = ibmvnic_send_crq(adapter, &crq);
3790         if (rc) {
3791                 netdev_err(netdev, "failed to request error information\n");
3792                 goto err_info_fail;
3793         }
3794
3795         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3796                 netdev_err(netdev, "timeout waiting for error information\n");
3797                 goto err_info_fail;
3798         }
3799
3800         return;
3801
3802 err_info_fail:
3803         spin_lock_irqsave(&adapter->error_list_lock, flags);
3804         list_del(&error_buff->list);
3805         spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3806
3807         kfree(error_buff->buff);
3808         kfree(error_buff);
3809 }
3810
3811 static void handle_error_indication(union ibmvnic_crq *crq,
3812                                     struct ibmvnic_adapter *adapter)
3813 {
3814         struct device *dev = &adapter->vdev->dev;
3815
3816         dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3817                 crq->error_indication.flags
3818                         & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3819                 be32_to_cpu(crq->error_indication.error_id),
3820                 be16_to_cpu(crq->error_indication.error_cause));
3821
3822         if (be32_to_cpu(crq->error_indication.error_id))
3823                 request_error_information(adapter, crq);
3824
3825         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3826                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3827         else
3828                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3829 }
3830
3831 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3832                                  struct ibmvnic_adapter *adapter)
3833 {
3834         struct net_device *netdev = adapter->netdev;
3835         struct device *dev = &adapter->vdev->dev;
3836         long rc;
3837
3838         rc = crq->change_mac_addr_rsp.rc.code;
3839         if (rc) {
3840                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3841                 goto out;
3842         }
3843         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3844                ETH_ALEN);
3845 out:
3846         complete(&adapter->fw_done);
3847         return rc;
3848 }
3849
3850 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3851                                    struct ibmvnic_adapter *adapter)
3852 {
3853         struct device *dev = &adapter->vdev->dev;
3854         u64 *req_value;
3855         char *name;
3856
3857         atomic_dec(&adapter->running_cap_crqs);
3858         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3859         case REQ_TX_QUEUES:
3860                 req_value = &adapter->req_tx_queues;
3861                 name = "tx";
3862                 break;
3863         case REQ_RX_QUEUES:
3864                 req_value = &adapter->req_rx_queues;
3865                 name = "rx";
3866                 break;
3867         case REQ_RX_ADD_QUEUES:
3868                 req_value = &adapter->req_rx_add_queues;
3869                 name = "rx_add";
3870                 break;
3871         case REQ_TX_ENTRIES_PER_SUBCRQ:
3872                 req_value = &adapter->req_tx_entries_per_subcrq;
3873                 name = "tx_entries_per_subcrq";
3874                 break;
3875         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3876                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3877                 name = "rx_add_entries_per_subcrq";
3878                 break;
3879         case REQ_MTU:
3880                 req_value = &adapter->req_mtu;
3881                 name = "mtu";
3882                 break;
3883         case PROMISC_REQUESTED:
3884                 req_value = &adapter->promisc;
3885                 name = "promisc";
3886                 break;
3887         default:
3888                 dev_err(dev, "Got invalid cap request rsp %d\n",
3889                         crq->request_capability.capability);
3890                 return;
3891         }
3892
3893         switch (crq->request_capability_rsp.rc.code) {
3894         case SUCCESS:
3895                 break;
3896         case PARTIALSUCCESS:
3897                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3898                          *req_value,
3899                          (long int)be64_to_cpu(crq->request_capability_rsp.
3900                                                number), name);
3901
3902                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3903                     REQ_MTU) {
3904                         pr_err("mtu of %llu is not supported. Reverting.\n",
3905                                *req_value);
3906                         *req_value = adapter->fallback.mtu;
3907                 } else {
3908                         *req_value =
3909                                 be64_to_cpu(crq->request_capability_rsp.number);
3910                 }
3911
3912                 ibmvnic_send_req_caps(adapter, 1);
3913                 return;
3914         default:
3915                 dev_err(dev, "Error %d in request cap rsp\n",
3916                         crq->request_capability_rsp.rc.code);
3917                 return;
3918         }
3919
3920         /* Done receiving requested capabilities, query IP offload support */
3921         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3922                 union ibmvnic_crq newcrq;
3923                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3924                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3925                     &adapter->ip_offload_buf;
3926
3927                 adapter->wait_capability = false;
3928                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3929                                                          buf_sz,
3930                                                          DMA_FROM_DEVICE);
3931
3932                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3933                         if (!firmware_has_feature(FW_FEATURE_CMO))
3934                                 dev_err(dev, "Couldn't map offload buffer\n");
3935                         return;
3936                 }
3937
3938                 memset(&newcrq, 0, sizeof(newcrq));
3939                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3940                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3941                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3942                 newcrq.query_ip_offload.ioba =
3943                     cpu_to_be32(adapter->ip_offload_tok);
3944
3945                 ibmvnic_send_crq(adapter, &newcrq);
3946         }
3947 }
3948
3949 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3950                             struct ibmvnic_adapter *adapter)
3951 {
3952         struct device *dev = &adapter->vdev->dev;
3953         struct net_device *netdev = adapter->netdev;
3954         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3955         struct ibmvnic_login_buffer *login = adapter->login_buf;
3956         int i;
3957
3958         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3959                          DMA_TO_DEVICE);
3960         dma_unmap_single(dev, adapter->login_rsp_buf_token,
3961                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
3962
3963         /* If the number of queues requested can't be allocated by the
3964          * server, the login response will return with code 1. We will need
3965          * to resend the login buffer with fewer queues requested.
3966          */
3967         if (login_rsp_crq->generic.rc.code) {
3968                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
3969                 complete(&adapter->init_done);
3970                 return 0;
3971         }
3972
3973         netdev->mtu = adapter->req_mtu - ETH_HLEN;
3974
3975         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3976         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3977                 netdev_dbg(adapter->netdev, "%016lx\n",
3978                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3979         }
3980
3981         /* Sanity checks */
3982         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3983             (be32_to_cpu(login->num_rxcomp_subcrqs) *
3984              adapter->req_rx_add_queues !=
3985              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3986                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3987                 ibmvnic_remove(adapter->vdev);
3988                 return -EIO;
3989         }
3990         release_login_buffer(adapter);
3991         complete(&adapter->init_done);
3992
3993         return 0;
3994 }
3995
3996 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3997                                      struct ibmvnic_adapter *adapter)
3998 {
3999         struct device *dev = &adapter->vdev->dev;
4000         long rc;
4001
4002         rc = crq->request_unmap_rsp.rc.code;
4003         if (rc)
4004                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4005 }
4006
4007 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4008                                  struct ibmvnic_adapter *adapter)
4009 {
4010         struct net_device *netdev = adapter->netdev;
4011         struct device *dev = &adapter->vdev->dev;
4012         long rc;
4013
4014         rc = crq->query_map_rsp.rc.code;
4015         if (rc) {
4016                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4017                 return;
4018         }
4019         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4020                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4021                    crq->query_map_rsp.free_pages);
4022 }
4023
4024 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4025                                  struct ibmvnic_adapter *adapter)
4026 {
4027         struct net_device *netdev = adapter->netdev;
4028         struct device *dev = &adapter->vdev->dev;
4029         long rc;
4030
4031         atomic_dec(&adapter->running_cap_crqs);
4032         netdev_dbg(netdev, "Outstanding queries: %d\n",
4033                    atomic_read(&adapter->running_cap_crqs));
4034         rc = crq->query_capability.rc.code;
4035         if (rc) {
4036                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4037                 goto out;
4038         }
4039
4040         switch (be16_to_cpu(crq->query_capability.capability)) {
4041         case MIN_TX_QUEUES:
4042                 adapter->min_tx_queues =
4043                     be64_to_cpu(crq->query_capability.number);
4044                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4045                            adapter->min_tx_queues);
4046                 break;
4047         case MIN_RX_QUEUES:
4048                 adapter->min_rx_queues =
4049                     be64_to_cpu(crq->query_capability.number);
4050                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4051                            adapter->min_rx_queues);
4052                 break;
4053         case MIN_RX_ADD_QUEUES:
4054                 adapter->min_rx_add_queues =
4055                     be64_to_cpu(crq->query_capability.number);
4056                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4057                            adapter->min_rx_add_queues);
4058                 break;
4059         case MAX_TX_QUEUES:
4060                 adapter->max_tx_queues =
4061                     be64_to_cpu(crq->query_capability.number);
4062                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4063                            adapter->max_tx_queues);
4064                 break;
4065         case MAX_RX_QUEUES:
4066                 adapter->max_rx_queues =
4067                     be64_to_cpu(crq->query_capability.number);
4068                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4069                            adapter->max_rx_queues);
4070                 break;
4071         case MAX_RX_ADD_QUEUES:
4072                 adapter->max_rx_add_queues =
4073                     be64_to_cpu(crq->query_capability.number);
4074                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4075                            adapter->max_rx_add_queues);
4076                 break;
4077         case MIN_TX_ENTRIES_PER_SUBCRQ:
4078                 adapter->min_tx_entries_per_subcrq =
4079                     be64_to_cpu(crq->query_capability.number);
4080                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4081                            adapter->min_tx_entries_per_subcrq);
4082                 break;
4083         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4084                 adapter->min_rx_add_entries_per_subcrq =
4085                     be64_to_cpu(crq->query_capability.number);
4086                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4087                            adapter->min_rx_add_entries_per_subcrq);
4088                 break;
4089         case MAX_TX_ENTRIES_PER_SUBCRQ:
4090                 adapter->max_tx_entries_per_subcrq =
4091                     be64_to_cpu(crq->query_capability.number);
4092                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4093                            adapter->max_tx_entries_per_subcrq);
4094                 break;
4095         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4096                 adapter->max_rx_add_entries_per_subcrq =
4097                     be64_to_cpu(crq->query_capability.number);
4098                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4099                            adapter->max_rx_add_entries_per_subcrq);
4100                 break;
4101         case TCP_IP_OFFLOAD:
4102                 adapter->tcp_ip_offload =
4103                     be64_to_cpu(crq->query_capability.number);
4104                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4105                            adapter->tcp_ip_offload);
4106                 break;
4107         case PROMISC_SUPPORTED:
4108                 adapter->promisc_supported =
4109                     be64_to_cpu(crq->query_capability.number);
4110                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4111                            adapter->promisc_supported);
4112                 break;
4113         case MIN_MTU:
4114                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4115                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4116                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4117                 break;
4118         case MAX_MTU:
4119                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4120                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4121                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4122                 break;
4123         case MAX_MULTICAST_FILTERS:
4124                 adapter->max_multicast_filters =
4125                     be64_to_cpu(crq->query_capability.number);
4126                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4127                            adapter->max_multicast_filters);
4128                 break;
4129         case VLAN_HEADER_INSERTION:
4130                 adapter->vlan_header_insertion =
4131                     be64_to_cpu(crq->query_capability.number);
4132                 if (adapter->vlan_header_insertion)
4133                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4134                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4135                            adapter->vlan_header_insertion);
4136                 break;
4137         case RX_VLAN_HEADER_INSERTION:
4138                 adapter->rx_vlan_header_insertion =
4139                     be64_to_cpu(crq->query_capability.number);
4140                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4141                            adapter->rx_vlan_header_insertion);
4142                 break;
4143         case MAX_TX_SG_ENTRIES:
4144                 adapter->max_tx_sg_entries =
4145                     be64_to_cpu(crq->query_capability.number);
4146                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4147                            adapter->max_tx_sg_entries);
4148                 break;
4149         case RX_SG_SUPPORTED:
4150                 adapter->rx_sg_supported =
4151                     be64_to_cpu(crq->query_capability.number);
4152                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4153                            adapter->rx_sg_supported);
4154                 break;
4155         case OPT_TX_COMP_SUB_QUEUES:
4156                 adapter->opt_tx_comp_sub_queues =
4157                     be64_to_cpu(crq->query_capability.number);
4158                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4159                            adapter->opt_tx_comp_sub_queues);
4160                 break;
4161         case OPT_RX_COMP_QUEUES:
4162                 adapter->opt_rx_comp_queues =
4163                     be64_to_cpu(crq->query_capability.number);
4164                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4165                            adapter->opt_rx_comp_queues);
4166                 break;
4167         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4168                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4169                     be64_to_cpu(crq->query_capability.number);
4170                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4171                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4172                 break;
4173         case OPT_TX_ENTRIES_PER_SUBCRQ:
4174                 adapter->opt_tx_entries_per_subcrq =
4175                     be64_to_cpu(crq->query_capability.number);
4176                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4177                            adapter->opt_tx_entries_per_subcrq);
4178                 break;
4179         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4180                 adapter->opt_rxba_entries_per_subcrq =
4181                     be64_to_cpu(crq->query_capability.number);
4182                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4183                            adapter->opt_rxba_entries_per_subcrq);
4184                 break;
4185         case TX_RX_DESC_REQ:
4186                 adapter->tx_rx_desc_req = crq->query_capability.number;
4187                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4188                            adapter->tx_rx_desc_req);
4189                 break;
4190
4191         default:
4192                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4193                            crq->query_capability.capability);
4194         }
4195
4196 out:
4197         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4198                 adapter->wait_capability = false;
4199                 ibmvnic_send_req_caps(adapter, 0);
4200         }
4201 }
4202
4203 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4204                                struct ibmvnic_adapter *adapter)
4205 {
4206         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4207         struct net_device *netdev = adapter->netdev;
4208         struct device *dev = &adapter->vdev->dev;
4209         u64 *u64_crq = (u64 *)crq;
4210         long rc;
4211
4212         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4213                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4214                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4215         switch (gen_crq->first) {
4216         case IBMVNIC_CRQ_INIT_RSP:
4217                 switch (gen_crq->cmd) {
4218                 case IBMVNIC_CRQ_INIT:
4219                         dev_info(dev, "Partner initialized\n");
4220                         adapter->from_passive_init = true;
4221                         adapter->failover_pending = false;
4222                         complete(&adapter->init_done);
4223                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4224                         break;
4225                 case IBMVNIC_CRQ_INIT_COMPLETE:
4226                         dev_info(dev, "Partner initialization complete\n");
4227                         send_version_xchg(adapter);
4228                         break;
4229                 default:
4230                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4231                 }
4232                 return;
4233         case IBMVNIC_CRQ_XPORT_EVENT:
4234                 netif_carrier_off(netdev);
4235                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4236                         dev_info(dev, "Migrated, re-enabling adapter\n");
4237                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4238                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4239                         dev_info(dev, "Backing device failover detected\n");
4240                         adapter->failover_pending = true;
4241                 } else {
4242                         /* The adapter lost the connection */
4243                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4244                                 gen_crq->cmd);
4245                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4246                 }
4247                 return;
4248         case IBMVNIC_CRQ_CMD_RSP:
4249                 break;
4250         default:
4251                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4252                         gen_crq->first);
4253                 return;
4254         }
4255
4256         switch (gen_crq->cmd) {
4257         case VERSION_EXCHANGE_RSP:
4258                 rc = crq->version_exchange_rsp.rc.code;
4259                 if (rc) {
4260                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4261                         break;
4262                 }
4263                 dev_info(dev, "Partner protocol version is %d\n",
4264                          crq->version_exchange_rsp.version);
4265                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4266                     ibmvnic_version)
4267                         ibmvnic_version =
4268                             be16_to_cpu(crq->version_exchange_rsp.version);
4269                 send_cap_queries(adapter);
4270                 break;
4271         case QUERY_CAPABILITY_RSP:
4272                 handle_query_cap_rsp(crq, adapter);
4273                 break;
4274         case QUERY_MAP_RSP:
4275                 handle_query_map_rsp(crq, adapter);
4276                 break;
4277         case REQUEST_MAP_RSP:
4278                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4279                 complete(&adapter->fw_done);
4280                 break;
4281         case REQUEST_UNMAP_RSP:
4282                 handle_request_unmap_rsp(crq, adapter);
4283                 break;
4284         case REQUEST_CAPABILITY_RSP:
4285                 handle_request_cap_rsp(crq, adapter);
4286                 break;
4287         case LOGIN_RSP:
4288                 netdev_dbg(netdev, "Got Login Response\n");
4289                 handle_login_rsp(crq, adapter);
4290                 break;
4291         case LOGICAL_LINK_STATE_RSP:
4292                 netdev_dbg(netdev,
4293                            "Got Logical Link State Response, state: %d rc: %d\n",
4294                            crq->logical_link_state_rsp.link_state,
4295                            crq->logical_link_state_rsp.rc.code);
4296                 adapter->logical_link_state =
4297                     crq->logical_link_state_rsp.link_state;
4298                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4299                 complete(&adapter->init_done);
4300                 break;
4301         case LINK_STATE_INDICATION:
4302                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4303                 adapter->phys_link_state =
4304                     crq->link_state_indication.phys_link_state;
4305                 adapter->logical_link_state =
4306                     crq->link_state_indication.logical_link_state;
4307                 break;
4308         case CHANGE_MAC_ADDR_RSP:
4309                 netdev_dbg(netdev, "Got MAC address change Response\n");
4310                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4311                 break;
4312         case ERROR_INDICATION:
4313                 netdev_dbg(netdev, "Got Error Indication\n");
4314                 handle_error_indication(crq, adapter);
4315                 break;
4316         case REQUEST_ERROR_RSP:
4317                 netdev_dbg(netdev, "Got Error Detail Response\n");
4318                 handle_error_info_rsp(crq, adapter);
4319                 break;
4320         case REQUEST_STATISTICS_RSP:
4321                 netdev_dbg(netdev, "Got Statistics Response\n");
4322                 complete(&adapter->stats_done);
4323                 break;
4324         case QUERY_IP_OFFLOAD_RSP:
4325                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4326                 handle_query_ip_offload_rsp(adapter);
4327                 break;
4328         case MULTICAST_CTRL_RSP:
4329                 netdev_dbg(netdev, "Got multicast control Response\n");
4330                 break;
4331         case CONTROL_IP_OFFLOAD_RSP:
4332                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4333                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4334                                  sizeof(adapter->ip_offload_ctrl),
4335                                  DMA_TO_DEVICE);
4336                 complete(&adapter->init_done);
4337                 break;
4338         case COLLECT_FW_TRACE_RSP:
4339                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4340                 complete(&adapter->fw_done);
4341                 break;
4342         case GET_VPD_SIZE_RSP:
4343                 handle_vpd_size_rsp(crq, adapter);
4344                 break;
4345         case GET_VPD_RSP:
4346                 handle_vpd_rsp(crq, adapter);
4347                 break;
4348         default:
4349                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4350                            gen_crq->cmd);
4351         }
4352 }
4353
4354 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4355 {
4356         struct ibmvnic_adapter *adapter = instance;
4357
4358         tasklet_schedule(&adapter->tasklet);
4359         return IRQ_HANDLED;
4360 }
4361
4362 static void ibmvnic_tasklet(void *data)
4363 {
4364         struct ibmvnic_adapter *adapter = data;
4365         struct ibmvnic_crq_queue *queue = &adapter->crq;
4366         union ibmvnic_crq *crq;
4367         unsigned long flags;
4368         bool done = false;
4369
4370         spin_lock_irqsave(&queue->lock, flags);
4371         while (!done) {
4372                 /* Pull all the valid messages off the CRQ */
4373                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4374                         ibmvnic_handle_crq(crq, adapter);
4375                         crq->generic.first = 0;
4376                 }
4377
4378                 /* remain in tasklet until all
4379                  * capabilities responses are received
4380                  */
4381                 if (!adapter->wait_capability)
4382                         done = true;
4383         }
4384         /* if capabilities CRQ's were sent in this tasklet, the following
4385          * tasklet must wait until all responses are received
4386          */
4387         if (atomic_read(&adapter->running_cap_crqs) != 0)
4388                 adapter->wait_capability = true;
4389         spin_unlock_irqrestore(&queue->lock, flags);
4390 }
4391
4392 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4393 {
4394         struct vio_dev *vdev = adapter->vdev;
4395         int rc;
4396
4397         do {
4398                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4399         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4400
4401         if (rc)
4402                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4403
4404         return rc;
4405 }
4406
4407 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4408 {
4409         struct ibmvnic_crq_queue *crq = &adapter->crq;
4410         struct device *dev = &adapter->vdev->dev;
4411         struct vio_dev *vdev = adapter->vdev;
4412         int rc;
4413
4414         /* Close the CRQ */
4415         do {
4416                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4417         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4418
4419         /* Clean out the queue */
4420         memset(crq->msgs, 0, PAGE_SIZE);
4421         crq->cur = 0;
4422
4423         /* And re-open it again */
4424         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4425                                 crq->msg_token, PAGE_SIZE);
4426
4427         if (rc == H_CLOSED)
4428                 /* Adapter is good, but other end is not ready */
4429                 dev_warn(dev, "Partner adapter not ready\n");
4430         else if (rc != 0)
4431                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4432
4433         return rc;
4434 }
4435
4436 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4437 {
4438         struct ibmvnic_crq_queue *crq = &adapter->crq;
4439         struct vio_dev *vdev = adapter->vdev;
4440         long rc;
4441
4442         if (!crq->msgs)
4443                 return;
4444
4445         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4446         free_irq(vdev->irq, adapter);
4447         tasklet_kill(&adapter->tasklet);
4448         do {
4449                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4450         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4451
4452         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4453                          DMA_BIDIRECTIONAL);
4454         free_page((unsigned long)crq->msgs);
4455         crq->msgs = NULL;
4456 }
4457
4458 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4459 {
4460         struct ibmvnic_crq_queue *crq = &adapter->crq;
4461         struct device *dev = &adapter->vdev->dev;
4462         struct vio_dev *vdev = adapter->vdev;
4463         int rc, retrc = -ENOMEM;
4464
4465         if (crq->msgs)
4466                 return 0;
4467
4468         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4469         /* Should we allocate more than one page? */
4470
4471         if (!crq->msgs)
4472                 return -ENOMEM;
4473
4474         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4475         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4476                                         DMA_BIDIRECTIONAL);
4477         if (dma_mapping_error(dev, crq->msg_token))
4478                 goto map_failed;
4479
4480         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4481                                 crq->msg_token, PAGE_SIZE);
4482
4483         if (rc == H_RESOURCE)
4484                 /* maybe kexecing and resource is busy. try a reset */
4485                 rc = ibmvnic_reset_crq(adapter);
4486         retrc = rc;
4487
4488         if (rc == H_CLOSED) {
4489                 dev_warn(dev, "Partner adapter not ready\n");
4490         } else if (rc) {
4491                 dev_warn(dev, "Error %d opening adapter\n", rc);
4492                 goto reg_crq_failed;
4493         }
4494
4495         retrc = 0;
4496
4497         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4498                      (unsigned long)adapter);
4499
4500         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4501         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4502                          adapter);
4503         if (rc) {
4504                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4505                         vdev->irq, rc);
4506                 goto req_irq_failed;
4507         }
4508
4509         rc = vio_enable_interrupts(vdev);
4510         if (rc) {
4511                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4512                 goto req_irq_failed;
4513         }
4514
4515         crq->cur = 0;
4516         spin_lock_init(&crq->lock);
4517
4518         return retrc;
4519
4520 req_irq_failed:
4521         tasklet_kill(&adapter->tasklet);
4522         do {
4523                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4524         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4525 reg_crq_failed:
4526         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4527 map_failed:
4528         free_page((unsigned long)crq->msgs);
4529         crq->msgs = NULL;
4530         return retrc;
4531 }
4532
4533 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4534 {
4535         struct device *dev = &adapter->vdev->dev;
4536         unsigned long timeout = msecs_to_jiffies(30000);
4537         u64 old_num_rx_queues, old_num_tx_queues;
4538         int rc;
4539
4540         adapter->from_passive_init = false;
4541
4542         old_num_rx_queues = adapter->req_rx_queues;
4543         old_num_tx_queues = adapter->req_tx_queues;
4544
4545         init_completion(&adapter->init_done);
4546         adapter->init_done_rc = 0;
4547         ibmvnic_send_crq_init(adapter);
4548         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4549                 dev_err(dev, "Initialization sequence timed out\n");
4550                 return -1;
4551         }
4552
4553         if (adapter->init_done_rc) {
4554                 release_crq_queue(adapter);
4555                 return adapter->init_done_rc;
4556         }
4557
4558         if (adapter->from_passive_init) {
4559                 adapter->state = VNIC_OPEN;
4560                 adapter->from_passive_init = false;
4561                 return -1;
4562         }
4563
4564         if (adapter->resetting && !adapter->wait_for_reset &&
4565             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4566                 if (adapter->req_rx_queues != old_num_rx_queues ||
4567                     adapter->req_tx_queues != old_num_tx_queues) {
4568                         release_sub_crqs(adapter, 0);
4569                         rc = init_sub_crqs(adapter);
4570                 } else {
4571                         rc = reset_sub_crq_queues(adapter);
4572                 }
4573         } else {
4574                 rc = init_sub_crqs(adapter);
4575         }
4576
4577         if (rc) {
4578                 dev_err(dev, "Initialization of sub crqs failed\n");
4579                 release_crq_queue(adapter);
4580                 return rc;
4581         }
4582
4583         rc = init_sub_crq_irqs(adapter);
4584         if (rc) {
4585                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4586                 release_crq_queue(adapter);
4587         }
4588
4589         rc = init_stats_buffers(adapter);
4590         if (rc)
4591                 return rc;
4592
4593         rc = init_stats_token(adapter);
4594         if (rc)
4595                 return rc;
4596
4597         return rc;
4598 }
4599
4600 static struct device_attribute dev_attr_failover;
4601
4602 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4603 {
4604         struct ibmvnic_adapter *adapter;
4605         struct net_device *netdev;
4606         unsigned char *mac_addr_p;
4607         int rc;
4608
4609         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4610                 dev->unit_address);
4611
4612         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4613                                                         VETH_MAC_ADDR, NULL);
4614         if (!mac_addr_p) {
4615                 dev_err(&dev->dev,
4616                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4617                         __FILE__, __LINE__);
4618                 return 0;
4619         }
4620
4621         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4622                                    IBMVNIC_MAX_QUEUES);
4623         if (!netdev)
4624                 return -ENOMEM;
4625
4626         adapter = netdev_priv(netdev);
4627         adapter->state = VNIC_PROBING;
4628         dev_set_drvdata(&dev->dev, netdev);
4629         adapter->vdev = dev;
4630         adapter->netdev = netdev;
4631
4632         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4633         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4634         netdev->irq = dev->irq;
4635         netdev->netdev_ops = &ibmvnic_netdev_ops;
4636         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4637         SET_NETDEV_DEV(netdev, &dev->dev);
4638
4639         spin_lock_init(&adapter->stats_lock);
4640
4641         INIT_LIST_HEAD(&adapter->errors);
4642         spin_lock_init(&adapter->error_list_lock);
4643
4644         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4645         INIT_LIST_HEAD(&adapter->rwi_list);
4646         mutex_init(&adapter->reset_lock);
4647         mutex_init(&adapter->rwi_lock);
4648         adapter->resetting = false;
4649
4650         adapter->mac_change_pending = false;
4651
4652         do {
4653                 rc = init_crq_queue(adapter);
4654                 if (rc) {
4655                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4656                                 rc);
4657                         goto ibmvnic_init_fail;
4658                 }
4659
4660                 rc = ibmvnic_init(adapter);
4661                 if (rc && rc != EAGAIN)
4662                         goto ibmvnic_init_fail;
4663         } while (rc == EAGAIN);
4664
4665         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4666         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4667         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4668
4669         rc = device_create_file(&dev->dev, &dev_attr_failover);
4670         if (rc)
4671                 goto ibmvnic_init_fail;
4672
4673         netif_carrier_off(netdev);
4674         rc = register_netdev(netdev);
4675         if (rc) {
4676                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4677                 goto ibmvnic_register_fail;
4678         }
4679         dev_info(&dev->dev, "ibmvnic registered\n");
4680
4681         adapter->state = VNIC_PROBED;
4682
4683         adapter->wait_for_reset = false;
4684
4685         return 0;
4686
4687 ibmvnic_register_fail:
4688         device_remove_file(&dev->dev, &dev_attr_failover);
4689
4690 ibmvnic_init_fail:
4691         release_sub_crqs(adapter, 1);
4692         release_crq_queue(adapter);
4693         free_netdev(netdev);
4694
4695         return rc;
4696 }
4697
4698 static int ibmvnic_remove(struct vio_dev *dev)
4699 {
4700         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4701         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4702
4703         adapter->state = VNIC_REMOVING;
4704         unregister_netdev(netdev);
4705         mutex_lock(&adapter->reset_lock);
4706
4707         release_resources(adapter);
4708         release_sub_crqs(adapter, 1);
4709         release_crq_queue(adapter);
4710
4711         release_stats_token(adapter);
4712         release_stats_buffers(adapter);
4713
4714         adapter->state = VNIC_REMOVED;
4715
4716         mutex_unlock(&adapter->reset_lock);
4717         device_remove_file(&dev->dev, &dev_attr_failover);
4718         free_netdev(netdev);
4719         dev_set_drvdata(&dev->dev, NULL);
4720
4721         return 0;
4722 }
4723
4724 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4725                               const char *buf, size_t count)
4726 {
4727         struct net_device *netdev = dev_get_drvdata(dev);
4728         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4729         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4730         __be64 session_token;
4731         long rc;
4732
4733         if (!sysfs_streq(buf, "1"))
4734                 return -EINVAL;
4735
4736         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4737                          H_GET_SESSION_TOKEN, 0, 0, 0);
4738         if (rc) {
4739                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4740                            rc);
4741                 return -EINVAL;
4742         }
4743
4744         session_token = (__be64)retbuf[0];
4745         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4746                    be64_to_cpu(session_token));
4747         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4748                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4749         if (rc) {
4750                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4751                            rc);
4752                 return -EINVAL;
4753         }
4754
4755         return count;
4756 }
4757
4758 static DEVICE_ATTR_WO(failover);
4759
4760 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4761 {
4762         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4763         struct ibmvnic_adapter *adapter;
4764         struct iommu_table *tbl;
4765         unsigned long ret = 0;
4766         int i;
4767
4768         tbl = get_iommu_table_base(&vdev->dev);
4769
4770         /* netdev inits at probe time along with the structures we need below*/
4771         if (!netdev)
4772                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4773
4774         adapter = netdev_priv(netdev);
4775
4776         ret += PAGE_SIZE; /* the crq message queue */
4777         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4778
4779         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4780                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4781
4782         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4783              i++)
4784                 ret += adapter->rx_pool[i].size *
4785                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4786
4787         return ret;
4788 }
4789
4790 static int ibmvnic_resume(struct device *dev)
4791 {
4792         struct net_device *netdev = dev_get_drvdata(dev);
4793         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4794
4795         if (adapter->state != VNIC_OPEN)
4796                 return 0;
4797
4798         tasklet_schedule(&adapter->tasklet);
4799
4800         return 0;
4801 }
4802
4803 static const struct vio_device_id ibmvnic_device_table[] = {
4804         {"network", "IBM,vnic"},
4805         {"", "" }
4806 };
4807 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4808
4809 static const struct dev_pm_ops ibmvnic_pm_ops = {
4810         .resume = ibmvnic_resume
4811 };
4812
4813 static struct vio_driver ibmvnic_driver = {
4814         .id_table       = ibmvnic_device_table,
4815         .probe          = ibmvnic_probe,
4816         .remove         = ibmvnic_remove,
4817         .get_desired_dma = ibmvnic_get_desired_dma,
4818         .name           = ibmvnic_driver_name,
4819         .pm             = &ibmvnic_pm_ops,
4820 };
4821
4822 /* module functions */
4823 static int __init ibmvnic_module_init(void)
4824 {
4825         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4826                 IBMVNIC_DRIVER_VERSION);
4827
4828         return vio_register_driver(&ibmvnic_driver);
4829 }
4830
4831 static void __exit ibmvnic_module_exit(void)
4832 {
4833         vio_unregister_driver(&ibmvnic_driver);
4834 }
4835
4836 module_init(ibmvnic_module_init);
4837 module_exit(ibmvnic_module_exit);